Professional Documents
Culture Documents
Editorial
IMPORTANCE Clinical trials have documented that lowering blood pressure reduces Supplemental content
cardiovascular disease and premature deaths. However, the optimal target for reduction
of systolic blood pressure (SBP) is uncertain.
OBJECTIVE To assess the association of mean achieved SBP levels with the risk of
cardiovascular disease and all-cause mortality in adults with hypertension treated with
antihypertensive therapy.
DATA SOURCES MEDLINE and EMBASE were searched from inception to December 15, 2015,
supplemented by manual searches of the bibliographies of retrieved articles.
STUDY SELECTION Studies included were clinical trials with random allocation to an
antihypertensive medication, control, or treatment target. Studies had to have reported a
difference in mean achieved SBP of 5 mm Hg or more between comparison groups.
DATA EXTRACTION AND SYNTHESIS Data were extracted from each study independently and
in duplicate by at least 2 investigators according to a standardized protocol. Network
meta-analysis was used to obtain pooled randomized results comparing the association of
each 5mm Hg SBP category with clinical outcomes after adjusting for baseline risk.
RESULTS Forty-two trials, including 144 220 patients, met the eligibility criteria. In general,
there were linear associations between mean achieved SBP and risk of cardiovascular disease
and mortality, with the lowest risk at 120 to 124 mm Hg. Randomized groups with a mean
achieved SBP of 120 to 124 mm Hg had a hazard ratio (HR) for major cardiovascular disease
of 0.71 (95% CI, 0.60-0.83) compared with randomized groups with a mean achieved SBP
of 130 to 134 mm Hg, an HR of 0.58 (95% CI, 0.48-0.72) compared with those with a mean
achieved SBP of 140 to 144 mm Hg, an HR of 0.46 (95% CI, 0.34-0.63) compared with those
with a mean achieved SBP of 150 to 154 mm Hg, and an HR of 0.36 (95% CI, 0.26-0.51)
compared with those with a mean achieved SBP of 160 mm Hg or more. Likewise, Author Affiliations: Department of
randomized groups with a mean achieved SBP of 120 to 124 mm Hg had an HR for all-cause Epidemiology, Tulane University
mortality of 0.73 (95% CI, 0.58-0.93) compared with randomized groups with a mean School of Public Health and Tropical
Medicine, New Orleans, Louisiana
achieved SBP of 130 to 134 mm Hg, an HR of 0.59 (95% CI, 0.45-0.77) compared with those (Bundy, Li, Stuchlik, Bu, Kelly, Mills,
with a mean achieved SBP of 140 to 144 mm Hg, an HR of 0.51 (95% CI, 0.36-0.71) compared H. He, Chen, Whelton, J. He);
with those with a mean achieved SBP of 150 to 154 mm Hg, and an HR of 0.47 (95% CI, Department of Epidemiology, School
of Public Health, Medical College of
0.32-0.67) compared with those with a mean achieved SBP of 160 mm Hg or more.
Soochow University, Suzhou, China
(Bu); Department of Medicine,
CONCLUSIONS AND RELEVANCE This study suggests that reducing SBP to levels below Tulane University School of Medicine,
currently recommended targets significantly reduces the risk of cardiovascular disease and New Orleans, Louisiana (Chen,
Whelton, J. He).
all-cause mortality. These findings support more intensive control of SBP among adults
Corresponding Author: Jiang He,
with hypertension.
MD, PhD, Department of
Epidemiology, Tulane University
School of Public Health and Tropical
Medicine, 1440 Canal St, Ste 2000,
JAMA Cardiol. doi:10.1001/jamacardio.2017.1421 New Orleans, LA 70118
Published online May 31, 2017. (jhe@tulane.edu).
(Reprinted) E1
2017 American Medical Association. All rights reserved.
H
ypertension is the leading global preventable risk fac-
tor for cardiovascular disease (CVD) and premature Key Points
death.1,2 Observational epidemiologic studies have
Question What is the optimal target for reduction of systolic
shown a strong, independent, and log-linear association be- blood pressure among patients with hypertension?
tween usual systolic blood pressure (SBP) and mortality from
Findings In this systematic review and network meta-analysis of
CVD and all causes, with no evidence of a threshold down to
42 trials, including 144 220 patients, linear associations were seen
at least 115 mm Hg.3 Randomized clinical trials have docu-
between mean achieved systolic blood pressure and risk of
mented that lowering blood pressure (BP) with commonly used cardiovascular disease and mortality, with the lowest risk at a
regimens reduces the risk of CVD and all-cause mortality.4,5 systolic blood pressure of 120 to 124 mm Hg.
