Blood pressure targets in patients with coronary artery disease: observations from traditional and Bayesian random effects meta-analysis of randomised trials
- 1Division of Cardiology, New York University School of Medicine, New York, New York, USA
- 2Division of Cardiology, University of Nebraska Medical Centre, Omaha, Nebraska, USA
- 3Division of Cardiology, St. Luke's Roosevelt Hospital, Columbia University College of Physicians & Surgeons, New York, New York, USA
- Correspondence to Dr Sripal Bangalore, Director of Research, Assistant Professor of Medicine, Cardiovascular Clinical Research Center, New York University School of Medicine, The Leon H. Charney Division of Cardiology, New York, NY 10016, USA;
Contributors SB had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: SB and FHM. Acquisition of data: SB, SK and AV. Analysis and interpretation of data: SB and FHM. Drafting of the manuscript: SB, SK, AV and FHM. Critical revision of the manuscript for important intellectual content: SB and FHM. Statistical analysis: SB. Study supervision: SB and FHM. All authors have read and approved the manuscript and the authors' conflicts of interest are listed in the manuscript.
- Received 6 March 2012
- Revised 29 June 2012
- Accepted 17 July 2012
- Published Online First 21 August 2012
Context Most guidelines for treatment of hypertension including the Joint National Committee-7 recommend a blood pressure (BP) goal of <140/90 mm Hg for hypertensive patients and a more aggressive goal of <130/80 mm Hg for patients with coronary artery disease (CAD), based largely on expert consensus.
Objective To evaluate the BP targets in patients with CAD
Data Sources PUBMED, EMBASE and CENTRAL Study Selection: Randomised clinical trials (RCTs) of antihypertensive therapy in patients with CAD, enrolling at least 100 patients, with achieved systolic pressure of <=135 mm Hg in the ‘intensive BP’ group and <=140 mm Hg in the ‘standard BP’ group with follow-up for at least 1 year and evaluating cardiovascular outcomes.
Data Extraction The following efficacy outcomes were extracted- all-cause mortality, cardiovascular mortality, myocardial infarction, stroke, angina pectoris, heart failure and revascularisation.
Results We identified 15 RCTs enrolling 66 504 participants with 276 328 patient-years of follow-up. Intensive BP group (≤135 mm Hg) was associated with a 15% decrease in heart failure rate and 10% decrease in stroke rate, driven largely by trials with a more intensive BP group (≤130 mm Hg), with similar outcomes for death and cardiovascular death and was associated with a 105% increase in the risk of hypotension. More intensive BP group (≤130 mm Hg) was also associated with a reduction in myocardial infarction and angina pectoris. The results were similar in a Bayesian random effects model. In addition, lower seemed to be better (based on regression analysis) for the outcomes of myocardial infarction, stroke, heart failure and perhaps angina.
Conclusions The present body of evidence suggests that in patients with CAD, intensive systolic BP control to ≤135 mm Hg and possibly to ≤130 mm Hg is associated with a modest reduction in stroke and heart failure but at the expense of hypotension. Lower was better, although not consistently so for myocardial infarction, stroke, heart failure and perhaps angina. Further trials are needed to prove these findings.
The seventh report of the Joint National Committee on prevention, detection, evaluation and treatment of high blood pressure (BP) recommends a systolic pressure goal of <140 mm Hg in patients with hypertension and a more aggressive goal of <130 mm Hg in patients at high risk.1 Largely based on expert consensus with scant clinical trial evidence, other major national and international guidelines have echoed this more aggressive BP goal in patients with cardiovascular disease/coronary artery disease (CAD).2 ,3
In the Action to Control Cardiovascular Risk in Diabetes (ACCORD) blood pressure trial (ACCORD BP) at the end of 4.7 years of follow-up, targeting a systolic pressure <120 mm Hg, as compared with <140 mm Hg, did not reduce the rate of fatal and non-fatal major cardiovascular events except stroke.4 However, in ACCORD only a third of the cohort had cardiovascular disease (including peripheral artery disease, stroke/transient ischaemic attacks, CAD) with an even smaller percentage of patients with known CAD. It is therefore unknown whether or not the results of ACCORD are applicable to patients with CAD, the highest risk subgroup, where aggressive strategies such as lower BP targets should prove to be most beneficial (if any). Data from observational studies and subgroup analyses of randomised trials seem to suggest that lower might not always be better for BP in patients with CAD.5–7 In a recent analysis, we have shown that in subjects with diabetes mellitus a target systolic BP goal of 130–135 mm Hg is ideal, with target organ heterogeneity below 130 mm Hg such that there is continued benefit for stroke but not for other outcomes and at the expense of increase in adverse events.8
Our objective was to evaluate target BP goals for subjects with CAD.
