|Home | About | Journals | Submit | Contact Us | Français|
In this study the authors used the meta-analytic approach to examine the effects of aerobic exercise on resting systolic and diastolic blood pressure in adults. Forty-seven clinical trials representing a total of 72 effect sizes in 2543 subjects (1653 exercise, 890 control) met the criteria for inclusion. Statistically significant exercise-minus-control decreases were found for changes in resting systolic and diastolic blood pressure in both hypertensive (systolic, -6 mm Hg, 95% CI, -8 to -3; diastolic, -5 mm Hg, 95% CI, -7 to -3) and normotensive (systolic, -2 mm Hg, 95% CI, -3 to -1; diastolic, -1 mm Hg, 95% CI, -2 to -1) groups. The differences between groups were statistically significant (systolic, p=0.008; diastolic, p=0.000). Relative decreases were approximately 4% (systolic) and 5% (diastolic) in hypertensives, and 2% (systolic) and 1% (diastolic) in normotensives. It was concluded that aerobic exercise reduces resting systolic and diastolic blood pressure in adults.
Recently, the Sixth Report of the Joint National Committee on the Prevention, Detection, Evaluation, and Treatment of High Blood Pressure,1 as well as the 1999 World Health Organization Guidelines for the Management of Hypertension,2 recommended adherence to a program of aerobic exercise for the purpose of lowering resting systolic and diastolic blood pressure in adults. We previously reported meta-analytic work in which aerobic exercise-induced reductions of approximately 4 and 3 mm Hg in resting systolic and diastolic blood pressure, respectively, were found.3,4 However, this work was limited to studies published up to January of 1992. Since that time, a number of additional studies have been conducted and/or located. In addition, more sophisticated and objective approaches have been developed for meta-analysis of clinical trials. Given that the effect of selected interventions on selected outcomes can change over time as additional studies become available, it is critical that concise and up-to-date information be available for clinical practice so that one can identify the true effect of an intervention on the outcome of interest. Meta-analysis is an approach in which individual study findings are combined and analyzed in a concise and quantitative manner.5 The purpose of this study was to use the meta-analytic approach to provide a comprehensive, up-to-date examination of the effects of aerobic exercise on resting systolic and diastolic blood pressure in adult humans.
Computerized literature searches of articles indexed between January, 1966 and December, 1998 were performed with use of MEDLINE, Embase, Current Contents, Sport Discus, and Dissertation Abstracts International databases. The major keywords used were “exercise” and “blood pressure” In addition, the reference lists from both original and review articles were scanned in order to identify any studies that had not been previously identified and appeared to contain information on the topic of interest. An expert on exercise and blood pressure, Dr. James Hagberg, reviewed our reference list for thoroughness and completeness.
Inclusion criteria for this study were as follows: 1) the studies were randomized, parallel-group trials that included a nonexercise control group; 2) aerobic exercise was the only intervention; 3) these subjects were adult humans age 18 and older; 4) the reports were journal articles, dissertations, and masters theses published in the English-language literature; 5) studies were published and indexed between January, 1966 and December, 1998; 6) resting systolic and/or diastolic blood pressure was assessed; and 7) training studies lasted a minimum of 4 weeks. In order to avoid multiple publication bias, all included studies were examined to ensure that each study was independent of all others.
Coding sheets that could hold 246 items were developed and utilized in this investigation. In order to avoid inter-coder bias, all data were independently extracted. The authors then met and reviewed every item for accuracy and consistency. Disagreements were resolved by consensus. The major categories of variables coded included: 1) study characteristics; 2) physical characteristics of subjects; 3) blood pressure assessment characteristics; and 4) exercise program characteristics.
