One of the 2 major goals of Healthy People 2010 is to increase the quality and years of healthy life.1 In the United States, the number 1 cause of mortality is coronary heart disease (CHD), a condition that comprised 452,300 (20%) of all deaths in 2004.2 In terms of expenditures, it is estimated that the annual total direct and indirect costs associated with CHD will be $151.6 billion in 2007.2 The 6 major risk factors for CHD are cigarette smoking, physical inactivity, diabetes, hypertension, hyperlipidemia, and overweight/obesity.2 Although 2 of these risk factors are behaviors (smoking and physical inactivity), the other 4 are related to behaviors.2 In terms of the prevalence of these risk factors, approximately 15.2 million US adults are reported to have physician-diagnosed type 1 or type 2 diabetes, 95% of which are type 2 diabetes while 72, 79.3, and 140 million adults, respectively, have hypertension or hyperlipidemia or are classified as overweight or obese.2 Aerobic exercise, a low-cost, low-risk, nonpharmacologic intervention that is available to the vast majority of the general public, has been recommended for improving these risk factors.2 These recommendations have traditionally been derived from the inclusion of meta-analyses of randomized controlled trials,3 considered by some to be one of the highest forms of evidence for determining the clinical utility of an intervention on an outcome.4 Unfortunately, most meta-analyses focus on only one primary outcome, for example, glycosylated hemoglobin (HbA1c) in patients with type 2 diabetes as opposed to examining multiple CHD risk factors such as HbA1c, resting blood pressure (BP), cholesterol, and body mass index (BMI). From a practical perspective, this is problematic because the multiple benefits of an intervention such as aerobic exercise on CHD risk factors may not be well appreciated. Despite its importance, we are not aware of any previous work that has attempted to quantify the benefits of aerobic exercise on multiple CHD risk factors from meta-analyses of previously conducted randomized controlled trials. Given this gap and its potential significance, the purpose of this review was to determine the efficacy of aerobic exercise on multiple selected CHD risk factors in adults using data from previously published meta-analyses.
The authors examined the effects of aerobic exercise on selected coronary heart disease (CHD) risk factors using data from previously published meta-analyses. Using a random effects model, the effects of aerobic exercise on glycosylated hemoglobin (HbA1c) (mean, 95% confidence interval, −0.9%, −1.9% to 0.03%), resting systolic blood pressure (−6.9 mm Hg, −9.1 to −4.6 mm Hg), low-density lipoprotein cholesterol (−3.1 mg/dL, −6.1 to 0 mg/dL), and body mass index (−1.3 kg/m2, −2.5 to −0.1 kg/m2) were either statistically significant or demonstrated a trend for statistical significance. Changes were equivalent to relative reductions of −8.5%, −4.7%, −2.0%, and −4.5%, respectively. Changes corresponded to estimated 5-year reductions in CHD mortality of 14%, 17%, 1.5%, and 5%, respectively. The results of this review reinforce the idea that aerobic exercise is an important nonpharmacologic intervention for improving selected CHD risk factors.
Electronic computer searches (PubMed, Sport Discus, Cochrane Database of Systematic Reviews) were conducted to identify previously published meta-analyses that included primary outcome data on the effects of aerobic exercise on ≥1 of the following 4 CHD risk factors: type 2 diabetes, hypertension, abnormal lipid and lipoprotein levels, and overweight/obesity. The major keywords used in the searches included exercise, BP, cholesterol, obesity, and diabetes. All searches were limited to meta-analyses published and indexed in the English-language literature between January 1, 1977, and May 16, 2007.
The inclusion criteria for this analysis was the most recent English-language meta-analysis in which primary outcome data could be derived that examined the independent effects of aerobic exercise (no other interventions such as diet) on (1) HbA1c in patients with type 2 diabetes, (2) resting systolic BP (SBP) in patients with hypertension (resting systolic and/or diastolic BP ≥140/90 mm Hg), (3) fasting low-density lipoprotein cholesterol (LDL-C) in hyperlipidemic patients (LDL-C ≥130 mg/dL or 3.3 mmol/L), and (4) BMI in overweight/obese patients (BMI ≥25 kg/m2). In addition, meta-analyses were limited to those that included randomized controlled trials, used a random-effects model to analyze data or provided sufficient data so that a random-effects analysis could be conducted, and included patients 18 years of age and older. All included meta-analyses were selected by the first author and then reviewed and confirmed by the second author to ensure that they met the selection criteria.