We conducted PUBMED, EMBASE and CENTRAL searches using the term ‘coronary artery disease’ in humans from 1990 until February 2012 using the limits ‘randomized controlled trials’. The search criteria were fairly broad to avoid missing studies with a restricted search. We checked the reference lists of review articles, meta-analyses and original studies identified by the electronic searches to find other eligible trials. There was no language restriction for the search. The authors of publications were contacted when results were unclear or when relevant data were not reported.
Eligible trials had to fulfil the following criteria to be included in this analysis: (1) randomised clinical trials (RCTs) of participants with CAD but without heart failure or acute myocardial infarction randomised to antihypertensive agent or placebo; (2) reporting 1 year or longer-term outcomes; (3) enrolling at least 100 patients (to avoid bias associated with small trials); and (4) achieving systolic pressure at the end of follow-up of ≤140 mm Hg in both arms. Additionally, since the objective was to test outcomes based on two BP targets, the following additional criteria were required: (1) the final achieved systolic pressure in the ‘intensive BP’ group ≤135 mm Hg; (2) the final achieved systolic pressure in the ‘standard BP’ group ≤140 mm Hg; and (3) the systolic pressure difference between the intensive and standard BP groups of at least 1 mm Hg. In a sensitivity analysis, we also tested a difference of 3 mm Hg, as was used in our prior analysis.8 We chose this cut point as a difference less than this is likely clinically not relevant and will not result in differential clinical outcomes based on BP alone. Studies where there was no difference in BP between the groups, defined here as those where final BP was ≤140 mm Hg but where there was no difference in BP between the two groups, were excluded. For example, if a study evaluated two antihypertensive agents, but uptitrated or added medication to ensure no difference in final systolic pressures, they were excluded as such studies are not expected to provide information on BP targets.
Selection and quality assessment
Three authors (SB, SK and AV) independently assessed trial eligibility and trial bias risk and extracted data. Disagreements were resolved by consensus. The bias risk of the trials was assessed using the components recommended by the Cochrane Collaboration:9 (1) sequence generation of allocation; (2) allocation concealment; (3) blinding of participants, personnel and outcome assessors; (4) incomplete outcome data; (5) selective outcome reporting; and (6) other sources of bias. Of note, the studies did not differ for quality components 4 through 6. Trials with high or unclear risk for bias for any one of the first three components were considered as trials with high-risk of bias. Otherwise, they were considered as low-risk of bias trials.
Data extraction and synthesis
For the purpose of this analysis, the intensive BP group was defined as the group where the final achieved systolic pressure was ≤135 mm Hg and the standard BP group as where the final achieved systolic pressure was ≤140 mm Hg. Of note, these terms are based on the trial mean achieved systolic pressure, are used for descriptive purposes for this manuscript and not necessarily the strategy employed in the trial (ie, no trial tested a BP strategy). The intensive BP group was further divided into a more intensive group with an achieved BP of ≤130 mm Hg and a less intensive group with an achieved BP of >130–≤135 mm Hg (figure 1).
Long-term efficacy and safety outcomes were evaluated. The efficacy outcomes were: all-cause mortality, cardiovascular mortality, myocardial infarction, stroke, angina pectoris, heart failure and revascularisation. The safety outcome evaluated was hypotension as reported between the two groups.