However, post hoc analyses based on achieved BP in some clini-
Meaning Reducing systolic blood pressure below currently
cal trials, in which the results were not analyzed according to
recommended targets with commonly used antihypertensive
the randomized treatment assignment, identified a J-shaped medications may significantly reduce the risk of cardiovascular
association between achieved BP and risk of CVD and all- disease and all-cause mortality.
cause mortality, especially between achieved BP and coro-
nary heart disease (CHD).6,7
The uncertainty of optimal goals for treatment for pa-
tients with hypertension has resulted in inconsistent recom- Study Selection
mendations for BP targets in clinical practice guidelines.4,8,9 Titles and abstracts of retrieved articles were independently
For example, compared with the 2003 Seventh Report of the screened by at least 2 of us (J.D.B., C.L., P.S., and X.B.). Ar-
Joint National Committee on Prevention, Detection, Evalua- ticles deemed potentially eligible by either reviewer were
tion, and Treatment of High Blood Pressure,4 the 2014 Evi- retrieved for full-text review. Disagreements on full-text re-
dence-Based Guideline for the Management of High Blood Pres- view were resolved by discussion and consensus.
sure in Adults8 raised the recommended SBP treatment goal Studies were included if they met the following criteria:
from less than 130 mm Hg to less than 140 mm Hg for patients (1) participants were randomly allocated to an antihyperten-
with type 2 diabetes or chronic kidney disease and from less sive medication, control, or treatment target; (2) the alloca-
than 140 mm Hg to less than 150 mm Hg for individuals 60 tion to antihypertensive treatment was independent of other
years of age or older. Recently, the Systolic Blood Pressure In- treatment regimens; (3) the sample size was 100 patients or
tervention Trial (SPRINT) reported that intensive treatment more in each treatment group; (4) trial duration was 6 months
(targeting an SBP of <120 mm Hg), as compared with stan- or more; (5) one or more events for an outcome of interest were
dard treatment (targeting an SBP of <140 mm Hg), signifi- reported in each treatment group; (6) mean achieved SBP level
cantly reduced CVD and all-cause mortality among adults was reported for each treatment group, and the difference in
with hypertension who were at high risk for CVD, but with- mean achieved SBP between the comparison groups was 5 mm
out diabetes or stroke.10 The data from SPRINT support a Hg or more; and (7) outcomes included major CVD, stroke, CHD,
more intensive SBP treatment goal, although concerns re- CVD mortality, or all-cause mortality. Clinical trials with mean
main regarding its generalizability to populations at large with achieved SBP of 160 mm Hg or more in both comparison groups
hypertension. were excluded because they do not contribute information to
Finding the optimal SBP target could have far-reaching im- the optimal target for SBP treatment. For studies with mul-
plications for the reduction of CVD and premature death in gen- tiple publications, data from the article with the longest trial
eral populations. By using a network meta-analysis to com- follow-up time were included.
bine available data from randomized clinical trials, we
compared the association of different levels of SBP reduction Data Extraction and Quality Assessment
with the risk of major CVD, stroke, CHD, CVD mortality, and Data abstraction was conducted by 2 of us (J.D.B., C.L., P.S.,
all-cause mortality. and X.B.) who independently used a predefined, standard-
ized protocol and data collection instrument. Information was
recorded on sample size, demographic characteristics, and
medical history of the trial participants; BP measurement meth-
Methods ods; mean achieved BP during treatment; follow-up time; out-
Data Sources and Searches come ascertainment methods; and number of events for each
We searched MEDLINE and EMBASE using the following search outcome. The predefined outcomes were major CVD events
terms as medical subject headings and key words: (antihyper- (including CHD, stroke, heart failure, and CVD deaths), stroke,
tensive agents OR blood pressure lowering OR antihypertensive CHD, CVD mortality, and all-cause mortality.
treatment) AND (cardiovascular disease OR coronary disease OR Risk of bias was assessed by 2 of us (J.D.B., C.L., P.S., and
myocardial infarction OR stroke OR heart failure OR mortal- X.B.) using the Cochrane Collaborations risk of bias tool, based
ity). The searches were conducted without language or date on 7 domains11: random sequence generation, allocation con-
restriction, from inception to December 15, 2015. We limited cealment, blinding of participants and outcome assessment,
searches to randomized clinical trials in human adults. Addi- incomplete data, selective reporting, intention-to-treat analy-
tional trials were identified by hand-searching bibliographies sis, and other sources of bias. Disagreement was resolved by
from included studies, reviews, and meta-analyses. consensus.