The primary outcomes (effect sizes) in this study were changes in resting systolic and diastolic blood pressure, analyzed separately. Since all studies were parallel trials, net changes in blood pressure were calculated as the difference (exercise minus control) of the changes (initial minus final) in these mean values. Pooled effect sizes were calculated by assigning weights equal to the inverse of the total variance for net changes in blood pressure. Ninety-five percent confidence intervals were used to establish statistical significance. If the 95% confidence interval included zero (0.00) it was concluded that there was no effect of aerobic exercise on resting blood pressure. Heterogeneity of net changes in resting systolic and diastolic blood pressure was examined by use of the Q statistic.6 A random-effects model was used for all analyses.7 In order to examine the influence (sensitivity) of each study on the overall results, analyses were also performed with each study deleted from the model. Publication bias (the tendency for studies to be published that yield statistically significant and positive results) was examined by means of the Kendall tau rank correlation test (rt).8 This consisted of correlating observed outcomes, i.e., changes in resting systolic and diastolic blood pressure, with sample size. In addition, we estimated the number of unpublished outcomes yielding null results that would be needed to reverse our decision that statistically significant decreases in resting systolic and diastolic blood pressure had occurred.9 Study quality was assessed with a three-item questionnaire designed to detect bias—specifically, randomization, blinding, and withdrawals/dropouts.10 The minimum and maximum number of points possible were 0 and 5. All questions were designed to elicit “yes” (1 point) or “no” (0 points) responses. The questionnaire required less than 10 minutes per study. The questionnaire has been shown to be both valid (face validity) and reliable (researcher inter-rater agreement: r=0.77, 95% confidence interval = 0.60–0.86).10 We chose this scale over numerous others11 because it appears to be the most valid and reliable scale currently available and has been successfully used in the past.12 Secondary outcomes, i.e., changes in body weight, percent body fat, maximum oxygen consumption, and resting heart rate, were examined by the same methods as those used to examine net changes in resting systolic and diastolic blood pressure.
For categorical variables as well as study quality, subgroup analyses were performed with ANOVA-like procedures for meta-analysis.6 Net changes in resting systolic and diastolic blood pressure were examined when data were partitioned according to the following study characteristics: source of publication (journal vs. other), country in which the study was conducted (U.S. vs. other), and study quality (<2 compared to ≥2). Subject characteristics were examined when data were partitioned according to gender, antihypertensive medications (yes vs. no), cigarette smoking (yes vs. no), previous physical activity (yes vs. no), and blood pressure category, defined as a resting systolic and/or diastolic blood pressure of ≥140/90 mm Hg (hypertensives) vs. <140/90 mm Hg (normotensives). Blood pressure assessment characteristics were examined when data were partitioned according to type of blood pressure instrument used (electronic compared to manual), use of a mercury vs. a random-zero sphygmomanometer, position of subjects when blood pressure was assessed (sitting vs. supine), and Korotkoff sound at which diastolic blood pressure was recorded (fourth vs. fifth).
To examine the influence of continuous variables on changes in resting systolic and diastolic blood pressure, random-effects least squares simple regression models, calculated with each effect size weighted by the reciprocal of its variance, were used.6 Subject characteristics examined included percent dropout; initial resting blood pressure; age; height; initial, as well as changes in, body weight; percent body fat; maximum oxygen consumption; and resting heart rate. Blood pressure assessment characteristics were limited to the rest period before assessment of resting blood pressure. Training program characteristics included length, frequency, intensity, duration, total minutes (length × frequency × duration), and compliance, defined as the percentage of exercise sessions attended. We were unable to conduct multiple regression analyses because of missing data.
An independent t-test was used to compare differences in study quality between articles published in journals and dissertations. Unless otherwise noted, all data are reported as either mean ±SD or mean and 95% confidence interval (CI). The alpha level for a type I error was set at p<0.05. Bonferroni adjustments were not made because of the increased risk of a type II error.
Fifty-four studies met the criteria for inclusion;13–66 however, we were unable to include seven of these studies because of the inability to obtain missing blood pressure data.18,24,37,42,44,50,66 Thus, the percent loss among studies that met our inclusion criteria was approximately 13% and the final number of studies included in the analysis was 47. The per person time to code each study once ranged from 0.50–4.00 hours (1.14 ± 0.62 hours). Thirty-eight studies were published in journals, and nine were doctoral dissertations. Twenty-eight of the studies were conducted in the U.S., five in Japan, three in Australia, two in Ireland, and one each in Belgium, Canada, Finland, New Zealand, Nigeria, Norway, Sweden, Switzerland, and the U.K. The 47 studies included in the final analysis represented initial and final blood pressure assessment in a total of 2543 subjects (1653 exercise, 890 control). There were 69 exercise and 49 control groups, from which a total of 72 primary outcomes were generated (some studies had more than one group and/or assessed blood pressure in more than one position). The average number of subjects in each group ranged from 6–166 in the exercisers (24 ± 22) and 4–81 in the controls (18 ± 14). Among those groups for which data were available, percent dropout, defined as the number of subjects who did not complete the study, ranged from 0%–60% in the exercise groups (16% ± 17%) and 0%–52% in the control groups (10% ± 14%). Study quality ranged from 1–3 (2 ± 1). There was no statistically significant difference in study quality between journals and dissertations (p=0.17).
