Walking and resting blood pressure: An inter-individual response difference meta-analysis of randomized controlled trials

Abstract

Whether true inter-individual response differences (IIRD) exist with respect to walking training and chronic changes in resting systolic blood pressure (SBP) and diastolic blood pressure (DBP) in adults is not known. To address this gap, data from a meta-analysis representing up to 5060 normotensive and high normal/hypertensive (SBP ≥ 130 and/or DBP ≥ 85 mmHg) participants 16 to 84 years of age (2881 walking, 2179 control) nested in 73 randomized controlled trials were included. Walking and control group change outcome standard deviations treated as point estimates for both resting SBP and DBP were used to calculate true IIRD from each study. The inverse variance heterogeneity (IVhet) model was used to pool IIRD as well as traditional pairwise results. Both 95% confidence intervals (CI) and prediction intervals (PI) were calculated. While statistically significant reductions in resting SBP ( $\bar{X}$ , −3.9 mmHg, 95% CI, −5.4 to −2.3 mmHg) and DBP ( $\bar{X}$ , −1.4 mmHg, 95% CI, −2.5 to −0.3 mmHg) were found, true IIRD were neither statistically significant nor clinically important for both SBP ( $\bar{X}$ , −1.4 mmHg, 95% CI, −2.0 to 2.8 mmHg) and DBP ( $\bar{X}$ , 0.9 mmHg, 95% CI, −2.5 to 2.8 mmHg). The 95% prediction interval for true IIRD was −2.1 to 2.8 mmHg for SBP and −3.2 to 3.4 mmHg for DBP.

Conclusions

While walking is associated with reductions in resting SBP and DBP, a lack of true IIRD exists, suggesting that factors other than training response variation (random variation, physiological responses associated with behavioral changes that are not the result of walking) are responsible for the observed variation in resting SBP and DBP.

Keywords

Walking blood pressure hypertension adults meta-analysis

Introduction

Hypertension has been reported to be the major cause of all-cause mortality and disability worldwide.^1,2 In 2015, the global prevalence of elevated blood pressure (BP) in adult men and women was estimated to be 24.1% and 20.1%, respectively,³ with an increase from 594 million to 1.13 billion adults between the years 1975 and 2015.³ In addition, the worldwide incidence of hypertension is expected to increase to 60% by the year 2025.⁴ With respect to costs, a study published in 2009 reported that over a 10-year period, the future global costs associated with elevated BP could be nearly $1 trillion.⁵

One potential nonpharmacologic therapy for reducing resting systolic blood pressure (SBP) and diastolic blood pressure (DBP) is walking, a low-impact, low-cost activity that is generally safe and available for most adults, including older adults and those with overweight and obesity. In the United States (US), walking is the most commonly reported type of physical activity for both women and men,⁶ with recent estimates reported to be 31.7% for transportation walking and 52.1% for leisure-time walking.⁷

A recent Cochrane Database systematic review with meta-analysis of 73 randomized controlled trials that included normotensive and hypertensive participants 16 years of age and older reported that walking reduces resting SBP by 4.1 mmHg (73 studies, 5060 participants, 95% CI, −5.2, −3.0 mmHg) and resting DBP by 1.8 mmHg (69 studies, 4711 participants, 95% CI, −2.5, −1.1 mmHg).⁸ No statistically significant differences were observed when results were partitioned according to age, sex, and BP status (normotensives versus hypertensives).⁸ Other meta-analyses have yielded similar overall results.^9–11

The mechanisms associated with changes in resting BP, both acute and chronic, as a result of activities such as walking, include, but are not necessarily limited to, an immediate decrease in nitric oxide mediated endothelial function¹² and decreased sympathetic drive.¹³ Along those lines, chronic aerobic training has been shown to improve endothelial function of conduit¹⁴ and resistance vessels¹⁵ suggestive of a BP-lowering effect. In addition, other potential mechanisms include decreases in systemic vascular resistance, plasma norepinephrine, and activity of the renin-angiotensin system.¹⁶ With respect to genetics, a narrative review by Rankinen and Bouchard suggested that there is a genetic component that modifies the BP response to endurance training, possibly as a result of DNA sequence variation in hypertensive candidate genes.¹⁷ Along those lines, a systematic review with meta-analysis addressing the BP response to acute and chronic aerobic exercise concluded that only one genetic variant, angiotensinogen M235T (rs699), was associated with changes in DBP after training.¹⁸ However, this accounted for <1% of the variance.¹⁸ No candidate genes were associated with changes in resting SBP.¹⁸

Precision medicine, defined by some as the use of prevention and treatment approaches that account for individual variability,^19,20 has been recommended as an important approach in the treatment and management of hypertension in adults.^21–25 However, one of the assumptions of this approach is that true inter-individual response differences (IIRD) are the result of the treatment itself versus other factors such as random variation (biological day-to-day variation, measurement error) and/or the physiological responses associated with behavioral changes that are not the result of a treatment such as walking (sleep, diet, etc.).^26–28 Distinguishing whether true IIRD are associated with walking and changes in resting BP is crucial before examining for possible moderators and mediators, including potential genetic interactions.²⁸ Not examining for such could lead to false conclusions as well as a waste of time and resources searching for potential moderators and mediators, including genetic interactions.^28,29 From an applied perspective, the potential deleterious consequences of such include a reliance on personalized recommendations versus broader, evidence-based, public-health oriented recommendations that may have the greatest reach.

To the best of the investigative team's knowledge, no previous study has examined for true IIRD associated with walking training and chronic changes in resting SBP and DBP in humans. Therefore, the purpose of this study was to fill this gap by determining whether true IIRD associated with walking interventions and chronic changes in resting SBP and DBP in humans exist.