Figure 2. Network of Treatment Comparisons for Cardiovascular Disease Figure 3. Hazard Ratios and 95% CIs for Major Cardiovascular Disease
and Mortality According to Achieved Systolic Blood Pressure Categories Associated With More Intensive Reductions in Systolic Blood Pressure
Among 42 Clinical Trials
Favors Favors
125129 Lower Higher
120124 Mean Achieved Systolic Hazard Ratio Blood Blood
mm Hg
mm Hg Blood Pressure, mm Hg (95% CI) Pressure Pressure
1
Reduction to 120-124
2 1 1
120-124 vs 125-129 0.82 (0.67-0.97)
135139
mm Hg 1 <120 mm Hg 120-124 vs 130-134 0.71 (0.60-0.83)
6
4 120-124 vs 135-139 0.68 (0.55-0.85)
2 1
3 120-124 vs 140-144 0.58 (0.48-0.72)
120-124 vs 145-149 0.55 (0.42-0.72)
3
140144 5 130134 120-124 vs 150-154 0.46 (0.34-0.63)
mm Hg mm Hg 120-124 vs 155-159 0.41 (0.32-0.54)
2 120-124 vs 160 0.36 (0.26-0.51)
1 1 Reduction to 130-134
3 130-134 vs 135-139 0.96 (0.83-1.14)
2 3 160 mm Hg
145149 130-134 vs 140-144 0.83 (0.74-0.94)
mm Hg 130-134 vs 145-149 0.78 (0.63-0.98)
130-134 vs 150-154 0.65 (0.51-0.85)
155159
150154 mm Hg 130-134 vs 155-159 0.58 (0.48-0.72)
mm Hg
130-134 vs 160 0.51 (0.39-0.69)
Reduction to 140-144
Each node (blue circle) represents an achieved systolic blood pressure category.
140-144 vs 145-149 0.94 (0.74-1.20)
The size of the nodes corresponds to the number of trials of the categories.
Comparisons are linked with a line, the thickness of which corresponds to the 140-144 vs 150-154 0.79 (0.63-0.99)
number of trials that assessed the comparison. Numbers next to every line 140-144 vs 155-159 0.70 (0.60-0.84)
indicate the number of trials directly comparing the categories. 140-144 vs 160 0.62 (0.48-0.80)
Reduction to 150-154
150-154 vs 155-159 0.90 (0.68-1.19)
all-cause mortality, and CVD mortality (Figure 3 and Figure 4; 150-154 vs 160 0.79 (0.66-0.94)
eFigure 2 and eTables 5-7 in the Supplement). The lowest risks 0.1 1.0 2
for major CVD, CHD, all-cause mortality, and CVD mortality Hazard Ratio (95% CI)
were at a mean achieved SBP of 120 to 124 mm Hg, whereas
the lowest risk for stroke was at a mean achieved SBP of less Square markers indicate hazard ratios for major cardiovascular disease events
comparing lower mean achieved systolic blood pressure with higher mean
than 120 mm Hg.
achieved systolic blood pressure. Error bars indicate 95% CIs.
