Effect of Aerobic Exercise on Blood Pressure

Context Exercise can lower blood pressure, but by how much and in whom? Contribution This meta-analysis of 54 trials showed that previously sed-entary adults could decrease systolic blood pressure by 3.8 mm Hg (95% CI, 2.7 to 5.0 mm Hg) and diastolic blood pressure by 2.6 mm Hg (CI, 1.8 to 3.4 mm Hg) with regu-lar aerobic exercise. Exercise lowered blood pressure in people who were normotensive or hypertensive; overweight or of normal weight; and black, white, or Asian. All frequencies, intensities, and types of aerobic exercise lowered blood pressure. Cautions Trials lasting longer than 6 months showed smaller reduc-tions in blood pressure, perhaps because of difficulties in sustaining regular exercise. The Editors High blood pressure, which affects nearly 50 million Americans, is a serious public health challenge for the United States (1). Cardiovascular disease has been the leading cause of death in the United States for more than 80 years. It is estimated that more than $320 billion is spent annually on the approximately 60 million Americans with cardiovascular disease, for which high blood pressure is an important modifiable risk factor (2). Epidemiologic studies indicate that uncontrolled elevated blood pressure leads to stroke, coronary heart disease, congestive heart failure, and end-stage renal disease (3). Clinical trials have demonstrated that lowering blood pressure reduces incidence of and death from cardiovascular disease (3, 4). These studies also indicate that a decrease of as little as 2 mm Hg in mean diastolic blood pressure in the general population could substantially reduce the risk for disease associated with elevated blood pressure (5). Physical inactivity is a major risk factor for cardiovascular disease, and persons who are less active and less fit have a 30% to 50% greater risk for high blood pressure (2). Several recent clinical trials have demonstrated that physical activity reduces blood pressure in hypertensive and normotensive persons, independent of weight loss (6-9). However, evidence regarding the magnitude of exercise-related reductions in blood pressure is inconsistent, both in general and among subgroups of the population. Pooling results from individual clinical trials provides more precise and accurate information on the effect of aerobic exercise on blood pressure and allows exploration of variation in intervention effect among subgroups of interest. Methods Study Selection We conducted a comprehensive literature search of MEDLINE (1966 to September 2001) using the Medical Subject Headings exercise, physical fitness, hypertension, and blood pressure and the keywords physical activity and aerobic exercise. We also searched the SPORTDiscus database using the same strategy. In addition, we conducted a manual search by examining reference lists from original research papers and review articles. Initially, three of the authors used predetermined selection criteria to identify and independently review 104 original research reports that included 121 trials. Disagreements were resolved by discussion and, when necessary, by deliberation with a fourth investigator. We included studies that were published in English-language journals; were conducted in persons at least 18 years of age; randomly assigned patients to intervention and concurrent control groups; limited differences between treatment and control groups to aerobic physical activity; lasted for at least 2 weeks; and reported changes in blood pressure (systolic, diastolic, or both) from baseline to follow-up, as well as variances or data to estimate them. Adherence to the program of physical activity and sample size were not defined as inclusion criteria, but their influence on blood pressure reduction was identified as an issue to be investigated. Fifty-four trials (from 38 reports) met the eligibility criteria and were included in the meta-analysis (10-47). Sixty-seven trials did not meet the eligibility criteria for the meta-analysis (Figure 1). The main reasons for exclusion were nonrandomized assignment of the trial participants (13 trials), lack of an appropriate concurrent control group (18 trials), a difference other than aerobic exercise between the active treatment and control groups (13 trials), inclusion of participants younger than 18 years of age (4 trials), a follow-up period of less than 2 weeks (12 trials), and lack of data on blood pressure outcome (7 trials). Results from these trials were not included in the current meta-analysis; a list of excluded clinical trials can be requested from the corresponding author. In addition, 1 trial was missing data on systolic blood pressure and 4 trials were missing data on diastolic blood pressure. These trials, however, were included in analyses of diastolic blood pressure and systolic blood pressure, respectively. Figure 1. Study selection Data Abstraction Three of the authors used a standardized protocol and reporting