Abstract
Over the past several decades, numerous large cohort studies have attempted to quantify the protective effect of physical activity on cardiovascular and all-cause mortality. The aim of the authors’ review was to provide an up-to-date overview of the study results.
In a systematic MEDLINE search conducted in May 2007, the authors included cohort studies that assessed the primary preventive impact of physical activity on all-cause and cardiovascular mortality. The authors reported risk reductions on the basis of comparison between the least active and the most active population subgroups, with the least active population subgroup as the reference group. Random-effect models were used for meta-analysis.
A total of 33 studies with 883372 participants were included. Follow-up ranged from 4 years to over 20 years. The majority of studies reported significant risk reductions for physically active participants. Concerning cardiovascular mortality, physical activity was associated with a risk reduction of 35% (95% confidence interval, 30–40%). All-cause mortality was reduced by 33% (95% confidence interval, 28–37%). Studies that used patient questionnaires to assess physical activity reported lower risk reductions than studies that used more objective measures of fitness.
Physical activity is associated with a marked decrease in cardiovascular and all-cause mortality in both men and women, even after adjusting for other relevant risk factors.
Introduction
Physical inactivity increases the risk of developing a variety of diseases, including overweight, diabetes, hypertension, osteoporosis, and depression. Moreover, inactivity has been associated with higher all-cause and cardiovascular mortality [1,2]. Numerous large cohort studies assessed self-reported physical activity, objective measures of physical fitness, activities of daily living, and systematic exercise training to determine the risks associated with a sedentary lifestyle in men and women [1]. According to a study by Mokdad et al. [3], physical inactivity, along with smoking and malnutrition, is one of the most important lifestyle-related risk factors.
Although most studies have reported a significant relative reduction in mortality for physically active participants, the range of benefit has varied considerably. For example, Myers et al. [4] reported a reduction in all-cause mortality of 72% between their most and least fit male participants during 6 years of follow-up, whereas Lee et al. [5] found a risk reduction of only 13%. Risk reduction may also vary depending on adjustment for important covariables such as blood pressure or diabetes. However the extent to which the risk reduction achieved through physical activity is attenuated by adjusting for other important risk factors remains unclear [6,7].
Study characteristics
| Study | Population | Sex | Age | Physical activity assessed by | Follow-up in years (mean) | Results adjusted for |
| Arraiz et al. [8] | 9792 | Men and women | 30–69 | Fitness test | 7 | Age, sex, BMI, smoking |
| Barengo et al. [9] | 32677 | 15853 men; 16824 women | 30–59 | Self-report | 20 | Age, study year, BMI, systolic blood pressure, cholesterol, education, smoking, other physical activity |
| Blair et al. [10]a | 9777 | Men | 20–82 | Fitness test | 5 | Age |
| Blair et al. [11] | 32421 | 25341 men; 7080 women | 20–88 | Fitness test | 8 | Age, smoking, cholesterol, blood pressure, health status |
| Bucksch [12] | 7187 | 3742 men; 3445 women | 30–69 | Self-report | 16 | Age, smoking, BMI, cardiovascular risk factors, alcohol use, diet, chronic disease index, social status |
| Carlsson et al. [13] | 27734 | Women | 51–83 | Self-report | 5 | Age, smoking, education, number of children, hormone replacement, fruit/vegetable intake, BMI, blood pressure, comorbidity |
| Church et al. [14] | 22167 | Men | Mean 43 | Fitness test | 14 | Age, examination year, alcohol use, smoking, family history of CVD, cholesterol, BMI, fasting glucose |
| Davey et al. [15] | 6702 | Men | 40–64 | Self-report | 25 | Age, employment, smoking, BMI, forced expiratory volume |
| Evenson et al. 2004 [16] | 5712 | 3000 men, 2712 women | ≥ 30 | Fitness test | 24, 26 | Age, smoking, education, alcohol use, BMI, race, cholesterol |
| Fang et al. [17] | 9790 | 3730 men, 6060 women | 25–74 | Self-report | 17 | Age, sex, race, family income, education, diabetes, smoking, systolic blood pressure, cholesterol, illnesses affecting eating |
| Gregg et al. [18] | 7553 | Women | ≥ 65 | Self-report | 7 | Age, smoking, BMI, stroke, diabetes, hypertension, self-rated health status |
| Hillsdon et al. [19] | 10522 | Men and women | 35–64 | Self-report | 10 | Age, sex, smoking, alcohol use, health status, social class |