Randomized groups with a mean achieved SBP of 120 to
124 mm Hg had an HR for major CVD of 0.71 (95% CI, 0.60-
0.83) compared with randomized groups with a mean achieved to 144 mm Hg, an HR of 0.51 (95% CI, 0.36-0.71) compared with
SBP of 130 to 134 mm Hg, an HR of 0.58 (95% CI, 0.48-0.72) those with a mean achieved SBP of 150 to 154 mm Hg, and an
compared with those with a mean achieved SBP of 140 to 144 HR of 0.47 (95% CI, 0.32-0.67) compared with those with a
mm Hg, an HR of 0.46 (95% CI, 0.34-0.63) compared with those mean achieved SBP of 160 mm Hg or more (Figure 4; eTable 7
with a mean achieved SBP of 150 to 154 mm Hg, and an HR of in the Supplement). Randomized groups with a mean achieved
0.36 (95% CI, 0.26-0.51) compared with those with a mean SBP of 120 to 124 mm Hg had an HR for CVD mortality of 0.67
achieved SBP of 160 mm Hg or more (Figure 3; eTable 5 in the (95% CI, 0.40-1.22) compared with randomized groups with
Supplement). Randomized groups with a mean achieved SBP a mean achieved SBP of 130 to 134 mm Hg, an HR of 0.55 (95%
of 120 to 124 mm Hg had an HR for stroke of 0.69 (95% CI, 0.40- CI, 0.30-1.07) compared with those with a mean achieved SBP
1.07) compared with randomized groups with a mean achieved of 140 to 144 mm Hg, an HR of 0.43 (0.22-0.93) compared with
SBP of 130 to 134 mm Hg, an HR of 0.51 (95% CI, 0.26-0.87) those with a mean achieved SBP of 150 to 154 mm Hg, and an
compared with those with a mean achieved SBP of 140 to 144 HR of 0.34 (0.17-0.76) compared with those with a mean
mm Hg, an HR of 0.36 (95% CI, 0.17-0.68) compared with those achieved SBP of 160 mm Hg or more (eFigure 2 and eTable 7
with a mean achieved SBP of 150 to 154 mm Hg, and an HR of in the Supplement).
0.27 (95% CI, 0.12-0.51) compared with those with a mean In a sensitivity analysis excluding SPRINT, HRs and 95%
achieved SBP of 160 mm Hg or more (eFigure 2 and eTable 5 CIs were consistent with results from the main analyses for ma-
in the Supplement). A similar but weaker association be- jor CVD, CHD, and all-cause mortality, indicating the lowest
tween mean achieved SBP and CHD was observed (eFigure 2 risk at an SBP of 120 to 124 mm Hg for these outcomes (eTables
and eTable 6 in the Supplement). 8-10 in the Supplement). However, in the sensitivity analysis,
Randomized groups with a mean achieved SBP of 120 to the lowest-risk group for stroke was the group with an SBP of
124 mm Hg had an HR for all-cause mortality of 0.73 (95% CI, 120 to 124 mm Hg, and the lowest-risk group for CVD mortal-
0.58-0.93) compared with randomized groups with a mean ity was the group with an SBP of less than 120 mm Hg. In the
achieved SBP of 130 to 134 mm Hg, an HR of 0.59 (95% CI, 0.45- main analyses, the lowest-risk group for stroke was the group
0.77) compared with those with a mean achieved SBP of 140 with an SBP of less than 120 mm Hg, and the lowest-risk group
Figure 4. Hazard Ratios and 95% CIs for All-Cause Mortality Associated
With More Intensive Reductions in Systolic Blood Pressure Discussion
Favors Favors This network meta-analysis of randomized clinical trials
Lower Higher
Mean Achieved Systolic Hazard Ratio Blood Blood documented significant and linear associations between
Blood Pressure, mm Hg (95% CI) Pressure Pressure mean achieved SBP and the risk of CVD and all-cause mor-
Reduction to 120-124 tality. The lowest risks for CVD and all-cause mortality were
120-124 vs 125-129 0.74 (0.57-0.97)
among randomized groups with a mean achieved SBP of 120
120-124 vs 130-134 0.73 (0.58-0.93)
to 124 mm Hg. These findings support recently published
120-124 vs 135-139 0.79 (0.59-1.05)
120-124 vs 140-144 0.59 (0.45-0.77) results from SPRINT10 and suggest a benefit of reducing SBP
120-124 vs 145-149 0.71 (0.50-1.00) below the currently recommended target among adults with
120-124 vs 150-154 0.51 (0.36-0.71) hypertension.8
120-124 vs 155-159 0.49 (0.34-0.67) The SPRINT trial randomly assigned 9361 persons 50 years
120-124 vs 160 0.47 (0.32-0.67) of age or older with an SBP of 130 to 180 mm Hg who had an
Reduction to 130-134
increased risk of CVD, but without diabetes or stroke, to re-
130-134 vs 135-139 1.08 (0.90-1.29)
ceive intensive treatment or standard treatment of SBP.10 Blood
130-134 vs 140-144 0.82 (0.68-0.93)
130-134 vs 145-149 0.97 (0.75-1.26) pressure was measured in accordance with a prespecified, stan-
130-134 vs 150-154 0.71 (0.53-0.90) dardized protocol. The mean achieved SBP was 121.5 mm Hg
130-134 vs 155-159 0.68 (0.51-0.85) in the intensive-treatment group and 134.6 mm Hg in the stan-
130-134 vs 160 0.68 (0.47-0.85) dard-treatment group during the intervention. During a me-
Reduction to 140-144 dian follow-up of 3.26 years, a significant 25% reduction in the
140-144 vs 145-149 1.20 (0.93-1.59)
primary composite outcome of CVD events (HR, 0.75; 95% CI,
140-144 vs 150-154 0.87 (0.69-1.08)
0.64-0.89; P < .001) and a 27% reduction in all-cause mortal-
140-144 vs 155-159 0.83 (0.67-1.01)
140-144 vs 160 0.80 (0.62-1.03) ity (HR, 0.73; 95% CI, 0.60-0.90; P = .003) were reported.