Intention-to-treat meta-analysis was performed in line with recommendations from the Cochrane Collaboration and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement,10 ,11 using standard software (Stata V.9.0, Stata corporation).12 Heterogeneity was assessed using the I2 statistic.13 I2 is the proportion of total variation observed between the trials attributable to differences between trials rather than sampling error (chance) with I2 <25% considered as low and I2 >75% as high. The analysis employed rates per 1000 patient-years of follow-up rather than events. This is more appropriate because they incorporate and control for the varying duration of the trials. Patient-years of follow-up were calculated by multiplying the sample size for each trial with the mean follow-up duration. The results are expressed as rate ratios rather than relative risk. If trials were homogeneous (p>0.05), a fixed effect model was used to calculate pooled effect sizes. Otherwise, a random effects model of DerSimonian and Laird14 was applied to calculate overall differences. However, given the clinical heterogeneity between the trials, regardless of statistical heterogeneity, a random effects model was used to make inferences (unless otherwise stated). Publication bias was estimated visually by funnel plots and using the Begg's test and the weighted regression test of Egger et al.15 Analyses were performed after further stratifying the studies based on the final achieved systolic pressure in the intensive group: systolic pressure >130 but ≤135 mm Hg (less intensive group) versus systolic pressure ≤130 mm Hg (more intensive group). We estimated the difference between the estimates of the subgroups according to tests of interaction.16 A p value <0.05 indicates that the effects of treatment differ between the tested subgroups.
A meta-regression analysis was performed to explore the relationship between systolic pressure (final achieved) and outcomes. For this purpose, the mean achieved systolic pressure was used as a continuous variable. We used residual maximum likelihood to estimate the additive (between-study) component of variance τ2 for the meta-regression analysis. Bootstrap analyses were performed using a Monte Carlo permutation test for meta-regression using 1000 random permutations.17
In a meta-analysis of clinical trials with binary outcomes such as the one described above, a normal approximation for the summary treatment effect measure in each trial may not be appropriate when some of the trials in the meta-analysis are small or the observed risks are close to 0 or 1. In order to avoid this problem, direct use of the binomial distribution within trials can be used as described by Warn et al.18 The advantages of Bayesian methods include a modelling framework which overcomes issues such as the appropriate treatment of small trials, and the ability to consider distributions other than normal for the random effects. In order to confirm the results from the traditional meta-analysis, Bayesian random effects meta-analyses were performed. The BUGS code for implementing the model is as described by Warn et al.18 Minimally informative prior distributions were used, so the findings and interpretation are close to those obtained with frequentist methods. All Bayesian analyses were conducted using WinBUGS 1.4.3 London, United Kingdom.
Further sensitivity analysis was performed by restricting the analysis to low bias risk trials. In addition, further analyses were performed after including trials which narrowly missed the inclusion criteria (where intensive BP group >135 mm Hg), such as the Heart Outcomes Prevention Evaluation (HOPE) trial (achieved systolic BP of 139 and 136 mm Hg),19 Telmisartan Randomized AssessmeNt Study in aCE iNtolerant subjects with cardiovascular Disease (TRANSCEND) trial (achieved systolic BP of 136.4 and 140.2 mm Hg)20 and the Japan Multicenter Investigation for Cardiovascular Diseases-B (JMIC-B) randomised trial (CAD subgroup) (achieved systolic BP of 136 and 138 mm Hg).21 Moreover, further analyses were conducted restricting trials where the BP difference between arms was at least 3 mm Hg, such as that used in our prior analysis.8
Role of the funding source
This work was not funded and hence there was no role of any funding source in the conception, data synthesis, analysis, data interpretation or in the drafting of the manuscript.
We identified 15 RCTs that fulfilled the inclusion criteria and were chosen for this analysis (figure 1). The Randomized Olmesartan and Diabetes Microalbuminuria Prevention (ROADMAP) trial included was the CAD subgroup only. In addition, three other trials (HOPE, TRANSCEND, JMIC-B) that narrowly missed the inclusion criteria were included in a sensitivity analysis as described above.
Characteristics of the trials
The baseline characteristics, inclusion criteria and bias–risk assessment are summarised in tables 1 and 2. The 15 RCTs enrolled 66 504 participants with 276 328 patient-years of follow-up: 37 842 (50.5%) participants to the group with achieved systolic BP ≤135 mm Hg (intensive BP group) and 28 662 (49.5%) participants to the group with achieved systolic BP ≤140 mm Hg (standard BP group) were followed-up for 3.4±1.2 years (weighted mean). Of note, none of the trials were designed to test a BP strategy.