| Hu et al. [20] | 80921 | Women | 30–55 | Self-report | 24 | Age, smoking, family history of CVD, menopausal status, hormone use |
| Hu et al. [7] | 47212 | 22528 men; 24684 women | 25–64 | Self-report | 18 | Age, study year, smoking, systolic blood pressure, cholesterol, BMI, diabetes, education |
| Kaplan et al. [21] | 6131 | Men and women | 16–94, mean 43 | Self-report | 28 | Age, sex, race, education, health status, social isolation |
| Katzmarzyk et al. [22] | 19223 | Men | 20–83, mean 43 | Fitness test | 10 | Age, year of examination, alcohol use, smoking, family history of CVD, BMI |
| Katzmarzyk et al. [23] | 5421 | Women | 20–69 | Self-report | 12 | Age, smoking, alcohol use, waist circumference |
| Khaw et al. [24] | 22191 | 9984 men; 12207 women | 45–79 | Self-report | 8 | Age, BMI, systolic blood pressure, cholesterol, smoking, alcohol use, diabetes, social status |
| Kohl et al. [25] | 8108 | Men | 20–84, mean 42 | Fitness test | 8 | Age, glycemic status, systolic blood pressure, cholesterol, BMI, family history of CVD, smoking |
| Kushi et al. [26] | 33154 | Women | 55–69 | Self-report | 7 | Age, age at menarche/menopause/first live birth, parity, smoking, alcohol use, BMI, waist-to-hip ratio, energy intake, oestrogen use, hypertension, diabetes, education, marital status |
| Lee et al. [5] | 17231 | Men | Mean 46 | Self-report | 22, 26 | Age, smoking, hypertension, diabetes, early parental death |
| Leon et al. [27] | 12138 | Men | 35–57, mean 46 | Self-report | 16 | Age, education, smoking, cholesterol, diastolic blood pressure, BMI |
| Matthews et al. [28] | 67143 | Women | 40–70, mean 52 | Self-report | 6 | Age, marital status, education, income, smoking, alcohol use, number of pregnancies, oral contraceptive use, menopausal status, diabetes, hypertension, respiratory disease, chronic hepatitis |
| Richardson et al. [29] | 9611 | Men and women | 51–61 | Self-report | 8 | Age, sex, race, self-rated health, history of cancer, obesity, CVD risk |
| Rockhill et al. [30] | 80348 | Women | 30–55 | Self-report | 14 | Age, smoking, alcohol use, height, BMI, hormone use |
| Schooling et al. [31] | 56167 | 18759 men; 37417 women | ≥ 65 | Self-report | 4 | Age, sex, education, alcohol use, smoking, income, BMI |
| Stevens et al. [32] | 5366 | 2860 men; 2506 women | Mean 46 | Fitness test | 22, 26 | Age, education, smoking, alcohol use, diet |
| Trolle-Lagerros et al. [33] | 99099 | Women | 30–49 | Self-report | 11 | Age, education, BMI, alcohol use, smoking |
| Vatten et al. [34] | 54248 | 26515 men; 27769 women | ≥ 20 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure, blood pressure medication |
| Wannamethee et al. [35]b | 4311 | Men | 40–59 | Self-report | 4 | Age, smoking, BMI, social class, self-rated health, other physical activity |
| Wei et al. [36] | 25714 normal weight | Men | Mean 44 | Fitness test | 10 | Age, examination year, BMI, family history of CVD, CVD, diabetes, cholesterol, hypertension, smoking |
| Weller and Corey [37] | 6620 | Women | ≥ 30 | Self-report | 7 | Age |
| Wisloff et al. [38] | 56072 | 27143 men; 28929 women | Mean 47 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure |
| Study | Population | Sex | Age | Physical activity assessed by | Follow-up in years (mean) | Results adjusted for |
| Arraiz et al. [8] | 9792 | Men and women | 30–69 | Fitness test | 7 | Age, sex, BMI, smoking |
| Barengo et al. [9] | 32677 | 15853 men; 16824 women | 30–59 | Self-report | 20 | Age, study year, BMI, systolic blood pressure, cholesterol, education, smoking, other physical activity |
| Blair et al. [10]a | 9777 | Men | 20–82 | Fitness test | 5 | Age |
| Blair et al. [11] | 32421 | 25341 men; 7080 women | 20–88 | Fitness test | 8 | Age, smoking, cholesterol, blood pressure, health status |
| Bucksch [12] | 7187 | 3742 men; 3445 women | 30–69 | Self-report | 16 | Age, smoking, BMI, cardiovascular risk factors, alcohol use, diet, chronic disease index, social status |
| Carlsson et al. [13] | 27734 | Women | 51–83 | Self-report | 5 | Age, smoking, education, number of children, hormone replacement, fruit/vegetable intake, BMI, blood pressure, comorbidity |
| Church et al. [14] | 22167 | Men | Mean 43 | Fitness test | 14 | Age, examination year, alcohol use, smoking, family history of CVD, cholesterol, BMI, fasting glucose |
| Davey et al. [15] | 6702 | Men | 40–64 | Self-report | 25 | Age, employment, smoking, BMI, forced expiratory volume |
| Evenson et al. 2004 [16] | 5712 | 3000 men, 2712 women | ≥ 30 | Fitness test | 24, 26 | Age, smoking, education, alcohol use, BMI, race, cholesterol |
| Fang et al. [17] | 9790 | 3730 men, 6060 women | 25–74 | Self-report | 17 | Age, sex, race, family income, education, diabetes, smoking, systolic blood pressure, cholesterol, illnesses affecting eating |