Reduction to 150-154 In our network meta-analysis, compared with random-
150-154 vs 155-159 0.96 (0.71-1.29) ized groups with a mean achieved SBP of 130 to 134 mm Hg,
150-154 vs 160 0.92 (0.77-1.09) CVD was reduced by 29% (HR, 0.71; 95% CI, 0.60-0.83), and
0.1 1.0 2
all-cause mortality was reduced by 27% (HR, 0.73; 95% CI, 0.58-
Hazard Ratio (95% CI) 0.93), among randomized groups with a mean achieved SBP
of 120 to 124 mm Hg. This agreement persisted even after ex-
Square markers indicate hazard ratios for all-cause mortality comparing lower cluding SPRINT in a sensitivity analysis. The findings from
mean achieved systolic blood pressure with higher mean achieved systolic
SPRINT10 and our network meta-analysis suggest that a more
blood pressure. Error bars indicate 95% CIs.
intensive treatment target than currently recommended
(eg, SBP of 120-124 mm Hg) provides additional benefits for
for CVD mortality was the group with an SBP of 120 to 124 mm prevention of CVD complications and all-cause mortality.
Hg. A second sensitivity analysis excluding trials with 4 or more Our study contributes additional information on SBP man-
categories deemed at high or unclear risk of bias did not agement strategies beyond SPRINT.17 First, our study in-
substantively change the results compared with the main analy- cluded 42 clinical trials conducted for 144 220 patients with
ses (eTables 11-13 in the Supplement). various comorbidities (including diabetes and stroke), age
Model fit for all outcomes was adequate according to the ranges, and mean BP levels at baseline. Therefore, these re-
Bayesian deviance information criterion, and the baseline risk sults are generalizable to populations at large with hyperten-
covariate did not significantly alter the models (eTables 14-18 sion. Second, our study compared multiple levels of achieved
in the Supplement). Heterogeneity was present for each out- SBP on the risk of CVD and all-cause mortality and found posi-
come, with random-effects models fitting better than fixed- tive and linear associations between achieved SBP and clini-
effects models according to the Bayesian deviance informa- cal outcomes. Our findings do not support the existence of a
tion criterion. The magnitude of heterogeneity was low to J-shaped association between achieved SBP and the risk of CVD
moderate, with a between-trial SD of 0.081 for major CVD and and all-cause mortality. Furthermore, our study indicates that
an SD ranging from 0.103 to 0.248 for the other outcomes there is a linear association between the magnitudes of SBP re-
(eTables 14-18 in the Supplement). duction and the risk of CVD and all-cause mortality. For ex-
There was no network-wide evidence of inconsistency ample, by lowering SBP by 10 mm Hg to achieve the treat-
between direct and indirect comparisons in any of the out- ment goal of 120 to 124 mm Hg, the risk of CVD was reduced
comes based on inconsistency models and design-by- by 29% (95% CI, 17%-40%), by lowering SBP by 20 mm Hg, the
treatment interaction analyses (eTables 19-23 in the Supple- risk of CVD was reduced by 42% (95% CI, 28%-52%), by low-
ment). However, inconsistency was present in a few individual ering SBP by 30 mm Hg, the risk of CVD was reduced by 54%
comparisons based on node-splitting analyses (major CVD, 125- (95% CI, 37%-66%), and by lowering SBP by 40 mm Hg or more,
129 vs 130-134 mm Hg; CHD, 125-129 vs 130-134 mm Hg; and the risk of CVD was reduced by 64% (95% CI, 49%-74%). These
CVD mortality, 120-124 vs 130-134 mm Hg, and 120-124 vs 135- data support a more intensive SBP management approach to
139 mm Hg). There was no evidence of publication bias. achieve a lower SBP goal.
