Articles

Effects of intensive blood pressure lowering on cardiovascular and renal outcomes: updated systematic review and meta-analysis

Xinfang Xie, Emily Atkins, Jicheng Lv, Alexander Bennett, Bruce Neal, Toshiharu Ninomiya, Mark Woodward, Stephen MacMahon, Fiona Turnbull, Graham S Hillis, John Chalmers, Jonathan Mant, Abdul Salam, Kazem Rahimi, Vlado Perkovic, Anthony Rodgers

Summary Background Recent guidelines have reversed previous recommendations for lower blood pressure Published Online targets in high-risk patients, such as those with cardiovascular disease, renal disease, or diabetes. This change November 6, 2015 http://dx.doi.org/10.1016/ represents uncertainty about whether more intensive blood pressure-lowering strategies are associated with greater S0140-6736(15)00805-3 reductions in risk of major cardiovascular and renal events. We aimed to assess the efficacy and safety of intensive See Online/Comment blood pressure-lowering strategies. http://dx.doi.org/10.1016/ S0140-6736(15)00816-8 Methods For this updated systematic review and meta-analysis, we systematically searched MEDLINE, Embase, and Renal Division, Department of the Cochrane Library for trials published between Jan 1, 1950, and Nov 3, 2015. We included randomised controlled Medicine, Peking University trials with at least 6 months’ follow-up that randomly assigned participants to more intensive versus less intensive First Hospital, Beijing, China (X Xie MD, Prof J Lv MD); The blood pressure-lowering treatment, with different blood pressure targets or different blood pressure changes from George Institute for Global baseline. We did not use any age or language restrictions. We did a meta-analysis of blood pressure reductions on Health, The University of relative risk (RR) of major cardiovascular events (myocardial infarction, stroke, heart failure, or cardiovascular death, Sydney, Sydney, NSW, Australia separately and combined), and non-vascular and all-cause mortality, end-stage disease, and adverse events, as (E Atkins BHlthSc, Prof J Lv, A Bennett BMedSc, well as albuminuria and progression of retinopathy in trials done in patients with diabetes. Prof B Neal MBChB, Prof M Woodward PhD, Findings We identified 19 trials including 44 989 participants, in whom 2496 major cardiovascular events were recorded Prof S MacMahon PhD, during a mean 3·8 years of follow-up (range 1·0–8·4 years). Our meta-analysis showed that after randomisation, F Turnbull PhD, Prof J Chalmers MBBS, patients in the more intensive blood pressure-lowering treatment group had mean blood pressure levels of A Salam MPharm, 133/76 mm Hg, compared with 140/81 mm Hg in the less intensive treatment group. Intensive blood pressure-lowering Prof V Perkovic MBBS, treatment achieved RR reductions for major cardiovascular events (14% [95% CI 4–22]), myocardial infarction (13% Prof A Rodgers MBChB); The George Institute for Global [0–24]), stroke (22% [10–32]), albuminuria (10% [3–16]), and retinopathy progression (19% [0–34]). However, more Health, Nuffield Department of intensive treatment had no clear effects on heart failure (15% [95% CI –11 to 34]), cardiovascular death (9% [–11 to 26]), Population Health, University total mortality (9% [–3 to 19]), or end-stage (10% [–6 to 23]). The reduction in major cardiovascular events of Oxford, Oxford, UK (Prof was consistent across patient groups, and additional blood pressure lowering had a clear benefit even in patients with M Woodward, Prof S MacMahon, Prof K Rahimi PhD); Department systolic blood pressure lower than 140 mm Hg. The absolute benefits were greatest in trials in which all enrolled of Medicine and Clinical Science, patients had vascular disease, renal disease, or diabetes. Serious adverse events associated with blood pressure lowering Graduate School of Medical were only reported by six trials and had an event rate of 1·2% per year in intensive blood pressure-lowering group Sciences, Kyushu University, participants, compared with 0·9% in the less intensive treatment group (RR 1·35 [95% CI 0·93–1·97]). Severe Japan (Prof T Ninomiya PhD); Department of Cardiology, hypotension was more frequent in the more intensive treatment regimen (RR 2·68 [1·21–5·89], p=0·015), but the Royal Perth Hospital, absolute excess was small (0·3% vs 0·1% per person-year for the duration of follow-up). Wellington Street, Perth, WA, Australia (Prof G S Hillis MBChB); Interpretation Intensive blood pressure lowering provided greater vascular protection than standard regimens. In Primary Care Unit, Department of Public Health and Primary high-risk patients, there are additional benefits from more intensive blood pressure lowering, including for those Care, University of Cambridge, with systolic blood pressure below 140 mmHg. The net absolute benefits of intensive blood pressure lowering in Cambridge, UK (Prof J Mant MD) high-risk individuals are large. Correspondence to: Prof Anthony Rodgers, Funding National Health and Medical Research Council of Australia. The George Institute for Global Health, The University of Sydney, PO Box M201, Missenden Road, Introduction Joint National Commitee guideline raised the target blood Sydney, NSW 2050, Australia Several major hypertension guidelines have recently pressure level for individuals older than 60 years of age to [email protected] raised target blood pressures for some high-risk patient 150/90 mm Hg.1 Globally, just under half of the total or 1–3 Prof Jicheng Lv, Renal Division, populations. Previous guidelines recommended target blood pressure-attributable disease burden occurs in Department Of Medicine, Peking blood pressure levels of around 130/85 mm Hg for people with systolic blood pressure lower than University First Hospital, Beijing, patients with cerebrovascular disease, coronary heart 140 mm Hg4 and most cardiovascular events occur in China disease, renal disease, and diabetes, whereas these people who have had a previous event.5 Therefore, [email protected] guidelines now recommend target levels of 140/90 mm Hg recommendations for treatment initiation, intensification, in these patient populations. Additionally, the Eighth or maintenance for high-risk patients who have systolic www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 1 Articles