Methods

Overview

The protocol for this study was registered in Open Science Framework (https://osf.io/6xp8e/) but has not been published in a peer-reviewed journal. Where applicable, we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.³⁰

Data source

For the current study, data were derived from a recent Cochrane Database systematic review with meta-analysis of randomized controlled trials that included normotensive and hypertensive (SBP ≥ 130 and/or DBP ≥ 85 mmHg) participants 16 years and older addressing the effects of walking training on chronic changes in resting SBP (73 studies, 5060 participants) and DBP (69 studies, 4711 participants).⁸ Details of the study are provided in the original meta-analysis.⁸ A general description of the studies included in this previous systematic review with meta-analysis is described here versus in the Results section of the manuscript because this information was previously collected and/or reported by the authors of the original systematic review with meta-analysis.⁸ Briefly, the included studies were conducted in 22 different countries and included males and females 16 to 84 years with 51% greater than 60 years of age and 39% 41 to 60 years of age.⁸ Interventions ranged from 4 to 64 weeks, 3 to 5 times per week, for 20 to 40 min per session, while the total minutes of walking ranged from 10 to 845 per week.⁸ The majority of studies reported moderate-intensity walking.⁸ For those studies that reported data, the majority of walking interventions took place in the home/community (50 studies) while 16 took place in a laboratory.⁸ Of those studies reporting information, 36 of 47 (76.6%) reported supervised walking.⁸

The rationale for using this existing meta-analysis⁸ versus conducting an updated systematic review with meta-analysis was based on (1) the recency of this work (2021),⁸ (2) criticism regarding the publication of redundant systematic reviews with meta-analysis,³¹ and (3) determination that an updated review was not necessary based on decision tree analysis according to procedures recommended by the Panel for Updating Systematic Reviews (PUGs).³²

Data extraction

Data from the included meta-analysis⁸ were extracted by the first two authors, independent of each other, using Microsoft® Excel® for Microsoft 365 MSO (16.0.13801.20442). Broadly, extracted data included (1) original study authors, (2) year of publication, (3) sample sizes, change outcome means and standard deviations in mmHg for both resting SBP and DBP for walking and control groups, and (4) treatment effect changes (walking minus control) and 95% confidence intervals (CIs) for both resting SBP and DBP. After extracting data, the first two authors met and reviewed their coding for agreement. Any disagreements were resolved by consensus. If consensus could not be achieved, the a priori plan was to have the third author render a recommendation. However, this was not necessary. Prior to meeting, Gwet's AC1 statistic was used to assess inter-rater agreement.^33,34

Risk of bias of included meta-analysis

The Assessment of Multiple Systematic Reviews (AMSTAR 2) instrument³⁵ was used to assess the risk of bias of the included systematic review with meta-analysis.⁸ Details regarding the AMSTAR 2 instrument are provided in the original article.³⁵ In brief, this instrument includes 16 questions for randomized controlled trials of healthcare interventions.³⁵ Response options, dependent on the question, consist of “Yes”, “No”, “Partial Yes” or “No-meta-analysis conducted”.³⁵ The selection of “Yes” and “Partial Yes” indicate that the item was adequately addressed.³⁵ While not specifically designed to produce an overall score, the following categories have been suggested in relation to overall confidence in the outcomes of the review: “High” (no or one non-critical weakness), “Moderate” (more than one non-critical weakness), “Low” (one critical flaw with or without non-critical weaknesses), or “Critically low” (more than one critical flaw with or without non-critical weaknesses).³⁵ AMSTAR2 assessments were performed using the same methods as for data extraction. Prior to meeting, Gwet's AC1 statistic was used to assess inter-rater agreement.^33,34

Data synthesis

Effect size metric

The original metric (mmHg) was used as the effect size for both resting SBP and DBP. Given the primary purpose of the study, and to avoid dependent effects from different treatment groups in the same study, multiple treatment arms, i.e., walking arms, in the same study were pooled so that only one effect size represented each study.

Traditional meta-analysis

Prior to conducting an IIRD meta-analysis, a traditional meta-analysis was performed by pooling treatment effects from each study. This was accomplished by computing separate change outcome differences in resting SBP and DBP between walking and control groups along with their standard deviations. Results were then pooled using the inverse variance heterogeneity (IVhet) model, a quasi-likelihood model that allows for the inclusion of heterogeneity into the analysis and has been shown to be more robust and theoretically reasonable than random-effects models.^36,37 Negative values were indicative of decreases in resting SBP and DBP favoring walking.

Heterogeneity was assessed using the Cochran Q statistic while I-squared (I²) was used to assess inconsistency.³⁸ Alpha values ≤0.10 for Q were judged as statistically significant heterogeneity while inconsistency based on I² values was classified as either very low (<25%), low (25% to <50%), moderate (50% to <75%), or high (>75%).³⁸ Tau ( $τ$ ), an absolute measure of between-study heterogeneity, was also calculated.³⁹

The Doi plot and Luis Furuya-Kanamori (LFK) index were used to assess small-study effects (publication bias, etc.).⁴⁰ The Doi plot has been reported to be more intuitive than the commonly used funnel plot while the LFK index has been reported to be more robust than Egger's regression intercept test.⁴⁰ Based on previously recommended cutpoints, LFK values <±1, between ±1 and ±2, and >±2, were used to indicate no, minor, and major asymmetry, respectively.⁴⁰ Outlier analysis was also conducted by excluding results for those effect sizes in which their 95% CI fall completely outside the pooled 95% CI. In addition, cumulative meta-analysis, ranked by year, was conducted to examine the accumulation of results over time.⁴¹

IIRD meta-analysis

For the primary aim of the current study, true IIRD between walking and control group standard deviations for resting SBP and DBP from each study were treated as point estimates and calculated as follows⁴²:

\sqrt{S D_{w}^{2} - S D_{c}^{2}}

where

{SD}_{w}^{2}

represents the standard deviation for the walking group and

S D_{c}^{2}

represents the standard deviation for the control group. The standard error of the variance for all point estimates was then calculated as follows^39,42:

S E = \sqrt{2 (\frac{S D_{w}^{4}}{D F_{w}} + \frac{S D_{c}^{4}}{D F_{c}})}

where DF represents the degrees of freedom (N−1) for the standard deviations of the walking and control groups. Negative values for IIRD were suggestive of greater variability for changes in resting SBP and DBP in the control versus walking groups.^39,42 Results were then pooled by combining individual response variances and their standard errors into one overall point estimate and 95% CI using the IVhet model.³⁶ The SD for point estimates and 95% CI were then calculated by taking the square root of each.³⁹ If the lower 95% CI was negative, the sign was first ignored, the square root calculated, and the sign reapplied. Absolute between-study heterogeneity was estimated using tau (

τ

).³⁹ In addition, outlier analysis was also conducted by excluding results for those effect sizes in which their 95% CI fall completely outside the pooled 95% CI.

Additional analyses

To provide a range of expected effects if a new study was performed, 95% prediction intervals (PIs) based on the pooled mean effect, standard error, and $τ$ were estimated for the results from both traditional and IIRD meta-analyses.⁴³ In addition, the magnitude of mean treatment effect changes for both resting SBP and DBP as well as IIRD were compared to a minimally clinically important difference (MCID) of 2 mmHg. A MCID of 2 mmHg was chosen based on prior studies showing that a reduction of 2 mmHg in resting SBP can decrease the risk of mortality from coronary heart disease, stroke, and all-causes by 4%, 6% and 3%, respectively.⁴⁴ Furthermore, Hardy et al. estimated that a 2 mmHg reduction in population-wide SBP can decrease the number of events per 100,000 person-years from coronary heart disease, stroke, and heart failure by 13.5, 12.1, and 20.3 in African-Americans and 17.9, 9.6 and 26.6 in Whites.⁴⁵

For resting DBP, data from the Framingham Heart Study demonstrated that a 2 mmHg population-wide decrease in resting DBP was associated with a 17% decrease in hypertension, 6% decrease in coronary heart disease, and 15% decrease in the risk of stroke and transient ischemic attacks.⁴⁶ In addition, the Blood Pressure Lowering Treatment Trialists’ Collaboration reported that a reduction of 2 mmHg in resting DBP was associated with relative risk reductions of 28% for stroke, 20% for coronary heart disease, 18% for heart failure, 22% for major cardiovascular events, 20% for cardiovascular death, and 12% for total mortality.⁴⁷

The following recommended probabilistic anchors were used to interpret the clinical importance of results: (1) <0.5% (most unlikely or almost certainly not), (2) 0.5 to 5% (very unlikely), (3) 5% to 25% (unlikely or probably not), (4) 25% to 75% (possibly), (5) 75% to 95% (likely or probably), (6) 95% to 99.5% (very likely), (7) >99.5% (most likely or almost certainly).⁴⁸

No analyses to test for potential covariates were conducted. Our rationale was based on the fact that for aggregate data meta-analysis, such analyses are considered to be observational in nature, and thus, would not support causal inferences.^49,50 Thus, such findings would need to be tested in original randomized controlled trials. In addition, meta-regression in an aggregate data meta-analysis has been criticized for being at a high risk for ecological bias.⁵¹ Furthermore, the original meta-analysis from which data for this study was derived found no statistically significant differences in resting SBP and DBP when results were partitioned according to age (SBP, p = 0.88; DBP, p = 0.18), sex (SBP, p = 0.67; DBP, p = 0.91), and BP category (SBP < 130 mmHg versus ≥130 mmHg, p = 0.45; DBP < 85 mmHg versus ≥85 mmHg, p = 0.65).⁸

Software used for analysis

Microsoft^® Excel^® for Microsoft 365 MSO (16.0.13801.20442), Meta XL (version 5.3),⁵² and the most recent user-written versions of metan_model and KAPPAETC within Stata (version 16)⁵³ were used for all analyses. While two-tailed alpha values ≤0.05 were considered statistically significant, the focus was on 95% CIs and PIs.

Results

Risk of bias

Results for AMSTAR 2 are shown in Supplementary file 1. Using Gwet's AC1 statistic, the overall agreement rate prior to correcting one disagreement was 0.93 (95% CI, 0.78 to 1.0). Final agreed-upon ratings for all 16 items were either “Yes” (n = 15) or “Partial Yes” (n = 1). Overall confidence in the systematic review with meta-analysis was considered “High”.

Data abstraction

A total of 1562 items were extracted from the original meta-analysis.⁸ Using Gwet's AC1 statistic, the overall agreement rate prior to correcting six discrepancies was 0.9996 (95% CI, 0.993 to 0.999).

Treatment effect results for SBP and DBP

Treatment effect changes in resting SBP

Mean between-study baseline SBP ranged from 107 to 155 mmHg in the walking groups ( $\bar{X}$ ± SD, 131.3 ± 10.8 mmHg) and 108 to 155 mmHg in the control groups ( $\bar{X}$ ± SD, 132.1 ± 11.4 mmHg). Overall pooled treatment effect changes (walking minus control) for resting SBP are shown in Table 1 while study-level results can be found in Figure 1. As can be seen, statistically significant reductions (p < 0.001) equivalent to approximately 2.9% (95% CI, 1.8% to 4.1%) from baseline values, were observed. Statistically significant heterogeneity and moderate inconsistency were found while no small-study effects were observed (Supplementary file 2). Absolute between-study heterogeneity ( $τ$ ) was 3.2. The 95% PI included 0 (Table 1). The probability (% chance) that reductions in resting SBP would be greater than a MCID of 2 mmHg in a future trial was approximately 71% (possibly clinically important). Cumulative meta-analysis, ranked by year, showed that reductions in resting SBP have been statistically significant and remained relatively stable since the year 1996 (Supplementary file 3).