Several meta-analyses have examined the association with of 120 to 124 mm Hg. Another strength of our network meta-
CVD and mortality of more intensive vs less intensive treat- analysis is that it uses all available information (direct and in-
ment of BP.18,19 Recently, Xie and colleagues18 reported an up- direct comparisons) to compare the association of each mean
dated meta-analysis of 19 clinical trials, including 44 989 par- achieved SBP level with clinical outcomes. Therefore, it was pos-
ticipants, on the association of intensive BP reduction with CVD sible to base the comparisons between various SBP levels on a
outcomes. The mean achieved SBP was 133 mm Hg (range, 118- much larger number of clinical trials compared with similar
144 mm Hg) in the more intensive treatment group and 140 mm meta-analyses limited to trials examining more intensive com-
Hg (range, 124-154 mm Hg) in the less intensive treatment group. pared with less intensive therapy; our study included 42 trials
Intensive BP-lowering treatment was associated with a reduc- compared with the traditional analysis from Xie et al18 that in-
tion of 14% (95% CI, 4%-22%) for major CVD, 13% (95% CI, 0%- cluded 19 trials. Our data indicate that there was no significant
24%) for myocardial infarction, and 22% (95% CI, 10%-32%) for difference between direct and indirect comparisons at the net-
stroke. However, more intensive treatment had no significant work level. In addition, we used a systematic and comprehen-
association with CVD mortality (9%; 95% CI, 11% to 26%) or all- sive search strategy to identify a wide coverage of available an-
cause mortality (9%; 95% CI, 3% to 19%). Another recent meta- tihypertensive clinical trials. Most of the included trials had low
analysis conducted by Ettehad and colleagues19 suggested that risk of bias; a sensitivity analysis indicated that trials with un-
every 10mm Hg reduction in SBP, including to levels less than clear risk of bias did not substantially influence our results. Fi-
130 mm Hg, significantly reduced the risk of major CVD and CHD. nally, our analyses included a large number of trials conducted
Our network meta-analysis results complement and expand on in diverse patient populations and were adjusted for differ-
the findings from these traditional meta-analyses. Our analy- ences in intervention duration and baseline risk among trials,
ses, based on many achieved SBP categories while maintain- which increases the generalizability of our findings.
ing randomized treatment assignments, show a beneficial lin- Our findings should be interpreted in light of several limi-
ear association between more intensively reduced mean tations, most of which have been common to all meta-
achieved SBPs and clinical outcomes, and identify the lowest analyses conducted in this topic area. First, we had limited
risk at a mean SBP of 120 to 124 mm Hg. sample size in some mean achieved SBP comparisons. For ex-
The association of intensive treatment in subgroups of pa- ample, only 3 trials achieved mean SBP levels below 120 mm
tients with certain comorbidities, especially type 2 diabetes, have Hg, with a combined sample size of 7333. Thus, most of the
been of particular interest.18,20-23 The Action to Control Cardio- evidence in our analyses is based on trials treating partici-
vascular Risk in Diabetes trial examined the association of an pants to achieve SBP levels above 120 mm Hg. Second, few trials
intensive SBP target (<120 mm Hg) compared with a standard reported heart failure outcomes, which resulted in an insuf-
SBP target (<140 mm Hg) for patients with diabetes, finding a ficiently connected network to analyze this outcome. Simi-
nonsignificant benefit on reducing risk for CVD events, which larly, we were unable to assess the association of intensive SBP
could be a consequence of reduced statistical power or use reduction with kidney disease outcomes, dementia, or ad-
of a factorial design.20,21 A 2012 meta-analysis conducted by verse events such as hypotension or falling, which have been
McBrien and colleagues22 reported a small reduction in the risk concerns with intensive treatment of BP.18 Furthermore, we
of stroke associated with more intensive BP reduction in adults were unable to conduct subgroup analyses by age, race/
with type 2 diabetes but found inconclusive results for mortal- ethnicity, history of CVD, stroke, chronic kidney disease, or dia-
ity and CHD. Another recent meta-analysis by Brunstrm and betes owing to insufficient data. Finally, we defined treat-
Carlberg23 reported increased risk of CVD mortality among pa- ment nodes according to the mean achieved SBP in each
tients with diabetes who had a baseline SBP of less than 140 mm randomization group, which does not consider the distribu-
Hg and reduced their level of SBP via treatment, suggesting a tion of individual SBP levels within groups. Thus, mean
J-shaped association. We were able to include many trials of pa- achieved SBP groups may represent a range of SBPs. In addi-
tients with diabetes and other comorbidities. Our findings do tion, analysis of mean achieved SBP does not guide treat-
not support the existence of a J-shaped association among popu- ment decisions regarding diastolic BP.