| Gregg et al. [18] | 7553 | Women | ≥ 65 | Self-report | 7 | Age, smoking, BMI, stroke, diabetes, hypertension, self-rated health status |
| Hillsdon et al. [19] | 10522 | Men and women | 35–64 | Self-report | 10 | Age, sex, smoking, alcohol use, health status, social class |
| Hu et al. [20] | 80921 | Women | 30–55 | Self-report | 24 | Age, smoking, family history of CVD, menopausal status, hormone use |
| Hu et al. [7] | 47212 | 22528 men; 24684 women | 25–64 | Self-report | 18 | Age, study year, smoking, systolic blood pressure, cholesterol, BMI, diabetes, education |
| Kaplan et al. [21] | 6131 | Men and women | 16–94, mean 43 | Self-report | 28 | Age, sex, race, education, health status, social isolation |
| Katzmarzyk et al. [22] | 19223 | Men | 20–83, mean 43 | Fitness test | 10 | Age, year of examination, alcohol use, smoking, family history of CVD, BMI |
| Katzmarzyk et al. [23] | 5421 | Women | 20–69 | Self-report | 12 | Age, smoking, alcohol use, waist circumference |
| Khaw et al. [24] | 22191 | 9984 men; 12207 women | 45–79 | Self-report | 8 | Age, BMI, systolic blood pressure, cholesterol, smoking, alcohol use, diabetes, social status |
| Kohl et al. [25] | 8108 | Men | 20–84, mean 42 | Fitness test | 8 | Age, glycemic status, systolic blood pressure, cholesterol, BMI, family history of CVD, smoking |
| Kushi et al. [26] | 33154 | Women | 55–69 | Self-report | 7 | Age, age at menarche/menopause/first live birth, parity, smoking, alcohol use, BMI, waist-to-hip ratio, energy intake, oestrogen use, hypertension, diabetes, education, marital status |
| Lee et al. [5] | 17231 | Men | Mean 46 | Self-report | 22, 26 | Age, smoking, hypertension, diabetes, early parental death |
| Leon et al. [27] | 12138 | Men | 35–57, mean 46 | Self-report | 16 | Age, education, smoking, cholesterol, diastolic blood pressure, BMI |
| Matthews et al. [28] | 67143 | Women | 40–70, mean 52 | Self-report | 6 | Age, marital status, education, income, smoking, alcohol use, number of pregnancies, oral contraceptive use, menopausal status, diabetes, hypertension, respiratory disease, chronic hepatitis |
| Richardson et al. [29] | 9611 | Men and women | 51–61 | Self-report | 8 | Age, sex, race, self-rated health, history of cancer, obesity, CVD risk |
| Rockhill et al. [30] | 80348 | Women | 30–55 | Self-report | 14 | Age, smoking, alcohol use, height, BMI, hormone use |
| Schooling et al. [31] | 56167 | 18759 men; 37417 women | ≥ 65 | Self-report | 4 | Age, sex, education, alcohol use, smoking, income, BMI |
| Stevens et al. [32] | 5366 | 2860 men; 2506 women | Mean 46 | Fitness test | 22, 26 | Age, education, smoking, alcohol use, diet |
| Trolle-Lagerros et al. [33] | 99099 | Women | 30–49 | Self-report | 11 | Age, education, BMI, alcohol use, smoking |
| Vatten et al. [34] | 54248 | 26515 men; 27769 women | ≥ 20 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure, blood pressure medication |
| Wannamethee et al. [35]b | 4311 | Men | 40–59 | Self-report | 4 | Age, smoking, BMI, social class, self-rated health, other physical activity |
| Wei et al. [36] | 25714 normal weight | Men | Mean 44 | Fitness test | 10 | Age, examination year, BMI, family history of CVD, CVD, diabetes, cholesterol, hypertension, smoking |
| Weller and Corey [37] | 6620 | Women | ≥ 30 | Self-report | 7 | Age |
| Wisloff et al. [38] | 56072 | 27143 men; 28929 women | Mean 47 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure |
aOutcome cardiovascular mortality; all-cause mortality was extracted from the larger 1996 study by Blair. b N=4311 men with no history of coronary heart disease. BMI, body mass index; CVD, cardiovascular disease.
Study characteristics
| Study | Population | Sex | Age | Physical activity assessed by | Follow-up in years (mean) | Results adjusted for |
| Arraiz et al. [8] | 9792 | Men and women | 30–69 | Fitness test | 7 | Age, sex, BMI, smoking |
| Barengo et al. [9] | 32677 | 15853 men; 16824 women | 30–59 | Self-report | 20 | Age, study year, BMI, systolic blood pressure, cholesterol, education, smoking, other physical activity |
| Blair et al. [10]a | 9777 | Men | 20–82 | Fitness test | 5 | Age |
| Blair et al. [11] | 32421 | 25341 men; 7080 women | 20–88 | Fitness test | 8 | Age, smoking, cholesterol, blood pressure, health status |
| Bucksch [12] | 7187 | 3742 men; 3445 women | 30–69 | Self-report | 16 | Age, smoking, BMI, cardiovascular risk factors, alcohol use, diet, chronic disease index, social status |
| Carlsson et al. [13] | 27734 | Women | 51–83 | Self-report | 5 | Age, smoking, education, number of children, hormone replacement, fruit/vegetable intake, BMI, blood pressure, comorbidity |
| Church et al. [14] | 22167 | Men | Mean 43 | Fitness test | 14 | Age, examination year, alcohol use, smoking, family history of CVD, cholesterol, BMI, fasting glucose |