Figure 1.

Treatment effect changes in resting systolic blood pressure (SBP). The black filled squares, sized according to the weight contributing to the overall effect, represent changes in SBP from each study while the left and right extremes of the squares represent the lower and upper 95% confidence intervals for changes in SBP from each study. The black diamond represents the pooled mean change in resting SBP while the left and right extremes of the diamond represent the pooled lower and upper 95% confidence intervals for changes in SBP. The dashed vertical line represents the pooled mean effect for changes in SBP while the solid vertical line represents the zero (0) point.

Table 1.

Treatment effects and standard deviations for individual responses.

Variable	Studies (#)	Participants (#)	$\bar{X}$ (95% CI)^a	95% PI^b
TE^c (All)
SBP^d (mmHg)	73	5060	−3.9 (−5.4, −2.3)*	−10.4, 2.6
DBP^e (mmHg)	69	4711	−1.4 (−2.5, −0.3)*	−5.5, 2.7
TE (outliers deleted)
SBP (mmHg)	65	4340	−3.7 (−4.6, −2.8)*	−6.2, −1.2*
DBP (mmHg)	66	4573	−1.1 (−2.0, −0.1)*	−4.0, 1.9
SD_IR^f (All)
SBP (mmHg)	73	5060	1.4 (−2.0, 2.8)	−2.1, 2.8
DBP (mmHg)	69	4711	0.9 (−2.5, 2.8)	−3.2, 3.4
SD_IR (outliers deleted)
SBP (mmHg)	71	4963	1.5 (−1.9, 2.9)	−1.9, 2.9
DBP (mmHg)	66	4493	0.8 (−0.9, 1.4)	−0.9, 1.4

Notes: ^a95% CI, 95% confidence interval.

95% PI, 95% prediction interval.

TE, treatment effects.

SBP, systolic blood pressure.

DBP, diastolic blood pressure.

SD_IR, standard deviation of individual responses differences.

*, confidence intervals do not include zero (0).

When outliers (n = 8) were deleted from the model, results remained statistically significant (p < 0.001, Table 1), with no statistically significant heterogeneity (Q = 73.6, p = 0.19) and very low inconsistency overall (I², 13.1%, 95% CI, 0 to 36.7%). Absolute between-study heterogeneity ( $τ$ ) was 1.2. Relative reductions were equivalent to approximately 2.9% (95% CI, 2.2% to 3.6%) from baseline values. The 95% PI did not include zero (Table 1). The probability (% chance) that reductions in resting SBP would be greater than a MCID of 2 mmHg in a future trial was approximately 92% (likely or probably clinically important). Qualitative examination did not yield any potential reason(s) for outlier studies being substantively different than those that remained in the analysis.

IIRD for resting SBP

For the primary purpose of the current study, overall pooled IIRD differences (walking minus control) for resting SBP are shown in Table 1. As can be seen, the overall 95% CI included zero. Absolute between-study heterogeneity ( $τ$ ) was 0. The 95% PI also included 0. The probability (% chance) that IIRD as a result of the walking intervention would be greater than a MCID of 2 mmHg in a future trial was approximately 62% (possibly clinically important).

When two outliers were deleted from the model, both the 95% CI and 95% PI included 0 (Table 1). Absolute between-study heterogeneity ( $τ$ ) was 0. The probability (% chance) that IIRD as a result of the walking intervention would be greater than a MCID of 2 mmHg in a future trial was approximately 65% (possibly clinically important). Qualitative investigation did not yield any possible reason(s) for outlier studies being significantly different from those that remained in the analysis.

Treatment effect changes in resting DBP

Mean between-study baseline DBP ranged from 63 to 104 mmHg in the walking groups ( $\bar{X}$ ± SD, 80.5 ± 7.2 mmHg) and 63 to 105 mmHg in the control groups ( $\bar{X}$ ± SD, 80.8 ± 7.6 mmHg). Overall pooled treatment effect changes (walking minus control) for resting DBP are shown in Table 1 while study-level results can be found in Figure 2. As can be seen, statistically significant reductions (p = 0.02) equivalent to approximately 1.7% (95% CI, 0.3% to 3.1%) from baseline values were observed. Statistically significant heterogeneity and moderate inconsistency were observed while the risk for small study effects was considered minor (Supplementary file 4). Absolute between-study heterogeneity ( $τ$ ) was 2.0. The 95% prediction interval included 0 (Table 1). The probability (% chance) that reductions in resting DBP would be greater than a MCID of 2 mmHg in a future trial was approximately 38% (possibly clinically important). Cumulative meta-analysis, ranked by year, showed that reductions in resting DBP have been statistically significant since the year 2001, with a trend towards wider CIs to the zero-based line of no benefit since 2018 (Supplementary file 5).

Figure 2.

Treatment effect changes in resting diastolic blood pressure (DBP). The black filled squares, sized according to the weight contributing to the overall effect, represent changes in DBP from each study while the left and right extremes of the squares represent the lower and upper 95% confidence intervals for changes in DBP from each study. The black diamond represents the pooled mean change in resting DBP while the left and right extremes of the diamond represent the pooled lower and upper 95% confidence intervals for changes in DBP. The dashed vertical line represents the pooled mean effect for changes in DBP while the solid vertical line represents the zero (0) point.

When outliers (n = 3) were deleted from the model, results remained statistically significant (p = 0.02, Table 1). Statistically significant heterogeneity (Q = 101.8, p = 0.002) but low inconsistency overall (I², 36.2%, 95% CI, 13.7 to 52.8%) was observed. Absolute between-study heterogeneity ( $τ$ ) was 1.4. Relative reductions were equivalent to approximately 1.3% (95% CI, 0.2% to 2.4%) from baseline values. The 95% PI included zero (Table 1). The probability (% chance) that reductions in resting DBP would be greater than a MCID of 2 mmHg in a future trial was approximately 28% (possibly clinically important). Qualitative inspection did not yield any possible reason(s) for excluded outlier studies to be any different than those that remained in the analysis.