lations at large with hypertension. There are several implications for clinicians based on find-
ings from SPRINT,10 other meta-analyses, and our network
Strengths and Limitations meta-analysis. First, data suggest that treatment to achieve an
There are several strengths in this network meta-analysis com- SBP below currently recommended guidelines reduces the risk
pared with the previous meta-analyses that used traditional for major CVD and all-cause mortality in adults with hyper-
analysis methods. Network meta-analysis methods offer a tension. However, there may be a tradeoff between these ben-
unique advantage compared with traditional meta-regression efits and potential adverse effects of intensive SBP reduction,
techniques by allowing the simultaneous comparison of mul- including hypotension, electrolyte abnormalities, and kid-
tiple achieved SBP levels on clinical outcomes while preserv- ney injury.10,18 Thus, clinicians should continue to monitor
ing trial-level treatment randomization and its associated pro- acute adverse effects in individual patients and make treat-
tection against bias. Our study allowed for comparisons of a ment decisions based on accurate BP measurements, accord-
wider range of mean achieved SBP levels than has been pos- ing to standardized protocols similar to those used in clinical
sible in traditional meta-analyses, with a spread from less than trials. Second, although our analysis suggests that intensive
120 mm Hg to more than 160 mm Hg, and identified the lowest SBP reduction reduces risk for major CVD and all-cause mor-
risks for CVD and all-cause mortality at a mean achieved SBP tality in populations at large with hypertension, including in
ARTICLE INFORMATION age-sex specific all-cause and cause-specific 12. Lu G, Ades AE. Combination of direct and
Accepted for Publication: January 18, 2017. mortality for 240 causes of death, 1990-2013: indirect evidence in mixed treatment comparisons.
a systematic analysis for the Global Burden of Stat Med. 2004;23(20):3105-3124.
Published Online: May 31, 2017. Disease Study 2013. Lancet. 2015;385(9963):117-171.
doi:10.1001/jamacardio.2017.1421 13. Multi-parameter Evidence Synthesis Research
3. Lewington S, Clarke R, Qizilbash N, Peto R, Group. Mixed treatment comparisons: network
Author Contributions: Dr J. He had full access to all Collins R; Prospective Studies Collaboration. meta-analysis. http://www.bristol.ac.uk/social
the data in the study and takes responsibility for the Age-specific relevance of usual blood pressure to -community-medicine/projects/mpes/mtc/.
integrity of the data and the accuracy of the data vascular mortality: a meta-analysis of individual Accessed September 25, 2015.
analysis. data for one million adults in 61 prospective studies.
Study concept and design: Bundy, Kelly, Mills, J. He. 14. Achana FA, Cooper NJ, Dias S, et al. Extending
Lancet. 2002;360(9349):1903-1913. methods for investigating the relationship between
Acquisition, analysis, or interpretation of data:
Bundy, Li, Stuchlik, Bu, Mills, H. He, Chen, Whelton, 4. Chobanian AV, Bakris GL, Black HR, et al; Joint treatment effect and baseline risk from pairwise
J. He. National Committee on Prevention, Detection, meta-analysis to network meta-analysis. Stat Med.
Drafting of the manuscript: Bundy, Li, Bu, J. He. Evaluation, and Treatment of High Blood Pressure. 2013;32(5):752-771.
Critical revision of the manuscript for important National Heart, Lung, and Blood Institute; 15. Jackson D, Barrett JK, Rice S, White IR, Higgins
intellectual content: Bundy, Stuchlik, Kelly, Mills, National High Blood Pressure Education Program JPT. A design-by-treatment interaction model for
H. He, Chen, Whelton, J. He. Coordinating Committee. Seventh report of the network meta-analysis with random inconsistency
Statistical analysis: Bundy, Li, Stuchlik, Bu, Mills, Joint National Committee on Prevention, effects. Stat Med. 2014;33(21):3639-3654.
H. He, Whelton, J. He. Detection, Evaluation, and Treatment of High Blood
Pressure. Hypertension. 2003;42(6):1206-1252. 16. Dias S, Welton NJ, Caldwell DM, Ades AE.