| Davey et al. [15] | 6702 | Men | 40–64 | Self-report | 25 | Age, employment, smoking, BMI, forced expiratory volume |
| Evenson et al. 2004 [16] | 5712 | 3000 men, 2712 women | ≥ 30 | Fitness test | 24, 26 | Age, smoking, education, alcohol use, BMI, race, cholesterol |
| Fang et al. [17] | 9790 | 3730 men, 6060 women | 25–74 | Self-report | 17 | Age, sex, race, family income, education, diabetes, smoking, systolic blood pressure, cholesterol, illnesses affecting eating |
| Gregg et al. [18] | 7553 | Women | ≥ 65 | Self-report | 7 | Age, smoking, BMI, stroke, diabetes, hypertension, self-rated health status |
| Hillsdon et al. [19] | 10522 | Men and women | 35–64 | Self-report | 10 | Age, sex, smoking, alcohol use, health status, social class |
| Hu et al. [20] | 80921 | Women | 30–55 | Self-report | 24 | Age, smoking, family history of CVD, menopausal status, hormone use |
| Hu et al. [7] | 47212 | 22528 men; 24684 women | 25–64 | Self-report | 18 | Age, study year, smoking, systolic blood pressure, cholesterol, BMI, diabetes, education |
| Kaplan et al. [21] | 6131 | Men and women | 16–94, mean 43 | Self-report | 28 | Age, sex, race, education, health status, social isolation |
| Katzmarzyk et al. [22] | 19223 | Men | 20–83, mean 43 | Fitness test | 10 | Age, year of examination, alcohol use, smoking, family history of CVD, BMI |
| Katzmarzyk et al. [23] | 5421 | Women | 20–69 | Self-report | 12 | Age, smoking, alcohol use, waist circumference |
| Khaw et al. [24] | 22191 | 9984 men; 12207 women | 45–79 | Self-report | 8 | Age, BMI, systolic blood pressure, cholesterol, smoking, alcohol use, diabetes, social status |
| Kohl et al. [25] | 8108 | Men | 20–84, mean 42 | Fitness test | 8 | Age, glycemic status, systolic blood pressure, cholesterol, BMI, family history of CVD, smoking |
| Kushi et al. [26] | 33154 | Women | 55–69 | Self-report | 7 | Age, age at menarche/menopause/first live birth, parity, smoking, alcohol use, BMI, waist-to-hip ratio, energy intake, oestrogen use, hypertension, diabetes, education, marital status |
| Lee et al. [5] | 17231 | Men | Mean 46 | Self-report | 22, 26 | Age, smoking, hypertension, diabetes, early parental death |
| Leon et al. [27] | 12138 | Men | 35–57, mean 46 | Self-report | 16 | Age, education, smoking, cholesterol, diastolic blood pressure, BMI |
| Matthews et al. [28] | 67143 | Women | 40–70, mean 52 | Self-report | 6 | Age, marital status, education, income, smoking, alcohol use, number of pregnancies, oral contraceptive use, menopausal status, diabetes, hypertension, respiratory disease, chronic hepatitis |
| Richardson et al. [29] | 9611 | Men and women | 51–61 | Self-report | 8 | Age, sex, race, self-rated health, history of cancer, obesity, CVD risk |
| Rockhill et al. [30] | 80348 | Women | 30–55 | Self-report | 14 | Age, smoking, alcohol use, height, BMI, hormone use |
| Schooling et al. [31] | 56167 | 18759 men; 37417 women | ≥ 65 | Self-report | 4 | Age, sex, education, alcohol use, smoking, income, BMI |
| Stevens et al. [32] | 5366 | 2860 men; 2506 women | Mean 46 | Fitness test | 22, 26 | Age, education, smoking, alcohol use, diet |
| Trolle-Lagerros et al. [33] | 99099 | Women | 30–49 | Self-report | 11 | Age, education, BMI, alcohol use, smoking |
| Vatten et al. [34] | 54248 | 26515 men; 27769 women | ≥ 20 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure, blood pressure medication |
| Wannamethee et al. [35]b | 4311 | Men | 40–59 | Self-report | 4 | Age, smoking, BMI, social class, self-rated health, other physical activity |
| Wei et al. [36] | 25714 normal weight | Men | Mean 44 | Fitness test | 10 | Age, examination year, BMI, family history of CVD, CVD, diabetes, cholesterol, hypertension, smoking |
| Weller and Corey [37] | 6620 | Women | ≥ 30 | Self-report | 7 | Age |
| Wisloff et al. [38] | 56072 | 27143 men; 28929 women | Mean 47 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure |
| Study | Population | Sex | Age | Physical activity assessed by | Follow-up in years (mean) | Results adjusted for |
| Arraiz et al. [8] | 9792 | Men and women | 30–69 | Fitness test | 7 | Age, sex, BMI, smoking |
| Barengo et al. [9] | 32677 | 15853 men; 16824 women | 30–59 | Self-report | 20 | Age, study year, BMI, systolic blood pressure, cholesterol, education, smoking, other physical activity |
| Blair et al. [10]a | 9777 | Men | 20–82 | Fitness test | 5 | Age |
| Blair et al. [11] | 32421 | 25341 men; 7080 women | 20–88 | Fitness test | 8 | Age, smoking, cholesterol, blood pressure, health status |
| Bucksch [12] | 7187 | 3742 men; 3445 women | 30–69 | Self-report | 16 | Age, smoking, BMI, cardiovascular risk factors, alcohol use, diet, chronic disease index, social status |