Initial subject characteristics for the exercise and control groups are shown in Table I. For those groups that reported such data, the percentage of males ranged from 0%–100% in both the exercise and control groups (exercise, 51%±45%; control, 56%±42%). Subjects in the studies were primarily white, but some studies also included blacks, Hispanics, Asians, Indians, and Japanese. However, insufficient data were available to examine changes in blood pressure according to race. Eleven exercise and nine control groups were reportedly taking antihypertensive medications during the studies. Sixteen exercise and five control groups included some subjects who smoked cigarettes, and 19 exercise and 11 control groups included subjects who consumed alcohol. Only one exercise and one control group reported a change in diet during the study. Only three exercise and one control group included subjects who had been physically active prior to the start of the investigation.
In 28 studies a manual sphygmomanometer was used to assess resting blood pressure, while in six an electronic sphygmomanometer was used. In 26 studies resting blood pressure was measured in the sitting position; eight studies used the supine position, and both sitting and supine positions were used in two. The number of measurements taken to arrive at a mean blood pressure ranged from 3–20. The rest period before assessment of resting blood pressure ranged from 5–15 minutes (10 ± 7 minutes), and the rest period between assessments ranged from 1–5 minutes (3 ± 2 minutes). In 10 studies the fifth Korotkoff sound was used to measure resting diastolic blood pressure, and in two studies the fourth Korotkoff sound was used. The time between the last exercise session before resting blood pressure was assessed ranged from 12–168 hours. The six reports that provided information on the time of blood pressure measurement stated that it was assessed in the morning.
Length of training in these studies ranged from 4–52 weeks (23 ± 15). The frequency ranged from one to five times per week (3 ± 1), the intensity from 45%–86% of Vo2max (67% ± 10%), and the duration from 15–60 minutes per session (40 ± 12). The total minutes of training (length × frequency × duration) ranged from 360–9360 (2978 ± 2158). The primary training modalities included one or more of the following: walking, jogging, cycling, aerobic dance, and swimming. Compliance, defined as the percentage of exercise sessions attended, ranged from 53%–100% (79% ± 14%).
Initial resting systolic blood pressure ranged from 105–169 mm Hg in the exercisers (129 ± 15 mm Hg) and from 103–158 mm Hg in controls (129 ± 14 mm Hg). For resting diastolic blood pressure, the initial values ranged from 65–108 mm Hg in the exercisers (82 ± 10 mm Hg) and from 60–113 mm Hg in controls (81 ± 10 mm Hg). Across all designs and categories, random-effects modeling yielded decreases of approximately 2% and 3% for resting systolic and diastolic blood pressure, respectively (systolic, x=-2, 95% CI=-3 to -1 mm Hg; diastolic, x=-2 mm Hg, 95% CI=-2 to -1). No statistically significant heterogeneity was observed for either resting systolic or diastolic blood pressure (systolic, Q=73.87, p=0.38; diastolic, Q=62.21, p=0.68), When analysis was limited to study results that appeared in journals, possible publication bias was observed for changes in both resting systolic and diastolic blood pressure (systolic, rt=0.22, p= 0.01; diastolic, rt=0.21, p=0.01). However, it was estimated that 434 and 410 unpublished systolic and diastolic outcomes yielding null results would be needed to reverse our findings that aerobic exercise reduces resting systolic and diastolic blood pressure. With each study deleted from the model once, changes ranged from -2 mm Hg (95% CI=-l to -3 mm Hg) to -3 mm Hg (95% CI=-4 to -1 mm Hg) for resting systolic blood pressure and -1 mm Hg (95% CI=-1 to -2 mm Hg) to -2 mm Hg (95% CI=-3 to -1 mm Hg) for resting diastolic blood pressure.