IIRD for resting DBP

For the primary aim of the current study, overall pooled IIRD differences (walking minus control) for resting DBP are shown in Table 1. As can be seen, the overall 95% CI included zero. Absolute between-study heterogeneity ( $τ$ ) was 2.0. In addition, the 95% PI included 0. The probability (% chance) that IIRD as a result of the walking intervention would be greater than a MCID of 2 mmHg in a future trial was approximately 48% (possibly clinically important).

When outliers (n = 3) were deleted from the model, both the 95% CI and 95% PI included 0 (Table 1). Absolute between-study heterogeneity ( $τ$ ) was 0. The probability (% chance) that IIRD as a result of the walking intervention would be greater than a MCID of 2 mmHg in a future trial was approximately 31% (possibly clinically important). Similar to other outlier analyses, qualitative review did not yield any possible reason(s) for the excluded outlier studies being substantively different than those that remained in the analysis.

Discussion

Overall findings

The primary purpose of the current study was to use the aggregate data meta-analytic approach to examine IIRD on resting SBP and DBP as a result of walking. The overall findings, as well as outlier analysis, suggest a lack of walking associated IIRD on changes in resting SBP and DBP. These results are supported by (1) overlapping 95% CI, (2) overlapping 95% PI, (3) findings that were categorized as only “possibly clinically important”, and (4) a lack of a large amount of inconsistency. Consequently, the variability observed for resting SBP and DBP appears to be the result of factors other than the walking intervention. Broadly, these include random variation (measurement error, biological day-to-day variability) and/or the physiological responses associated with behavioral changes that are not the result of participation in walking (diet, sleep, stress, depression, anxiety, etc.).^26,27 Based on the current findings, the significance of precision exercise with respect to walking for reducing resting SBP and DBP is questionable.²⁸ These findings are similar to the investigative team's previous work in which a lack of IIRD associated with isometric exercise was found for changes in resting SBP and DBP.⁵⁴

While not the primary purpose of the current study, treatment effect reductions were found for both resting SBP and DBP as a result of walking, including when outliers were deleted. However, with the exception of SBP when outliers were deleted, the 95% PI for what one might expect if they conducted their own randomized controlled trial included zero for all other analyses. In addition, while outlier results for changes in resting SBP were considered to be “likely or probably clinically important”, all other results were considered to be only “possibly clinically important”. Thus, it appears that walking may have a greater impact on resting SBP versus DBP. These findings were reinforced by the overall lack of a large amount of inconsistency between studies as well as a lack of small-study effects (publication bias, etc.). In addition, cumulative meta-analysis demonstrated that reductions in resting SBP and DBP have remained statistically significant since 1996 and 2001, respectively. Along those lines, one possible reason for the trend for wider 95% CIs extending towards the line of no benefit (0) since 2018 for changes in resting SBP might be the improved study designs in more recent years. Finally, our overall results for resting SBP and DBP were slightly smaller and with wider 95% CIs than those reported in the original meta-analysis.⁸ This is most likely the result of the more robust IVhet model used to pool results in the current study.^31,32

Implications for research

Several implications for future research are suggested with respect to IIRD, walking, and changes in resting SBP and DBP. First, prior to investigating for potential moderators and mediators, including genetic interactions, future original randomized controlled trials should appropriately assess and report IIRD results. More specifically, for those who plan to examine for potential moderators and mediators a priori, including genetic interactions, it is suggested that they first plan for an IIRD analysis to see if such pre-planned analyses are worth pursuing. For those conducting post hoc analyses, it is recommended that they also examine for IIRD before pursuing an examination for potential moderators, mediators, and genetic interactions. Thus, one would adhere to the following steps: (1) generate one's overall findings, (2) conduct IIRD analyses, and (3) conduct moderator and mediator analysis, including an examination for genetic interactions, only if one's IIRD results suggest that they are worth pursuing. Applied methods for such, and of which are similar to the approach used in the current meta-analysis, have been described elsewhere.⁵⁵ Another potential approach for individual randomized controlled trials is the concurrent repetition of an intervention within the trial.²⁷ However, the additional cost and effort associated with this study design needs to be considered.²⁷ The end result in examining for IIRD is the prevention of false conclusions as well as wasting time and resources.

Second, it is suggested that future aggregate data meta-analyses include an examination for IIRD before embarking on a search for potential moderators and mediators. This may be especially appropriate given recent research demonstrating the greater power associated with examining IIRD using an aggregate data meta-analysis versus data from a single individual trial.⁵⁶ An even more attractive approach to examine IIRD may be through the use of an individual participant data (IPD) meta-analysis,⁵⁷ although the retrieval of IPD may be challenging.^58–61

Third, to aid in conducting IIRD meta-analyses based on randomized controlled trials, it is suggested that original trial investigators report sample sizes as well as baseline, final, and change outcome means and standard deviations for both intervention and control groups. By doing so, this will help minimize potential bias as a result of excluding studies because of missing data.

Finally, an examination for true IIRD as a result of other pharmacologic and nonpharmacologic interventions on resting SBP and DBP, as well as other cardiometabolic outcomes, is recommended. For example, an examination of IIRD with respect to different types of acute and chronic exercise on various cardiometabolic outcomes (BP, lipids and lipoproteins, glycated hemoglobin, etc.) in children, adolescents, and adults, appears to be warranted.