| Carlsson et al. [13] | 27734 | Women | 51–83 | Self-report | 5 | Age, smoking, education, number of children, hormone replacement, fruit/vegetable intake, BMI, blood pressure, comorbidity |
| Church et al. [14] | 22167 | Men | Mean 43 | Fitness test | 14 | Age, examination year, alcohol use, smoking, family history of CVD, cholesterol, BMI, fasting glucose |
| Davey et al. [15] | 6702 | Men | 40–64 | Self-report | 25 | Age, employment, smoking, BMI, forced expiratory volume |
| Evenson et al. 2004 [16] | 5712 | 3000 men, 2712 women | ≥ 30 | Fitness test | 24, 26 | Age, smoking, education, alcohol use, BMI, race, cholesterol |
| Fang et al. [17] | 9790 | 3730 men, 6060 women | 25–74 | Self-report | 17 | Age, sex, race, family income, education, diabetes, smoking, systolic blood pressure, cholesterol, illnesses affecting eating |
| Gregg et al. [18] | 7553 | Women | ≥ 65 | Self-report | 7 | Age, smoking, BMI, stroke, diabetes, hypertension, self-rated health status |
| Hillsdon et al. [19] | 10522 | Men and women | 35–64 | Self-report | 10 | Age, sex, smoking, alcohol use, health status, social class |
| Hu et al. [20] | 80921 | Women | 30–55 | Self-report | 24 | Age, smoking, family history of CVD, menopausal status, hormone use |
| Hu et al. [7] | 47212 | 22528 men; 24684 women | 25–64 | Self-report | 18 | Age, study year, smoking, systolic blood pressure, cholesterol, BMI, diabetes, education |
| Kaplan et al. [21] | 6131 | Men and women | 16–94, mean 43 | Self-report | 28 | Age, sex, race, education, health status, social isolation |
| Katzmarzyk et al. [22] | 19223 | Men | 20–83, mean 43 | Fitness test | 10 | Age, year of examination, alcohol use, smoking, family history of CVD, BMI |
| Katzmarzyk et al. [23] | 5421 | Women | 20–69 | Self-report | 12 | Age, smoking, alcohol use, waist circumference |
| Khaw et al. [24] | 22191 | 9984 men; 12207 women | 45–79 | Self-report | 8 | Age, BMI, systolic blood pressure, cholesterol, smoking, alcohol use, diabetes, social status |
| Kohl et al. [25] | 8108 | Men | 20–84, mean 42 | Fitness test | 8 | Age, glycemic status, systolic blood pressure, cholesterol, BMI, family history of CVD, smoking |
| Kushi et al. [26] | 33154 | Women | 55–69 | Self-report | 7 | Age, age at menarche/menopause/first live birth, parity, smoking, alcohol use, BMI, waist-to-hip ratio, energy intake, oestrogen use, hypertension, diabetes, education, marital status |
| Lee et al. [5] | 17231 | Men | Mean 46 | Self-report | 22, 26 | Age, smoking, hypertension, diabetes, early parental death |
| Leon et al. [27] | 12138 | Men | 35–57, mean 46 | Self-report | 16 | Age, education, smoking, cholesterol, diastolic blood pressure, BMI |
| Matthews et al. [28] | 67143 | Women | 40–70, mean 52 | Self-report | 6 | Age, marital status, education, income, smoking, alcohol use, number of pregnancies, oral contraceptive use, menopausal status, diabetes, hypertension, respiratory disease, chronic hepatitis |
| Richardson et al. [29] | 9611 | Men and women | 51–61 | Self-report | 8 | Age, sex, race, self-rated health, history of cancer, obesity, CVD risk |
| Rockhill et al. [30] | 80348 | Women | 30–55 | Self-report | 14 | Age, smoking, alcohol use, height, BMI, hormone use |
| Schooling et al. [31] | 56167 | 18759 men; 37417 women | ≥ 65 | Self-report | 4 | Age, sex, education, alcohol use, smoking, income, BMI |
| Stevens et al. [32] | 5366 | 2860 men; 2506 women | Mean 46 | Fitness test | 22, 26 | Age, education, smoking, alcohol use, diet |
| Trolle-Lagerros et al. [33] | 99099 | Women | 30–49 | Self-report | 11 | Age, education, BMI, alcohol use, smoking |
| Vatten et al. [34] | 54248 | 26515 men; 27769 women | ≥ 20 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure, blood pressure medication |
| Wannamethee et al. [35]b | 4311 | Men | 40–59 | Self-report | 4 | Age, smoking, BMI, social class, self-rated health, other physical activity |
| Wei et al. [36] | 25714 normal weight | Men | Mean 44 | Fitness test | 10 | Age, examination year, BMI, family history of CVD, CVD, diabetes, cholesterol, hypertension, smoking |
| Weller and Corey [37] | 6620 | Women | ≥ 30 | Self-report | 7 | Age |
| Wisloff et al. [38] | 56072 | 27143 men; 28929 women | Mean 47 | Self-report | 16 | Age, BMI, marital status, education, alcohol use, smoking, blood pressure |
aOutcome cardiovascular mortality; all-cause mortality was extracted from the larger 1996 study by Blair. b N=4311 men with no history of coronary heart disease. BMI, body mass index; CVD, cardiovascular disease.
The aim of this systematic review was therefore to summarise the results of the largest cohort studies that examine the effects of physical activity on cardiovascular and all-cause mortality.