Data that were abstracted from each meta-analysis included the authors of the meta-analysis, source and year in which the meta-analysis was published, data on the risk factor examined (eg, means, standard errors, 95% confidence intervals), number of randomized controlled trials included, number of patients, age of patients in years, sex, physical activity status of patients before taking part in the study, medications that could effect the risk factor examined, and exercise program characteristics (length, frequency, intensity, duration, mode, compliance, supervision status). All data from each included meta-analysis were abstracted by the first author and then reviewed and confirmed by the second author.
The primary outcomes in this review were HbA1c, resting SBP, fasting LDL-C, and BMI. We used changes in HbA1c over other outcomes such as the oral glucose tolerance test because it has been shown to be more convenient and reproducible and is used to make therapeutic decisions.5 Changes in resting SBP were chosen over resting diastolic BP (DBP) because the former has been shown to be a better predictor of CHD morbidity and mortality.6,7 We chose LDL-C as our primary outcome because it is currently the primary target of lipid-lowering therapy.8 We used changes in BMI as the outcome of choice for overweight and obesity because it is the most commonly used measure for estimating overweight and obesity in large populations.
We limited our inclusion to 1 variable per risk factor (eg, LDL-C vs both LDL-C and high-density lipoprotein cholesterol) because we wanted to limit our analysis to 1 variable from each risk factor group. From the data provided in each meta-analysis, we abstracted and reported the mean pooled treatment effect, defined as the change outcome difference between the aerobic exercise and control groups, along with 95% confidence intervals for each CHD risk factor based on a random-effects model. If a fixed-effects analysis was performed and reported for the meta-analysis, we abstracted data from each study provided in the meta-analysis and conducted a random-effects analysis (inverse variance-weighted approach). We focused on a random-effects analysis across all outcomes for consistency as well as the fact that a random-effects model controls (statistically) for heterogeneity.9 If any of the pooled results from each meta-analysis did not meet our inclusion criteria (eg, including a group that took part in both aerobic exercise and a dietary intervention), we abstracted data only from those studies that were limited to aerobic exercise and then conducted a random-effects meta-analysis. If not reported in the original meta-analysis itself and/or the studies we included, meta-regression (random effects, method of moments of approach) was performed to examine the association between changes in body weight and changes in HbA1c, resting SBP, and fasting LDL-C. An α level ≤0.05 was considered to be statistically significant. In addition to estimating our effects using the original metric, we also calculated the relative treatment effect change (percent change) for each risk factor when compared with each risk factor's baseline values. This allowed for a comparison of the effects of aerobic exercise on each risk factor using a common metric. Finally, using data from previously published studies,10–13 we estimated the 5-year, relative-risk reduction in CHD mortality based on the changes observed for each risk factor in our analysis. All data analyses that needed to be performed were conducted using Stata S/E (version 8.2)14 and Comprehensive Meta-Analysis (version 2.2).15
Five meta-analyses met our inclusion criteria.16–20 A description of the study characteristic data that we abstracted from each meta-analysis is shown in Table I. As can be seen, no meta-analysis was published before 2004 and the number of trials that we included from the Type 2 diabetes and overweight/obesity meta-analyses was small.16,20 For the Type 2 diabetes meta-analysis we limited the abstraction of data to studies in which aerobic exercise was the only intervention.16 For the hypertension meta-analysis,17 we only included data reported for hypertensive adults although separate data were available for normotensive and prehypertensive subjects.