Implications for practice

The current findings suggest that movement away from precision exercise approaches and towards more general recommendations for walking may be more appropriate for reducing resting SBP and DBP. This may be especially true given that precision medicine, while intended otherwise, may marginalize some of the very populations it was intended to address (poor, selected race/ethnicities, etc.) given its cost as well as lack of access.⁶² Therefore, more general guidelines that have the potential for greater reach may be more appropriate. These include 150 min per week of brisk walking (moderate intensity), 75 min per week of fast walking (vigorous-intensity), or some combination of the two.⁶³ However, given the modest reductions observed in the current study, additional non-pharmacological, e.g., reduced sodium intake, as well as pharmacologic therapies may be necessary for those with elevated BP. Along those lines, adherence to the 2020 International Society of Hypertension global hypertension practice guidelines⁶⁴ or 2017 guidelines endorsed by the American Heart Association, American College of Cardiology, and 9 other health organizations, would seem appropriate.⁶⁵

Implications for policy

While policies and initiatives aimed at exercises like walking should be promoted for reducing resting SBP and DBP as well as a number of other physiological and psychological outcomes,⁶⁶ it may be preferable to focus on policies and initiatives that favor general versus precision oriented recommendations for exercises such as walking. For example, Heart Foundation Walking, a nationwide Australian community-based walking program, has been shown to be a low-cost program with significant reach and retention.⁶⁷ In the United States, initiatives such as the Surgeon General's “Step it up” program, aimed at promoting community-based walking, may be promising for similar reasons.⁶⁸ The former notwithstanding, it is important that one considers such factors as minimizing exercise risks, for example, the attenuation of BP increases during exercise, at the level of the individual.

Strengths and potential limitations

The major strength of the current study is that this is the first study, to the best of the authors’ knowledge, to examine for potential IIRD as a result of walking on resting SBP and DBP. This is considered timely and important given a recent report by others stating IIRD as one of the most important topics to address in exercise medicine.⁶⁹ Alternatively, several potential limitations exist. First, although PIs have been recommended by some investigators,^43,70 caution has been suggested given the possibility for poor coverage probabilities.⁷¹ Second, while we used a MCID of 2 mmHg for both resting SBP and DBP, use of a different MCID would have yielded different results for any analyses that included the MCID. Third, no meta-regression analyses to test for potential covariates were conducted. However, it's important to understand that meta-regression is observational in nature, and thus, would not support causal inferences when results are pooled using the aggregate data meta-analytic approach.⁴⁹ Rather, such analyses would need to be conducted in original randomized controlled trials. In addition, meta-regression using aggregate data has been criticized for being at a high risk of ecological bias.⁵¹ Finally, like any meta-analysis, one is beholden to the weaknesses and limitations of the included studies. For example, in the original meta-analysis, the overall certainty of the evidence was considered moderate for reducing resting SBP and low for reducing resting DBP.⁸

Conclusions

Supplemental Material

sj-docx-1-sci-10.1177_00368504221101636 - Supplemental material for Walking and resting blood pressure: An inter-individual response difference meta-analysis of randomized controlled trials

Supplemental material, sj-docx-1-sci-10.1177_00368504221101636 for Walking and resting blood pressure: An inter-individual response difference meta-analysis of randomized controlled trials by George A Kelley, Kristi S Kelley and Brian L Stauffer in Science Progress

Footnotes

Acknowledgements

None

Authors’ contributions

GAK was responsible for the conception and design, acquisition of data, analysis and interpretation of data, drafting the initial manuscript and revising it critically for important intellectual content. KSK was responsible for the conception and design, acquisition of data, and reviewing all drafts of the manuscript. BLS was responsible for the conception and design, interpretation of data and reviewing all drafts of the manuscript. All authors read and approved the final manuscript.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Informed consent/institutional review board approval

The proposed study was an aggregate data meta-analysis of previously reported summary data. Therefore, neither Informed Consent nor Institutional Review Board Approval were required.

Availability of data

All data for this study are available from the corresponding author upon reasonable request.

ORCID iD

George A Kelley

Supplemental material

Supplemental material for this article is available online.

Author biographies

George A Kelley is a Professor and Director of the Meta-Analytic Research Group in the School of Public Health's Department of Epidemiology and Biostatistics at West Virginia University in Morgantown, West Virginia. His research focuses on using the systematic review and meta-analytic approach to examine the effects of exercise and physical activity on health-related outcomes in humans.

Kristi S Kelley is a Research Instructor and member of the Meta-analytic Research Group in the School of Public Health's Department of Epidemiology and Biostatistics at West Virginia University in Morgantown, West Virginia. Her research focuses on using the systematic review and meta-analytic approach to examine the effects of exercise and physical activity on health-related outcomes in humans.

Brian L Stauffer is a Professor of Medicine/Cardiology and Integrated Physiology at the University of Colorado School of Medicine on the Anschutz Medical Campus. He is Chief of the Division of Cardiology at the Denver Health Medical Center. His research focuses on the development, prevention and treatment of atherosclerotic vascular disease and heart failure.

References

GBD 2015 Risk Factors Collaborators. Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet 2016; 388: 1659–1724. 2016/10/14.

World Health Organization. Guideline for the pharmacological treatment of hypertension in adults. Geneva, Switzerland: World Health Organization, 2021.

NCD Risk Factor Collaboration. Worldwide trends in blood pressure from 1975 to 2015: a pooled analysis of 1479 population-based measurement studies with 19·1 million participants. Lancet 2017; 389: 37–55. 2016/11/20.

Kearney

Whelton

Reynolds

, et al. Global burden of hypertension: analysis of worldwide data. Lancet 2005; 365: 217–223.

Gaziano

Bitton

Anand

, et al. The global cost of nonoptimal blood pressure. J Hypertens 2009; 27: 1472–1477. 2009/05/29.

Dai

Carroll

Watson

, et al. Participation in types of physical activities among US adults--National Health and Nutrition Examination Survey 1999–2006. J Phys Act Health 2015; 12(Suppl 1): S128–S140.

Ussery

Carlson

Whitfield

, et al. Transportation and leisure walking among U.S. Adults: trends in reported prevalence and volume, National Health Interview Survey 2005–2015. Am J Prev Med 2018; 55: 533–540.