Methods
In a systematic MEDLINE search conducted in May 2007, the authors identified prospective cohort studies that assessed the impact of physical activity on all-cause and cardiovascular mortality. Studies analysing populations with a history of cardiovascular or other serious diseases were excluded. To be included in the authors’ review, studies needed to be published in English or German and have a minimum population size of 5000 and a minimum follow-up of 3 years. The authors reported risk reductions on the basis of comparison between the least active and the most active population subgroups, with the least active population subgroup as the reference group. In cases where results were adjusted for relevant variables in addition to age, the authors selected the results of the most extensively adjusted models.
Meta-analyses were performed using a generic inverse variance approach with RevMan version 4.2 (The Cochrane Collaboration, Copenhagen, Denmark). Funnel plots were used to assess publication bias. The results of the meta-analysis were presented using forest plots stratified according to the mean by which physical activity had been assessed (i.e. self-reported or objectively measured). Random-effect models were used to pool study results. Heterogeneity was assessed using χ2 and I methods. The authors performed sensitivity and subgroup analyses to determine the impact of sex, the number of variables in the final model, and the number of physical activity categories.
Results
The authors’ MEDLINE search yielded 1768 articles, 1469 of which were excluded on the basis of a review of the title and the abstract. The remaining 299 studies were analysed in greater detail, and 33 studies fulfilled all inclusion criteria. The main study characteristics are presented in Table 1. The studies ultimately included in the authors’ analysis published between 1992 and 2007, had a total of 883372 participants, and included a roughly equal number of men and women (not all studies provided information on the number of men and women in their samples). The follow-up period in the included studies ranged from 4 years to more than 20 years. A total of nine studies used a fitness test (usually a treadmill test) to assess physical activity, and 24 studies used patient questionnaires. Most studies reported results adjusted for other known risk factors such as hypertension, high cholesterol, and obesity, and 15 studies also reported results for parsimonious models (usually adjusted only for age).
Funnel plots for cardiovascular and all-cause mortality are presented in Figs 1 and 2. For cardiovascular mortality, the distribution of studies is not symmetrical. Although the majority of studies is grouped around the pooled effect, some studies reported larger risk reductions, which is an indicator of publication bias. All studies reporting larger risk reductions were, however, based on objective fitness tests and had less precise effect estimates. The funnel plot for all-cause mortality shows similar nonsymmetrical distributions, though outliers are closer to the pooled estimate than in Fig. 1. Here, too, studies based on objective fitness tests tended to report larger risk reductions.
Associations between physical activity and cardiovascular mortality are presented in Fig. 3. Of all the studies included in the authors’ analysis, five reported nonsignificant results and 26 reported statistically significant risk reductions. Reductions in risk ranged from 11 to 81%. Most studies reported a risk reduction of 30–50%. Risk reductions were found for both men and women.
Funnel plot for physical activity and cardiovascular mortality.
Funnel plot for physical activity and cardiovascular mortality.
Funnel plot for physical activity and all-cause mortality.
Funnel plot for physical activity and all-cause mortality.
Relative risk of cardiovascular mortality in physically active versus physically inactive participants (fully adjusted models).
Relative risk of cardiovascular mortality in physically active versus physically inactive participants (fully adjusted models).
The overall pooled risk reduction was 35%. Studies that used a fitness test to assess physical activity reported larger risk reductions than studies based on self-reported activity (57 vs. 30%).
The results were similar for the association between physical activity and all-cause mortality (Fig. 4). Here, the reductions in risk ranged from 2 to 61%, with 33 of the 35 results being significant. The overall pooled risk reduction was 33%. Again, the largest reductions were found in studies that used a fitness test to assess physical activity (41 vs. 29%).
Sensitivity and subgroup analyses are presented in Table 2. Overall, 15 studies reported age-adjusted results in addition to fully adjusted models. The age-adjusted risk reductions were always larger than the fully adjusted reductions. Stratifying the fitness status of the study populations into more categories (e.g., five instead of three fitness categories) increased the risk reductions moderately. The overall results were similar in men and women. After stratifying the data according to the method that had been used to assess physical activity (i.e. by self-reported or objectively measured data), the risk reductions associated with physical activity were larger for women than for men (Table 2).
Relative risk of all-cause mortality in physically active versus physically inactive participants (fully adjusted models).
Relative risk of all-cause mortality in physically active versus physically inactive participants (fully adjusted models).