Among the 15 RCTs considered for this analysis, 11 were considered trials with a low risk of bias as described above and the others were considered trials with unclear or high risk of bias (table 2).
Intensive BP group (≤135 mm Hg) was not associated with any significant benefit for the outcomes of death (figure 2A), CV death (figure 2B) or revascularisation (figure 2E) when compared with standard BP (≤140 mm Hg). The results were similar for ‘more’ and ‘less’ intensive BP subgroups for the above outcomes (pinteraction >0.05). For the outcome of myocardial infarction (figure 2C), the test for interaction was significant (pinteraction=0.03) such that the more intensive BP subgroup was associated with greater reduction in myocardial infarction (figure 2C) compared with the less intensive BP subgroup. In addition, the more intensive BP subgroup was associated with an 8% reduction in myocardial infarction and angina pectoris (figure 2D) when compared with standard BP group in the fixed effect model but not the random effects model (figure 2C,D). There was low heterogeneity for the outcome of death and modest heterogeneity for the outcomes of angina pectoris, CV death and myocardial infarction (figure 2A–D) but high heterogeneity for the outcome of revascularisation (I2=77.2%).
Intensive BP (≤135 mm Hg) was associated with a 15% decrease in heart failure (figure 3A) and 10% decrease in stroke (figure 3B). More intensive BP control (≤130 mm Hg) was associated with a greater (27% and 17%) reduction in heart failure and stroke when compared with standard BP (figure 3A,B). There was no heterogeneity for the outcome of stroke but modest heterogeneity for the outcome of heart failure (figure 3A,B).
Bias was insignificant for any of the above analyses (online supplementary web appendix figure A1–7). The results were similar when the analysis was restricted to low bias risk trials (data not shown).
Intensive BP group (≤135 mm Hg) was associated with a 105% increase in hypotension rate (figure 3C) when compared with the standard BP group. The results were similar for more versus less intensive subgroups (pinteraction=0.97). There was high heterogeneity for this analysis but bias was insignificant (online supplementary web appendix figure A8). Analysis was performed to explore heterogeneity for this outcome. A Galbraith's plot revealed that the ONTARGET trial may be a possible outlier. Excluding this trial reduced the heterogeneity somewhat (I2=67.7%) but the results were similar with a 103% increase in hypotension rate. No other tested variable reduced the heterogeneity further and the residual heterogeneity could likely be the result of clinical heterogeneity between trials.
The relationship between final achieved systolic pressure and the risk of efficacy and safety outcomes is shown in figure 4A–F. For the outcomes of death and CV death, lower was not better for systolic BP (figure 4A,B) and was uniformly no different than the standard BP group (log rate ratio ∼0). For the outcomes of myocardial infarction, stroke, angina and heart failure, lower systolic BP was associated with a greater rate ratio reduction (figure 4C–F). The relationship between lower systolic BP and heart failure outcomes was significant even after bootstrap analyses performed using a Monte Carlo permutation test for meta-regression with 1000 random permutations (p=0.025), such that for each 10 mm Hg lower systolic pressure, there was a 50% decrease in the log risk ratio for heart failure. Similarly for the outcome of myocardial infarction, there was a trend towards (p=0.049) lower being better with each 10 mm Hg lower systolic pressure associated with a 24% lower log risk ratio. For the safety outcome of hypotension, the rate ratio was uniformly high with intensive BP across systolic pressures (figure 5).
Sensitivity analysis performed using a Bayesian random effects model showed similar results with an 18% reduction in the rate of heart failure and a 14% reduction in the rate of stroke with intensive BP control (≤135 mm Hg) when compared with standard BP control but with a 99% increase in the rate of hypotension with intensive BP control (table 3). In addition, a Bayesian random effects model evaluating the relationship between systolic BP and outcomes showed that lower was better for the outcomes of myocardial infarction and stroke (table 3). The results were similar when the analysis was restricted to low bias risk trials or using trials where the systolic pressure difference was at least 3 mm Hg (data not shown).