Lee

Mulvaney

Wong

, et al. Walking for hypertension. Cochrane Database Syst Rev 2021. DOI: 10.1002/14651858.CD008823.pub2

Kelley

Tran

. Walking and resting blood pressure in adults: a meta-analysis. Prev Med 2001; 33: 120–127.

10.

Murtagh

Nichols

Mohammed

, et al. The effect of walking on risk factors for cardiovascular disease: an updated systematic review and meta-analysis of randomised control trials. Prev Med 2015; 72C: 34–43.

11.

Hanson

Jones

. Is there evidence that walking groups have health benefits? A systematic review and meta-analysis. Br J Sports Med 2015; 49: 710–715.

12.

Dawson

Green

Cable

, et al. Effects of acute exercise on flow-mediated dilatation in healthy humans. J Appl Physiol 2013; 115: 1589–1598. 2013/09/14.

13.

Fagard

Cornelissen

. Effect of exercise on blood pressure control in hypertensive patients. Eur J Cardiovasc Prev Rehabil 2007; 14: 12–17.

14.

Pedralli

Eibel

Waclawovsky

, et al. Effects of exercise training on endothelial function in individuals with hypertension: a systematic review with meta-analysis. J Am Soc Hypertens 2018; 12: e65–e75.

15.

Higashi

Sasaki

Kurisu

, et al. Regular aerobic exercise augments endothelium-dependent vascular relaxation in normotensive as well as hypertensive subjects: role of endothelium-derived nitric oxide. Circulation 1999; 100: 1194–1202.

16.

Cornelissen

Fagard

. Effects of endurance training on blood pressure, blood pressure-regulating mechanisms, and cardiovascular risk factors. Hypertension 2005; 46: 667–675.

17.

Rankinen

Bouchard

. Genetics and blood pressure response to exercise, and its interactions with adiposity. Prev Cardiol 2002; 5: 138–144. 2002/07/02.

18.

Bruneau

Jr Johnson

Huedo-Medina

, et al. The blood pressure response to acute and chronic aerobic exercise: a meta-analysis of candidate gene association studies. J Sci Med Sport 2016; 19: 424–431.

19.

Collins

Varmus

. A new initiative on precision medicine. N Engl J Med 2015; 372: 793–795. 2015/01/31.

20.

Denny

Collins

. Precision medicine in 2030—seven ways to transform healthcare. Cell 2021; 184: 1415–1419.

21.

Padmanabhan

Joe

. Towards precision medicine for hypertension: a review of genomic, epigenomic, and microbiomic effects on blood pressure in experimental rat models and humans. Physiol Rev 2017; 97: 1469–1528. 2017/09/22.

22.

Padmanabhan

Dominiczak

. Genomics of hypertension: the road to precision medicine. Nat Rev Cardiol 2021; 18: 235–250. 2020/11/22.

23.

Melville

Byrd

. Personalized medicine and the treatment of hypertension. Curr Hypertens Rep 2019; 21: 13. 2019/02/13.

24.

Kotchen

Cowley

Jr Liang

. Ushering hypertension into a new era of precision medicine. JAMA 2016; 315: 343–344.

25.

Mattson

Liang

. Hypertension: from GWAS to functional genomics-based precision medicine. Nat Rev Neurol 2017; 13: 195–196. 2017/03/07.

26.

Walsh

Bonafiglia

Goldfield

, et al. Interindividual variability and individual responses to exercise training in adolescents with obesity. Appl Physiol Nutr Metab 2020; 45: 45–54. 2019/05/23.

27.

Hecksteden

Kraushaar

Scharhag-Rosenberger

, et al. Individual response to exercise training—a statistical perspective. J Appl Physiol 2015; 118: 1450–1459. 2015/02/11.

28.

Atkinson

Batterham

. True and false interindividual differences in the physiological response to an intervention. Exp Physiol 2015; 100: 577–588. 2015/05/13.

29.

Senn

. Mastering variation: variance components and personalised medicine. Stat Med 2016; 35: 966–977. 2015/09/28.

30.

Page

Moher

Bossuyt

, et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews. Br Med J 2021; 372: n160.

31.

Ioannidis

JPA

. The mass production of redundant, misleading, and conflicted systematic reviews and meta-analyses. Milbank Q 2016; 94: 485–514.

32.

Garner

Hopewell

Chandler

, et al. When and how to update systematic reviews: consensus and checklist. Br Med J 2016; 354: i3507.

33.

Gwet

. Computing inter-rater reliability and its variance in the presence of high agreement. Br J Math Stat Psychol 2008; 61: 29–48. 2008/05/17.

34.

Klein

. KAPPAETC: Stata module to evaluate interrater agreement. 2019.

35.

Shea

Reeves

Wells

, et al. AMSTAR 2: a critical appraisal tool for systematic reviews that include randomised or non-randomised studies of healthcare interventions, or both. Br Med J 2017; 358: j4008. 2017/09/25.

36.

Doi

Barendregt

Khan

, et al. Advances in the meta-analysis of heterogeneous clinical trials I: the inverse variance heterogeneity model. Contemp Clin Trials 2015; 45: 130–138.

37.

Doi

SAR

Furuya-Kanamori

Thalib

, et al. Meta-analysis in evidence-based healthcare: a paradigm shift away from random effects is overdue. Int J Evid Based Healthc 2017; 15: 152–160. 2017/11/15.

38.

Higgins

JPT

Thompson

Deeks

, et al. Measuring inconsistency in meta-analyses. Br Med J 2003; 327: 557–560.

39.

Williamson

Atkinson

Batterham

. Inter-individual differences in weight change following exercise interventions: a systematic review and meta-analysis of randomized controlled trials. Obes Rev 2018; 19: 960–975.

40.