|Reference||Risk Factor||Study Description||Subjects||Medications||Aerobic Exercise|
|Thomas et al. (2006)16||Type 2 diabetes (duration 1–13 years) (HbA1c, mean ± SD, 10.6±1.7%)||3 RCTs, 84 subjects (43 exercise, 41 control)||Sedentary, males and females, 45–60 years of age||Low (8 subjects) or moderate (9 subjects) insulin secreters; sulphonylureas (18 subjects) or sulphonylureas + metformin (10 subjects)||12 weeks; 3–7 times per week; 20–50 minutes per session; 70%–75% of VO2max; walking, running, cycling,swimming, cross-country skiing|
|Cornelissen & Fagard (2005)17||Hypertension (SBP, mean ± SD, 146.7±8.1 mm Hg)||28 RCTs, 31 exercise groups, 492 trained subjects||Mostly sedentary, males and females (Age, mean ± SD, 52.7±11.8 years)||Antihypertensive treatment in some or all subjects in 6 groups (19%); taken off medication in 6 groups (19%)||4–52 weeks (Md=12); 2–7 times per week (Md=3); 25 to 60 minutes per session (Md=40); 30%–87% of HRR (Md=65); mostly walking, running, & cycling; mostly supervised sessions|
|Kelley et al. (2004)18 Kelley and Kelley (2006)19||Cholesterol (LDL-C, mean ± SD, 153.6±15.0 mg/dL)||26 studies, 36 exercise groups, 1789 subjects (1038 exercise, 751 control)||Mostly sedentary males and females, 20–76 years of age (mean ± SD, 40.6±12.7 years)||Drugs that may have affected lipid levels taken by 6 groups (17%)||12–52 weeks (mean ± SD, 29.4±16.7); 17–60 minutes per session (mean ± SD, 41.4±11.7); 53% to 85% of VO2max (mean ± SD, 71±9); mostly walking, running, & cycling; mostly supervised (compliance, mean ± SD, 84.2%±15.4%)|
|Shaw et al. (2006)20a||Overweight /Obesity (BMI, mean ± SD, 28.7±3.5 kg/m2)||3 RCTs, 194 subjects (104 exercise, 90 control)||Sedentary males and females, 35–60 years of age||No medications taken for weight loss||12–52 weeks of primarily walking and/or jogging, 3 times per week, ≥30 minutes per session, 60% to 80% of MHR; mostly supervised|
|aAuthors of meta-analysis limited inclusion criteria to studies with a subject drop out rate <15%20; Abbreviations: BMI, body mass index; HbA1c, glycosylated hemoglobin; HRR, heart rate reserve; LDL-C, low-density lipoprotein cholesterol; Md, median; MHR, maximum heart rate; RCTs, randomized controlled trials; SBP, systolic blood pressure; VO2 max, maximum oxygen consumption.|
While our a priori decision was to only include the most recent meta-analysis in which LDL-C values were ≥130 mg/dL (3.3 mmol/L), no such study was available. Therefore, we merged data from 2 recent meta-analyses, one dealing with women18 and the other in men,20 since changes in LDL-C were similar across both studies. We then limited our random-effects analysis to those studies in which the mean initial LDL-C level was ≥130 mg/dL (3.3 mmol/L).
Changes in CHD Risk Factors
The results for changes in CHD risk factors are shown in Table II. As can be seen, the 95% CIs for treatment effect changes in SBP and BMI did not include zero (0) while a trend for reductions in HbA1c and LDL-C were observed. No statistically significant association was found between changes in body weight and changes in HbA1c (P=0.35) and resting SBP (P>0.05) as reported in the original meta-analysis.15 A trend was observed for fasting LDL-C (P=0.09). From a relative perspective, the greatest improvements were observed for HbA1c followed by SBP, BMI, and LDL-C. When 5-year relative risk estimates for mortality from CHD were calculated, the greatest reduction was found for resting SBP followed by HbA1c, BMI, and LDL-C.
|Variable||Studies (#)||Subjects (Total#)||Change (OM) mean (95% CI)||Change (%) mean (95% CI)||Change (RR%) mean (95% CI)|
|HbA1c, %||3||84||−0.9 (−1.9 to 0.03)||−8.5 (−17.9 to 0.3)||−14% (−30% to 0.5%)|
|SBP, mm Hg||28||492a||−6.9 (−9.1 to −4.6)a||−4.7 (−6.2 to −3.1)a||−17% (−23% to −12%)a|
|LDL-C, mg/dL||26||1038||−3.1 (−6.1 to 0)||−2.0 (−4.0 to 0)||−1.5% (−3% to 0%)|
|LDL-C, mmol/L||−0.08 (0.16 to 0)||____||____|
|BMI, kg/m2||3||194||−1.3 (−2.5 to −0.1)a||−4.5 (−8.7 to −0.3)a||−5% (−10% to −0.4%)a|
|Exercise subjects only. a95% CI does not cross zero. Abbreviations: BMI, body mass index; Change (OM), absolute change in the original metric; Change (%), percent change from baseline values; HbA1c, glycosylated hemoglobin; LDL-C, low-density lipoprotein cholesterol; RR%, five-year relative risk reduction in CHD mortality; SBP, systolic blood pressure.|
While the age-adjusted death rate for CHD in men and women aged 25 to 84 years decreased from 542.9 to 266.8 cases per 100,000 population (men) and 263.3 to 134.4 cases per 100,000 population (women) between 1980 and 2000,21 CHD still remains the number 1 cause of death among adults in the United States.2 The primary purpose of this review was to use a quantitative approach to determine the efficacy of aerobic exercise on selected CHD risk factors in adults using data from previously published meta-analyses. Our overall results reinforce the importance of aerobic exercise on CHD risk factors in subjects with hypertension as well as those who are overweight/obese, with a trend for positive effects in subjects with Type 2 diabetes and hyperlipidemia.