Sensitivity/subgroup analyses for physical activity and cardiovascular/all-cause mortality
| Fitness test | Self-report | Total | |
| Cardiovascular mortality | |||
| Only age adjusted | |||
| Relative risk (95% CI) | 0.27 (0.20–0.36) | 0.59 (0.53–0.66) | 0.53 (0.46–0.61) |
| Number of studies | N=3 | N=10 | N=13 |
| Fitness categories compared < =3 | |||
| Relative risk (95% CI) | 0.49 (0.36–0.68) | 0.69 (0.61–0.78) | 0.66 (0.59–0.74) |
| Number of studies | N=4 | N=9 | N=13 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.37 (0.23–0.60) | 0.70 (0.63–0.79) | 0.65 (0.59–0.72) |
| Number of studies | N=6 | N=12 | N=18 |
| Men | |||
| Relative risk (95% CI) | 0.48 (0.36–0.64) | 0.74 (0.70–0.79) | 0.65 (0.57–0.72) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.31 (0.17–0.56) | 0.65 (0.58–0.72) | 0.63 (0.56–0.71) |
| Number of studies | N=2 | N=10 | N=12 |
| All-cause mortality | |||
| Only age-adjusted | |||
| Relative risk (95% CI) | 0.46 (0.38–0.55) | 0.60 (0.54–0.67) | 0.59 (0.53–0.65) |
| Number of studies | N=3 | N=16 | N=19 |
| Fitness categories compared <=3 | |||
| Relative risk (95% CI) | 0.60 (0.52–0.69) | 0.73 (0.67–0.80) | 0.69 (0.64–0.75) |
| Number of studies | N=6 | N=12 | N=18 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.55 (0.46–0.66) | 0.67 (0.60–0.76) | 0.65 (0.58–0.72) |
| Number of studies | N=4 | N=13 | N=17 |
| Men | |||
| Relative risk (95% CI) | 0.61 (0.55–0.68) | 0.78 (0.70–0.68) | 0.70 (0.64–0.77) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.49 (0.38–0.62) | 0.65 (0.59–0.73) | 0.63 (0.57–0.70) |
| Number of studies | N=3 | N=13 | N=16 |
| Fitness test | Self-report | Total | |
| Cardiovascular mortality | |||
| Only age adjusted | |||
| Relative risk (95% CI) | 0.27 (0.20–0.36) | 0.59 (0.53–0.66) | 0.53 (0.46–0.61) |
| Number of studies | N=3 | N=10 | N=13 |
| Fitness categories compared < =3 | |||
| Relative risk (95% CI) | 0.49 (0.36–0.68) | 0.69 (0.61–0.78) | 0.66 (0.59–0.74) |
| Number of studies | N=4 | N=9 | N=13 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.37 (0.23–0.60) | 0.70 (0.63–0.79) | 0.65 (0.59–0.72) |
| Number of studies | N=6 | N=12 | N=18 |
| Men | |||
| Relative risk (95% CI) | 0.48 (0.36–0.64) | 0.74 (0.70–0.79) | 0.65 (0.57–0.72) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.31 (0.17–0.56) | 0.65 (0.58–0.72) | 0.63 (0.56–0.71) |
| Number of studies | N=2 | N=10 | N=12 |
| All-cause mortality | |||
| Only age-adjusted | |||
| Relative risk (95% CI) | 0.46 (0.38–0.55) | 0.60 (0.54–0.67) | 0.59 (0.53–0.65) |
| Number of studies | N=3 | N=16 | N=19 |
| Fitness categories compared <=3 | |||
| Relative risk (95% CI) | 0.60 (0.52–0.69) | 0.73 (0.67–0.80) | 0.69 (0.64–0.75) |
| Number of studies | N=6 | N=12 | N=18 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.55 (0.46–0.66) | 0.67 (0.60–0.76) | 0.65 (0.58–0.72) |
| Number of studies | N=4 | N=13 | N=17 |
| Men | |||
| Relative risk (95% CI) | 0.61 (0.55–0.68) | 0.78 (0.70–0.68) | 0.70 (0.64–0.77) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.49 (0.38–0.62) | 0.65 (0.59–0.73) | 0.63 (0.57–0.70) |
| Number of studies | N=3 | N=13 | N=16 |
Sensitivity/subgroup analyses for physical activity and cardiovascular/all-cause mortality
| Fitness test | Self-report | Total | |
| Cardiovascular mortality | |||
| Only age adjusted | |||
| Relative risk (95% CI) | 0.27 (0.20–0.36) | 0.59 (0.53–0.66) | 0.53 (0.46–0.61) |
| Number of studies | N=3 | N=10 | N=13 |
| Fitness categories compared < =3 | |||
| Relative risk (95% CI) | 0.49 (0.36–0.68) | 0.69 (0.61–0.78) | 0.66 (0.59–0.74) |
| Number of studies | N=4 | N=9 | N=13 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.37 (0.23–0.60) | 0.70 (0.63–0.79) | 0.65 (0.59–0.72) |
| Number of studies | N=6 | N=12 | N=18 |
| Men | |||
| Relative risk (95% CI) | 0.48 (0.36–0.64) | 0.74 (0.70–0.79) | 0.65 (0.57–0.72) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.31 (0.17–0.56) | 0.65 (0.58–0.72) | 0.63 (0.56–0.71) |
| Number of studies | N=2 | N=10 | N=12 |
| All-cause mortality | |||
| Only age-adjusted | |||
| Relative risk (95% CI) | 0.46 (0.38–0.55) | 0.60 (0.54–0.67) | 0.59 (0.53–0.65) |
| Number of studies | N=3 | N=16 | N=19 |
| Fitness categories compared <=3 | |||
| Relative risk (95% CI) | 0.60 (0.52–0.69) | 0.73 (0.67–0.80) | 0.69 (0.64–0.75) |
| Number of studies | N=6 | N=12 | N=18 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.55 (0.46–0.66) | 0.67 (0.60–0.76) | 0.65 (0.58–0.72) |
| Number of studies | N=4 | N=13 | N=17 |
| Men | |||
| Relative risk (95% CI) | 0.61 (0.55–0.68) | 0.78 (0.70–0.68) | 0.70 (0.64–0.77) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.49 (0.38–0.62) | 0.65 (0.59–0.73) | 0.63 (0.57–0.70) |
| Number of studies | N=3 | N=13 | N=16 |
| Fitness test | Self-report | Total | |
| Cardiovascular mortality | |||
| Only age adjusted | |||
| Relative risk (95% CI) | 0.27 (0.20–0.36) | 0.59 (0.53–0.66) | 0.53 (0.46–0.61) |
| Number of studies | N=3 | N=10 | N=13 |
| Fitness categories compared < =3 | |||
| Relative risk (95% CI) | 0.49 (0.36–0.68) | 0.69 (0.61–0.78) | 0.66 (0.59–0.74) |
| Number of studies | N=4 | N=9 | N=13 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.37 (0.23–0.60) | 0.70 (0.63–0.79) | 0.65 (0.59–0.72) |
| Number of studies | N=6 | N=12 | N=18 |
| Men | |||
| Relative risk (95% CI) | 0.48 (0.36–0.64) | 0.74 (0.70–0.79) | 0.65 (0.57–0.72) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.31 (0.17–0.56) | 0.65 (0.58–0.72) | 0.63 (0.56–0.71) |