Furuya-Kanamori

Barendregt

Doi

SAR

. A new improved graphical and quantitative method for detecting bias in meta-analysis. Int J Evid Based Healthc 2018; 16: 195–203. 2018/04/06.

41.

Lau

Schmid

Chalmers

. Cumulative meta-analysis of clinical trials builds evidence for exemplary medical care: the Potsdam International Consultation on meta-analysis. J Clin Epidemiol 1995; 48: 45–57.

42.

Hopkins

. Individual responses made easy. J Appl Physiol 2015; 118: 1444–1446. 2015/02/14.

43.

IntHout

Ioannidis

Rovers

, et al. Plea for routinely presenting prediction intervals in meta-analysis. BMJ Open 2016; 6: e010247.

44.

Stamler

Rose

Stamler

, et al. INTERSALT Study findings. Public health and medical care implications. Hypertension 1989; 14: 570–577.

45.

Hardy

Loehr

Butler

, et al. Reducing the blood pressure-related burden of cardiovascular disease: impact of achievable improvements in blood pressure prevention and control. J Am Heart Assoc 2015; 4: e002276. 2015/10/29.

46.

Cook

Cohen

Hebert

, et al. Implications of small reductions in diastolic blood pressure for primary prevention. Arch Intern Med 1995; 155: 701–709. 1995/04/10.

47.

Blood Pressure Lowering Treatment Trialists’ Collaboration. Effects of different blood-pressure-lowering regimens on major cardiovascular events: results of prospectively-designed overviews of randomised trials. The Lancet 2003; 362: 1527–1535.

48.

Hopkins

Marshall

Batterham

, et al. Progressive statistics for studies in sports medicine and exercise science. Med Sci Sports Exerc 2009; 41: 3–13.

49.

Borenstein

. Common mistakes in meta-analysis and how to avoid them. Englewood, NJ: Biostat, Inc., 2019, p.388.

50.

Littell

Corcoran

Pillai

. Systematic reviews and meta-analysis. New York: Oxford University Press, 2008.

51.

Fisher

Carpenter

Morris

, et al.

Meta-analytical methods to identify who benefits most from treatments: daft, deluded, or deft approach?

Br Med J 2017; 356: j573–j573.

52.

Meta XL. 5.3 ed. Queensland, Australia: EpiGear International Pty Ltd, 2016.

53.

StataCorp. Stata statistical software: release 16. College Station, TX: StataCorp LLC, 2019.

54.

Kelley

Stauffer

. Isometric exercise and inter-individual response differences on resting systolic and diastolic blood pressure in adults: a meta-analysis of randomized controlled trials. Blood Press 2021; 30: 310–321.

55.

Swinton

Hemingway

Saunders

, et al. A statistical framework to interpret individual response to intervention: paving the way for personalized nutrition and exercise prescription. Front Nutr 2018; 5: 41. 2018/06/13.

56.

Mills

Higgins

JPT

Morris

, et al. Detecting heterogeneity of intervention effects using analysis and meta-analysis of differences in variance between trial arms. Epidemiology 2021; 32: 846–854. 2021/08/26.

57.

Esteves

Swinton

Sale

, et al. Individual participant data meta-analysis provides no evidence of intervention response variation in individuals supplementing with beta-alanine. Int J Sport Nutr Exerc Metab 2021; 1. DOI: 10.1123/ijsnem.2021-0038

58.

Polanin

. Efforts to retrieve individual participant data sets for use in a meta-analysis result in moderate data sharing but many data sets remain missing. J Clin Epidemiol 2018; 98: 157–159. 2017/12/31.

59.

Kelley

. Retrieval of individual participant data for exercise meta-analyses may not be worth the time and effort. Biomed Res Int 2016; 2016: 1–5.

60.

Kelley

Tran

. Retrieval of individual patient data for an exercise meta-analysis. Am J Med Sport 2002; 4: 350–354.

61.

Duval

Vazquez

Baker

, et al. The collaborative study of obesity and diabetes in adults (CODA) project: meta-analysis design and description of participating studies. Obes Rev 2007; 8: 263–276.

62.

Geneviève

Martani

Shaw

, et al. Structural racism in precision medicine: leaving no one behind. BMC Med Ethics 2020; 21: 17.

63.

Centers for Disease Control and Prevention. Physical activity for arthritis, https://www.cdc.gov/arthritis/basics/physical-activity-overview.html (2018, accessed 7 April 2021).

64.

Unger

Borghi

Charchar

, et al. 2020 International society of hypertension global hypertension practice guidelines. Hypertension 2020; 75: 1334–1357.

65.

Whelton

Carey

Aronow

, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. J Am Coll Cardiol 2018; 71: e127–e248. 2017/11/18.

66.

2018 Physical Activity Guidelines Advisory Committee. 2018 physical activity guidelines advisory committee scientific report. 2nd ed. Washington, DC: U.S. Department of Health and Human Services, 2018.

67.

Ball

Abbott

Wilson

, et al. How to get a nation walking: reach, retention, participant characteristics and program implications of Heart Foundation Walking, a nationwide Australian community-based walking program. Int J Behav Nutr Phys Act 2017; 14: 61.

68.

U.S. Department of Health and Human Services. Step it up! The Surgeon General’s call to action to promote walking and walkable communities. Washington, DC: U.S. Deptartment of Health and Human Services, 2015.

69.

Ross

Goodpaster

Koch

, et al. Precision exercise medicine: understanding exercise response variability. Br J Sports Med 2019; 53: 1141–1153. 2019/03/14.

70.

Higgins

Thompson

Spiegelhalter

. A re-evaluation of random-effects meta-analysis. J R Stat Soc Series A 2009; 172: 137–159.

71.

Partlett

Riley

. Random effects meta-analysis: coverage performance of 95% confidence and prediction intervals following REML estimation. Stat Med 2017; 36: 301–317.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.59 MB