form to independently abstract data on characteristics of trial participants and study design, intervention method and duration, and study outcomes. Where unstated, ethnicity was assumed to be white if the study was conducted in Europe, Australia, or New Zealand; Asian if the study was conducted in Asia; and black if the study was conducted in Africa. Differences in duplicate data extraction among the primary reviewers were resolved by discussion and, where necessary, by obtaining additional input from a fourth author. Statistical Analysis The mean baseline body weight and blood pressure for each trial were calculated by combining mean values from the intervention and control groups, weighted by the number of participants. This mean was not used for calculations of net changes in weight or blood pressure. For parallel and factorial trials, net changes in blood pressure (BP) were calculated as (BP at the end of follow-up in the intervention group BP at baseline in the intervention group) (BP at the end of follow-up in the control group BP at baseline in the control group). For crossover and Latin-square trials, net changes in blood pressure were calculated as BP at the end of the intervention period BP at the end of the control period. To pool the overall effect size, we weighted each study by the reciprocal of the total variance for blood pressure change (separate values for systolic and diastolic blood pressure). Variances for net changes in blood pressure were calculated by using confidence intervals, P values, t-statistics, or individual variances for intervention and control groups (parallel and factorial design) or intervention or control periods (crossover and Latin-square design). For parallel trials that reported variance for paired differences separately for each group, we used standard methods to calculate the pooled variance for net change (48). If the variance for paired differences was not reported, it was calculated by using variances at baseline and at the end of the trial. We used the method of Follmann and colleagues (49), in which a correlation coefficient of 0.5 between initial and final values is assumed. Within each trial, equal variance was assumed between the control and intervention groups, as well as between the beginning and end of the trial. Fixed-effects and DerSimonian and Laird random-effects models (50) were used to calculate the estimated mean effect of aerobic exercise on blood pressure and associated 95% CIs. Although both models yielded similar results, we chose the random-effects model to present the results because the trials had significant heterogeneity in effect size (50) and were conducted among participants of different ethnic backgrounds, sexes, and hypertensive status, as well as other important covariables. We performed a series of prestated subgroup analyses to examine the influence of covariables. The subgroups were chosen on the basis of biological plausibility and knowledge of previous studies on the relationship between exercise and blood pressure. For each subgroup, pooled effects were calculated by using the random-effects model and statistical significance was tested by using one-way analysis of variance, weighted by the reciprocal of the total variance for change in blood pressure. Multivariate meta-regression analysis was not performed because many trials did not report important covariables, such as hypertensive status and ethnicity. We examined the potential for publication bias by using a funnel plot, in which sample size was plotted against net change in blood pressure (51). In addition, a nonparametric trim and fill method was used to test and adjust for potential publication bias (52, 53). This method may be used to estimate the number of missing studies that might exist in a meta-analysis and the influence that the missing studies might have had on the estimates of overall effect size. We used Stata, version 8.0 (Stata Corp., College Station, Texas), for the trim and fill method and SAS, version 8 (SAS Institute, Inc., Cary, North Carolina), for all other analyses. Role of the Funding Source The funding source had no role in the collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication. Results Characteristics of the Participants and Study Designs Characteristics of the 54 trials and their participants are shown in Appendix Table 1. The trials were conducted between 1986 and 2000 and varied in size from 8 to 247 participants (median, 28 participants). Overall, 2419 participants were evaluated, but 39 were included twice or three times in separate protocols (10, 11, 28). All trials were conducted in adults (mean age, 21 to 79 years). Of the 51 trials that reported sex distribution, 10 included predominantly ( 80%) men and 17 included predominantly women. Among the 37 trials that reported distribution of ethnicity, all or most of the participants ( 80%) were white in 23 trials, Asian in 6 trials, and black in 4 trials. Fifteen of the 47