For secondary outcomes, small but statistically significant decreases were found for percent body fat (-1%, 95% CI=-2% to -1%) and resting heart rate (-5 beats per minute, 95% CI=-6 to -3 beats per minute), while a statistically significant increase was found for maximum oxygen consumption (4 ml/kg-l/min-l, 95% CI=3–5 ml/kg-l/min-l). No statistically significant changes were found for body weight.
A statistically significant difference was found for changes in resting systolic blood pressure when data were partitioned according to whether subjects were hypertensive or normotensive (systolic hypertensive: -6 mm Hg, 95% CI=-8 to -3 mm Hg; systolic normotensive: -2 mm Hg, 95% CI=-3 to -1 mm Hg; Qb=6.99, p=0.008). Decreases in resting systolic blood pressure were equivalent to reductions of approximately 4% in hypertensives and 2% in normotensives. A statistically significant difference was also found when changes in diastolic blood pressure were partitioned according to blood pressure category (diastolic hypertensive: -5 mm Hg, 95% CI=-7 to -3 mm Hg; diastolic normotensive: -1 mm Hg, 95% CI=-2 to -1 mm Hg; Qb=22.26, p=0.00). Decreases in resting diastolic blood pressure were equivalent to reductions of approximately 5% in hypertensives and 1% in normotensives. Changes in resting systolic blood pressure were also significantly different when data were partitioned according to source of study (journal: -3 mm Hg, 95% CI=-4 to -2 mm Hg; dissertation: 1 mm Hg, 95% CI=-2 to 5 mm Hg; Q=5.41, p=0.02). No statistically significant differences were found for any of the other subgroup analyses performed.
Associations between changes in resting blood pressure and selected variables are shown in Table II. Larger effect size reductions in both systolic and diastolic blood pressure at rest were associated with higher initial blood pressure levels, older age, larger reductions in resting heart rate, and more minutes of training. For resting systolic blood pressure, larger reductions were associated with shorter stature as well as lower levels of training intensity. For resting diastolic blood pressure, larger reductions were associated with longer rest periods before the assessment of resting blood pressure, larger increases in maximum oxygen consumption, and shorter periods (weeks) of training. However, the proportion of the total variation for changes in resting blood pressure attributable to most of these variables was small. No statistically significant associations were observed for any of the other variables assessed, including changes in body weight and percent body fat.
The results of this study suggest that aerobic exercise reduces resting systolic and diastolic blood pressure in adults and that reductions are greater in hypertensives than in normotensives. These results were independent of changes in body weight and percent body fat.
The fact that the majority of the studies were published in journals and we found an indication of publication bias, as well as the fact that we found greater reductions in resting systolic blood pressure in studies published in journals than in unpublished dissertations, may be thought to have affected our results. However, it is highly unlikely that more than 400 unpublished outcomes yielding null results exist. Thus, we believe that our findings are representative of the effects of aerobic exercise on resting systolic and diastolic blood pressure in humans. These findings corroborate our previous meta-analytic work in both hypertensives and normotensives,3,4 as well as earlier meta-analytic work by others in which reductions of approximately 5/4 mm Hg and 4/2 mm Hg were found for resting systolic and diastolic blood pressure, respectively, in hypertensives and normotensives.67 However, the results of our review are somewhat less dramatic than those of an earlier meta-analysis68 in which reductions in resting systolic and diastolic blood pressure were reported to be approximately 3/3 mm Hg in normotensives, 5/4 mm Hg in borderline hypertensives, and 10/8 mm Hg in hypertensives. One possible reason for this discrepancy is that the latter study included both randomized and nonrandomized trials. It was recently demonstrated that the inclusion of nonrandomized trials yields larger effect sizes than randomized trials.69
We believe that the statistically significant reductions in resting blood pressure observed among both hypertensive and normotensive adults in this study are clinically important. For example, the 6-mm Hg reduction in resting systolic blood pressure found in hypertensives should confer benefit, as it has been reported that a 5-mm Hg reduction in systolic pressure has been estimated to reduce mortality from coronary heart disease, strokes, and all causes by 9%, 14%, and 7%, respectively.70 In addition, the decreases in resting diastolic blood pressure in hypertensives should also be clinically important, in that a reduction of 5 mm Hg in resting diastolic blood pressure has been associated with a 34% reduction in stroke and a 21% reduction in coronary heart disease.71
Although we do not want to underestimate the importance of reducing resting blood pressure in hypertensive adults, it is important to understand that while the relative risk of stroke and coronary heart disease is greater among hypertensive than normotensive adults, the absolute number of deaths from these causes is greater among normotensive adults. For example, MacMahon and Rogers72 reported that the greatest absolute number of strokes occurred in individuals with diastolic blood pressures between 80 and 89 mm Hg. Furthemore, it has been reported71 that the risk of stroke and coronary heart disease is directly related to the level of blood pressure throughout the normotensive and hypertensive range. While one may question the clinical importance of the small reductions in resting blood pressure found among normotensives in this study, it should be noted that the 2-mm Hg reduction in resting systolic blood pressure observed has been associated with reductions in mortality of 4% from coronary heart disease, 6% from stroke, and 3% from all causes.70 Thus, the lowering of blood pressure in both normotensive and hypertensive adults should be of major public health importance.