Further sensitivity analyses performed after inclusion of the HOPE, TRANSCEND and JMIC-B trials showed similar results with a significant benefit of intensive BP control (≤135 mm Hg) for the outcomes of heart failure and stroke at the expense of an increase in hypotension (table 4).
The principal finding of the present study is that compared with a BP target of ≤140 mm Hg, a more intensive BP target of ≤135 mm Hg is associated with significant reduction in stroke and heart failure but at the expense of increased rate of hypotension, consistently seen in both the traditional meta-analysis and using a Bayesian random effects model. In addition, lower seemed to be better (based on regression analysis) for the outcomes of myocardial infarction, stroke, heart failure and perhaps angina.
BP targets in patients without CAD
Data from observational studies involving more than 1 million individuals without pre-existing vascular disease indicate that death from both ischaemic heart disease and stroke increases progressively and linearly with BP.37 Consequently, the notion that ‘lower is better’ has been popular for management of hypertension. The seventh report of the Joint National Committee on prevention, detection, evaluation and treatment of high BP states ‘The relationship between BP and risk of cardiovascular events is continuous, consistent, and independent of other risk factors’.1 As a consequence, a BP of <120/80 mm Hg has been considered as ‘optimal’ or ‘normal’.1 However, this linear theory has been challenged for nearly 3 decades5–7 ,38 ,39 and the recently published ACCORD BP trial showed no benefit of lowering systolic pressure to <120 mm Hg, except for stroke.4 In addition, a recently published analysis from our group further confirms the findings from the ACCORD BP trial in subjects with diabetes.8
BP targets in patients with CAD
In the American Heart Association scientific statement on ‘Treatment of Hypertension in the Prevention and Management of Ischemic Heart Disease,’ a target of <130/80 mm Hg has been recommended in patients with CAD and acute coronary syndromes, although it was acknowledged that there were limited data to support this recommendation (Class IIa, level of evidence B).40 Similarly, other national and international guidelines recommend a lower target of ≤130/80 mm Hg in patients with established cardiovascular disease. However, the evidence to support this lower goal is scant.
Antihypertensive therapy for secondary prevention in patients with CAD is distinctly different from primary prevention of CAD (such as tested in ACCORD BP). If aggressive strategies such as intensive BP control were to be efficacious, they can be expected to be so in the highest risk subsets, such as those with known CAD. Increased BP is a significant risk factor for the development of heart failure, stroke and less so for myocardial infarction. On the contrary, lower BP has been shown to often compromise coronary perfusion, leading to increased risk of cardiovascular events.
We have shown in an analysis of 10 001 patients with CAD enrolled in the Treating to New Targets trial that the event rate (a composite of death from coronary disease, non-fatal myocardial infarction, resuscitated cardiac arrest, and fatal or non-fatal stroke) at the end of 4.9 years (median) of follow-up was the lowest at a BP of 146.3/81.4 mm Hg and that a very low BP (<110–120/<60–70 mm Hg) portends an increased risk of events.5 Similarly, a J-shaped relationship between BP and cardiovascular outcomes has been shown in subgroup analyses from other randomised trials (INVEST,7 ONTARGET,41 CAD cohorts of Cruickshank et al,38 Framingham Heart Study39 and Syst-Eur42). In addition, we found similar results in 4162 acute coronary syndrome patients enrolled in the PROVE-IT TIMI 22 trial (randomised to pravastatin 40 mg vs atorvastatin 80 mg).6 The nadir BP where the risk of primary outcome (death from any cause, myocardial infarction (MI), unstable angina requiring rehospitalisation, revascularisation after 30 days and stroke) was the lowest for a BP of 136/85 mm Hg, while in INVEST the nadir systolic BP was ∼119 mm Hg. In addition, in all of these analyses, the increased risk of cardiovascular outcomes with systolic pressures occurred at very low systolic pressures (<110 mm Hg) with a relatively shallow curve between 110 and 140 mm Hg.