The lack of a statistically significant relationship between changes in body weight and changes in our selected CHD risk factors was somewhat surprising. This may have been the result of the small changes in body weight that occurred in the majority of studies that comprised the meta-analyses we included in our analysis. Alternatively, this may have been the result of a lack of statistical power to detect this association.
We are not aware of any quantitative studies that have collectively determined the effects of aerobic exercise on CHD mortality that could be expected as a result of treatment effect changes in HbA1c, resting SBP, fasting LDL-C, and BMI. However, using data derived from non-exercise studies, our observed changes resulted in important 5-year relative risk reductions for CHD (Table II). In absolute terms and based on current prevalence estimates,2 this would result in the avoidance or postponement of 2, 12.2, 1.2, and 7 million CHD deaths, respectively, in subjects with type 2 diabetes, hypertension, hyperlipidemia, and overweight/obesity.10–13
The benefits of aerobic exercise observed in this analysis are important since pharmacologic interventions are traditionally targeted to 1 risk factor, carry greater risk of side effects, and are more expensive when compared with nonpharmacologic interventions such as aerobic exercise. This may be especially relevant given the prevalence of the metabolic syndrome, a condition that exists in an estimated 47 million adults ages 20 years and older in the United States22 and which has been associated with an estimated 9.1 CHD deaths per 1000 person-years in adults 30 to 75 years of age (average follow-up, 13 years).23 In addition, knowledge of these multiple benefits may provide greater motivation for health care providers, including clinicians, to promote and support the adoption of a regular and sustainable program of aerobic exercise for their clients. This is important since the benefits of aerobic exercise extend well beyond CHD.24 Unfortunately, the prevalence of participation in regular physical activity which includes aerobic exercise is less than optimal. For example, in 2004, the age-adjusted prevalence of regular leisure-time physical activity in the United States, defined as light to moderate activity for ≥30 minutes, ≥5 times per week, or vigorous activity for ≥20 minutes, ≥3 times per week, was estimated to be 30.1% in adults 18 years of age and older.2 Clearly, a need exists for improved approaches for increasing the physical activity levels of adults in the United States so that the benefits of this lifestyle change can be realized.
The results of our analysis need to be interpreted with respect to the following potential limitations.26 First, our results may not be generalized beyond the characteristics of the subjects and exercise training program characteristics of the studies we included from each meta-analysis.16–20,25 For example, similar to another recent quantitative analysis,27 most subjects were of North American or European descent. Consequently, the effects of aerobic exercise on the outcomes included in our analysis may differ in other groups of people. In addition, most of the studies included in these meta-analyses appeared to have subjects participate in more intense types of training like running. As a result, the effects of lower intensity types of training such as walking, the most popular type of physical activity among adults in the United States,28 did not appear to be very well represented. In addition, contemporary guidelines tend to emphasize moderate vs vigorous aerobic exercise.29,30 A second potential limitation, germane to any work that relies solely on the results of published studies, is publication bias, that is, the tendency for authors to not submit, and journal editors to not publish, studies that yield negative or null results. A third potential limitation may have been our strict inclusion criteria that resulted in only 3 studies being included from the meta-analysis in subjects with type 2 diabetes16 as well as the inclusion of only 3 studies in the original meta-analysis on overweight/obesity because of the <15% loss to follow-up criterion imposed by the authors of the original meta-analysis.20 While we have seen meta-analyses published with as few as 3 studies,31,32 our findings may not be very robust when applied beyond the characteristics of our included studies. A fourth possible limitation may have been the fact that some subjects in the included meta-analyses were taking various medications to treat their particular condition. Consequently, their responses to aerobic exercise may have differed when compared with those subjects who were not taking any medications or when compared with subjects who were taking different medications. Other potentially confounding variables include a wide age range, some meta-analyses weighted to one gender, and co-morbid conditions. Finally, by including data limited to such groups as hypertensive adults, we would expect greater reductions in SBP when compared with those with lower baseline values such as normotensives (SBP/DBP <120/80 mm Hg) and prehypertensives (SBP/DBP 120 to 139/80 to 89 mm Hg). In conclusion, the results of our analysis reinforce the importance of aerobic exercise as a nonpharmacologic intervention for improving CHD risk factors in adults.