The association found between initial resting blood pressure and changes in both resting systolic and diastolic blood pressure is supported by the larger reductions in resting blood pressure observed when we partitioned our data according to whether subjects were hypertensive or normotensive. However, the association between older age and larger reductions in resting systolic and diastolic blood pressure may be the result of higher initial blood pressure levels in older individuals and not age per se.73 In addition, our results conflict with a recent study74 in which larger reductions in resting blood pressure were found among younger than older adults. Since a reduction in resting heart rate is one of the basic physiologic adaptations that occurs as a result of aerobic exercise, such changes may be important in predicting changes in resting systolic and diastolic blood pressure. The fact that more minutes of exercise per session was associated with greater reductions in resting systolic and diastolic blood pressure suggests that more minutes of training per session may have the potential to confer greater benefits on resting blood pressure. While we have no plausible explanation for the association between shorter stature and greater reductions in resting systolic blood pressure, it may be that shorter individuals have higher initial resting systolic blood pressure levels. The association between lower levels of training intensity and greater reductions in resting systolic blood pressure suggests that one does not have to train at a high level of intensity in order to reduce resting systolic blood pressure. In support of this contention, Moreira et al.75 recently found that aerobic training programs at 20% and 60% of maximum work capacity have similar effects on ambulatory blood pressure. This may also be the case for resting systolic, and possibly diastolic, blood pressure. Since increases in maximum oxygen consumption are another of the basic physiologic adaptations that occur as a result of aerobic exercise, our results suggest that such increases may help to predict the magnitude of reduction in resting diastolic blood pressure. The association between greater reductions in resting diastolic blood pressure and longer rest periods before the assessment of blood pressure may have to do with the, quality of assessment; in some studies, subjects rested for a longer period of time in order that they be tested in a truly rested state. The association between greater reductions in resting diastolic blood pressure with shorter training protocols may reflect some unknown threshold for the reduction of resting diastolic blood pressure. However, it is important to realize that it is well established that the cessation of aerobic exercise will return resting blood pressure to the levels observed prior to the start of the exercise program. Despite these statistically significant associations, our results should be interpreted with caution because of the small amount of variation accounted for by most of the variables. In addition, the fact that we were unable to conduct any type of multiple regression analyses because of missing data probably resulted in multicollinearity between some of the independent variables (for example, age and initial blood pressure). Furthermore, given the larger number of statistical tests conducted, the possibility exists that some of the statistically significant results found in this study may have been nothing more than the result of chance. Finally, the fact that we included dissertations and masters theses can be questioned on the basis that the students' advisors may have decided not to publish these studies because of concerns over methodology and the execution of the studies in general. In contrast, it may be that these studies were not submitted for publication because they did not yield statistically significant and positive results (publication bias),76 For example, Stern and Simes76 found that approximately 96% of selected psychology journals and 85% of selected medical journals published studies that yielded a statistically significant result,76 The inclusion of unpublished data in a meta-analysis is a policy that is shared by the majority (80%) of methodologists, who feel that such material as dissertations and masters theses should definitely or probably be included in scientific overviews.77
In conclusion, the results of this study suggest that aerobic exercise reduces resting systolic and diastolic blood pressure in both hypertensive and normotensive adults.
This study was supported by a grant from the National Institutes of Health R01 Award HL56893.