Obtained funding: J. He. Checking consistency in mixed treatment
Administrative, technical, or material support: Mills. 5. Turnbull F; Blood Pressure Lowering Treatment comparison meta-analysis. Stat Med. 2010;29(7-8):
Study supervision: Chen, J. He. Trialists Collaboration. Effects of different 932-944.
Conflict of Interest Disclosures: All authors have blood-pressure-lowering regimens on major
cardiovascular events: results of prospectively- 17. Jones DW, Weatherly L, Hall JE. SPRINT: what
completed and submitted the ICMJE Form for remains unanswered and where do we go from
Disclosure of Potential Conflicts of Interest and designed overviews of randomised trials. Lancet.
2003;362(9395):1527-1535. here? Hypertension. 2016;67(2):261-262.
none were reported.
6. Messerli FH, Panjrath GS. The J-curve between 18. Xie X, Atkins E, Lv J, et al. Effects of intensive
Funding/Support: The research reported in this blood pressure lowering on cardiovascular and
publication was supported by award U01HL114197 blood pressure and coronary artery disease or
essential hypertension: exactly how essential? J Am renal outcomes: updated systematic review and
from the National Heart, Lung, and Blood Institute meta-analysis. Lancet. 2016;387(10017):435-443.
and by award P20GM109036 from the National Coll Cardiol. 2009;54(20):1827-1834.
Institute of General Medical Sciences. Dr Bu is 7. Mancia G, Grassi G. Aggressive blood pressure 19. Ettehad D, Emdin CA, Kiran A, et al. Blood
supported by research training grant D43TW009107 lowering is dangerous: the J-curve: pro side of the pressure lowering for prevention of cardiovascular
from the John E. Fogarty International Center of the argument. Hypertension. 2014;63(1):29-36. disease and death: a systematic review and
National Institutes of Health. Dr Mills is supported in meta-analysis. Lancet. 2016;387(10022):957-967.
8. James PA, Oparil S, Carter BL, et al. 2014
part by Johns Hopkins Cardiovascular Epidemiology Evidence-based guideline for the management of 20. Cushman WC, Evans GW, Byington RP, et al;
training grant T32HL007024 from the National high blood pressure in adults: report from the panel ACCORD Study Group. Effects of intensive
Heart, Lung, and Blood Institute. members appointed to the Eighth Joint National blood-pressure control in type 2 diabetes mellitus.
Role of the Funder/Sponsor: The funding sources Committee (JNC 8). JAMA. 2014;311(5):507-520. N Engl J Med. 2010;362(17):1575-1585.
had no role in the design and conduct of the study; 9. Weber MA, Schiffrin EL, White WB, et al. 21. Margolis KL, OConnor PJ, Morgan TM, et al.
collection, management, analysis, and Clinical practice guidelines for the management of Outcomes of combined cardiovascular risk factor
interpretation of the data; preparation, review, or hypertension in the community: a statement by management strategies in type 2 diabetes: the
approval of the manuscript; and decision to submit the American Society of Hypertension and the ACCORD randomized trial. Diabetes Care. 2014;37
the manuscript for publication. International Society of Hypertension. J Clin (6):1721-1728.
Disclaimer: The content is solely the responsibility Hypertens (Greenwich). 2014;16(1):14-26. 22. McBrien K, Rabi DM, Campbell N, et al.
of the authors and does not necessarily represent 10. Wright JT Jr, Williamson JD, Whelton PK, et al; Intensive and standard blood pressure targets in
the official views of the National Institutes of SPRINT Research Group. A randomized trial of patients with type 2 diabetes mellitus: systematic
Health. intensive versus standard blood-pressure control. review and meta-analysis. Arch Intern Med. 2012;
N Engl J Med. 2015;373(22):2103-2116. 172(17):1296-1303.
REFERENCES 23. Brunstrm M, Carlberg B. Effect of
11. Higgins JPT, Altman DG, Gtzsche PC, et al;
1. Kearney PM, Whelton M, Reynolds K, Muntner P, Cochrane Bias Methods Group; Cochrane Statistical antihypertensive treatment at different blood
Whelton PK, He J. Global burden of hypertension: Methods Group. The Cochrane Collaborations tool pressure levels in patients with diabetes mellitus:
analysis of worldwide data. Lancet. 2005;365 for assessing risk of bias in randomised trials. BMJ. systematic review and meta-analyses. BMJ. 2016;
(9455):217-223. 2011;343:d5928. 352:i717.
2. GBD 2013 Mortality and Causes of Death
Collaborators. Global, regional, and national