| Number of studies | N=2 | N=10 | N=12 |
| All-cause mortality | |||
| Only age-adjusted | |||
| Relative risk (95% CI) | 0.46 (0.38–0.55) | 0.60 (0.54–0.67) | 0.59 (0.53–0.65) |
| Number of studies | N=3 | N=16 | N=19 |
| Fitness categories compared <=3 | |||
| Relative risk (95% CI) | 0.60 (0.52–0.69) | 0.73 (0.67–0.80) | 0.69 (0.64–0.75) |
| Number of studies | N=6 | N=12 | N=18 |
| Fitness categories compared >3 | |||
| Relative risk (95% CI) | 0.55 (0.46–0.66) | 0.67 (0.60–0.76) | 0.65 (0.58–0.72) |
| Number of studies | N=4 | N=13 | N=17 |
| Men | |||
| Relative risk (95% CI) | 0.61 (0.55–0.68) | 0.78 (0.70–0.68) | 0.70 (0.64–0.77) |
| Number of studies | N=7 | N=8 | N=15 |
| Women | |||
| Relative risk (95% CI) | 0.49 (0.38–0.62) | 0.65 (0.59–0.73) | 0.63 (0.57–0.70) |
| Number of studies | N=3 | N=13 | N=16 |
Discussion
Physical activity is associated with a marked decrease in cardiovascular and all-cause mortality. Most studies included in the authors’ meta-analysis reported risk reductions of 30–50% for cardiovascular mortality and of 20–50% for all-cause mortality, with pooled risk reductions of 35% for the former and 33% for the latter. These results are based on studies including a total of nearly 900000 participants.
Most of these findings were based on regression models adjusted for important risk factors such as hypertension, hypercholesterolemia, and diabetes, which themselves are associated with physical inactivity. As a result, the risk reductions reported were rather conservative. Adjusting only for age, increased the already large risk reductions for cardiovascular mortality from 35 to 47% and for all-cause mortality from 33 to 41%. Accordingly, the largest reductions in the risk of cardiovascular mortality were reported by the two studies that did not adjust for important confounding variables [8,10]. In one study [7], including these variables attenuated the risk reduction for all-cause mortality compared to a model adjusted only for age by 6% for women and 10% for men, and for cardiovascular mortality by 8% for women and 13% for men.
Both men and women benefited from physical activity. According to the authors’ subgroup analyses, risk reductions were larger for women than for men after stratifying the populations according to the method that had been used to assess physical activity. With regard to age, most studies in this review focused on middle-aged populations. Three studies [18,26,31] included older participants and reported similar decreases in mortality.
Although the large majority of studies reported that physical activity had protective effects on both cardiovascular and all-cause mortality, the risk reductions varied considerably. The most important reason for heterogeneity was the differing methods used to assess physical activity. The authors included studies that assessed physical fitness (i.e. a physiological state measured by treadmill tests) as well as studies that relied on patient self-reports of the duration and intensity of physical activities. Studies with a more objective means of assessment tended to report larger risk reductions. Accordingly, a study by Myers et al. [4], which directly compared both methods found that exercise testing outperformed self-reported physical activity in predicting mortality. It seems likely that participants overestimated their levels of physical activity in self-reports, thus minimising the true protective effect. However, even studies based on the less reliable self-reported measures of activity have shown marked protective effects on mortality.
The authors’ study has several important limitations. Based on the authors’ inclusion and exclusion criteria, a bias in the selection of studies is possible. The authors included only large studies with more than 5000 participants. Though this was an arbitrary cut-off, the large number of studies that fulfilled this criterion (i.e. with a total of nearly 900000 participants) and the overall homogeneity in the direction of effects support the validity of the authors’ findings. Another limitation is that the participants were not classified into distinct activity groups in a homogenous manner across the studies, especially when classification was based on patient questionnaires. Moreover, the number of activity categories differed across studies. Whereas most studies compared three physical activity groups (e.g. low, moderate, and high), some studies categorized the populations into four or five subgroups, thus increasing the range between the least and most active groups. Including only studies that compared three distinct activity groups, however, had only small effects on overall risk reductions.
These results once again confirm the impressive protective effect of physical activity. Despite this knowledge, a large majority of individuals still lead a primarily sedentary life. Numerous studies have addressed the question of how to promote physical activity [39]. Positive results have been shown for information campaigns, behavioural and social interventions, and environmental interventions, such as providing easy access to sports or exercise facilities. The key is to translate these findings into permanent public health efforts.
Acknowledgements
Competing interests: none.
References
Comments