The result of the present analysis suggests that intensive BP to ≤135 mm Hg was consistently seen to have a modest effect at lowering stroke and heart failure, both in the traditional meta-analysis and the Bayesian random effect meta-analysis, but at the expense of hypotension. In addition, BP ≤130 mm Hg was associated with a modest benefit for myocardial infarction and angina when compared with standard BP group. Moreover, meta-regression analysis suggests sustained benefit of aggressive BP control for the outcomes of heart failure and myocardial infarction, with the slope suggesting benefit for outcomes of stroke and angina.
In the INVEST trial analysis described above, the event rate continued to decrease with a nadir systolic pressure of around 120 mm Hg, similar to the finding from this study. Our results are discordant with the main results from the ACCORD BP trial in subjects with diabetes, where lower was better only for the outcome of stroke. In the ACCORD BP trial, in a subgroup analysis in patients with known prior cardiovascular disease, an intensive BP strategy (<120 mm Hg) was associated with a numerically lower primary endpoint rate (2.98%/year) when compared with the standard BP strategy (<140 mm Hg) (3.43%/year), although this was not statistically significant. The analysis was underpowered with only 1593 patients with known cardiovascular disease.
Similarly, our own observation from non-randomised comparisons seems to suggest that lower is better but this only goes so far and systolic BP below 110 mm Hg is not advisable. We however did not see a J-curve association in the present analyses as there was no trial where the systolic BP was very low (below 110 mm Hg). More importantly, none of the trials included were designed to test a BP strategy. Though the findings are hypothesis generating and provide some evidence in the interim, this observation should be further investigated in future clinical trials. Of note, in the ACCORD BP trial only a third of patients had cardiovascular disease (including those with stroke, peripheral arterial disease and not necessarily all CAD) with even lower percentage of patients with CAD and hence the results of ACCORD BP cannot and should not be extrapolated to the CAD cohort.
Randomised controlled trials testing BP strategies (targeting <120 mm Hg vs <140 mm Hg) and powered for hard clinical outcomes (death or myocardial infarction) using therapies that have proven efficacy (ARBs, ACEi, CCB, β blockers and diuretics) are needed to test such an association. The design of such a trial should be an open label design with a treatment escalation strategy to achieve BP goals with a PROBE design with blinded outcome assessors.43
As in other meta-analyses, given the lack of data in each trial, we did not adjust our analyses for medications used. Though detailed analyses were undertaken, given the heterogeneity in the study designs and cohort enrolled, clinically relevant differences could have been missed and are best assessed in a meta-analysis of individual patient data. All of the trials did not report all of the outcomes and none were designed to test a BP strategy. The results are therefore best described as hypothesis generating to be further confirmed in future RCTs. In this analysis, we tested a BP goal of ≤140 mm Hg with that of ≤135 mm Hg and not of a lower goal such as ≤130 mm Hg for several reasons: (1) to ensure that we use data from as many relevant trials as possible as the number of trials if the criteria is tightened to ≤130 mm Hg was very small; (2) test if evidence exists for even a 135 mm Hg goal and finally and more importantly; (3) the cut points are less relevant as in our regression analysis, the mean achieved BP was treated as a continuous variable and our conclusions and recommendations are based on this. Moreover, the intensive group was substratified into a more intensive group (≤130 mm Hg) versus a less intensive group (>130 but ≤135 mm Hg). The relationship tested is for systolic BP targets only (one that is recommended by guidelines and used in clinical practice). Moreover, we did not test for the treatment effect of individual trials as the intention was to test the effect of a given BP target rather than the medication used to achieve such a target.
The present body of evidence suggests that intensive BP control to ≤135 mm Hg and possibly even to ≤130 mm Hg reduces heart failure and stroke in patients with CAD at the expense of increase in hypotension, with meta-regression analysis suggesting lower the better for myocardial infarction, stroke, heart failure and perhaps angina. Randomised controlled trials testing BP strategies are needed to conclusively prove the efficacy and safety of aggressive BP control in subjects with CAD.
Competing interests Franz H Messerli: Ad hoc consulting: Abbott, Novartis, Pfizer, Bayer, Forest, Takeda, Daiichi. Research/Grants: Novartis, Boehringer.
Provenance and peer review Not commissioned; externally peer reviewed.