blood pressure levels below 140 mm Hg carry substantial Data extraction and quality assessment clinical and public health importance. Published reports were obtained for each trial and The most frequently cited reason for the change in standard information was extracted and placed into a guideline recommendations for high-risk patients was spreadsheet. The data extracted from each paper included the findings from the ACCORD trial,6 which randomly baseline patient characteristics (age, sex, mean systolic assigned 4733 patients with type 2 diabetes to intensive and diastolic blood pressure levels, history of diabetes, or standard blood pressure-lowering therapy (target history of hypertension, and chronic kidney disease), systolic blood pressure <120 mm Hg vs <140 mm Hg) blood pressure control target in each group, blood and did not report a significant difference in overall pressure-lowering agents used, follow-up duration, mean cardiovascular event rates associated with a 14 mm Hg reduction of systolic and diastolic blood pressure during mean difference in systolic blood pressure. By contrast, a the trial, outcome events, and adverse events. We judged systematic review of trials of more versus less blood study quality by evaluating trial procedures for pressure lowering did note a significant reduction in randomisation, concealment of treatment allocation, major vascular events.7 In view of the uncertainty and the completeness of follow-up, and use of intention-to-treat completion of several additional trials8–10 (Mant J, analysis. The Cochrane Collaboration’s tool was used to University of Cambridge, personal communication), we assess risk of bias. Any disagreements in abstracted data sought to undertake an updated systematic review of all were adjudicated by a third reviewer (AR). trials comparing different blood pressure targets, with a particular focus on the efficacy and safety of additional Outcomes blood pressure lowering in high-risk individuals whose Outcomes of interest were: major cardiovascular events, systolic blood pressure is lower than 140 mm Hg. defined as a myocardial infarction, stroke, heart failure, or cardiovascular death, separately and combined; non- Methods vascular and all-cause mortality; end-stage kidney disease; Search strategy and selection criteria and adverse events. Progression of albuminuria (defined as We did an update of a systematic review7 of the literature, new onset of micro-albuminuria/macro-albuminuria or a using the same methods and to allow the reporting change from micro-albuminuria to macro-albuminuria) standards recommended by the PRISMA statement for and retinopathy (retinopathy progression of two or more See Online for appendix meta-analyses of intervention studies (appendix pp 11–13).11 steps) were also recorded for trials that were done in We identified relevant studies by searching the following patients with diabetes. databases using Ovid: MEDLINE (from Jan 1, 1950, to Nov 3, 2015), Embase (from 1966 to Nov 3, 2015) and the Statistical analysis Cochrane Library database (on Nov 3, 2015), using relevant For each study, we calculated individual relative risks (RRs) keywords and medical subject headings that included all and 95% CIs for each outcome before pooling. We obtained spellings of antihypertensive agents, target blood pressure, summary estimates of RRs by using a random-effects intensive blood pressure treatment, intensive blood DerSimonian model, with secondary analyses as a fixed- pressure control, strict blood pressure treatment, strict effects model. For blood pressure, we used the mean blood pressure control, tight blood pressure treatment, difference between the groups who received intensive and tight blood pressure control (appendix pp 21–25). versus standard blood pressure-lowering treatment during Trials were eligible for inclusion if they had different blood the trial. We estimated the percentage of variability across pressure targets or different blood pressure changes studies attributable to heterogeneity beyond chance using between the more versus less blood pressure-treated the I² statistic.12 We assessed small study bias by stratifying groups. Our search was limited to randomised controlled treatment effects by sample size. We explored evidence for trials with at least 6 months’ follow-up, but we did not use heterogeneity in estimates of treatment effect attributable any age or language restrictions. Reference lists from to the baseline characteristics of the trials by comparing identified trials and review articles were manually scanned summary results obtained from subsets of studies grouped to identify any other relevant studies. We also searched the by number of patients, cardiovascular event rate, age, ClinicalTrials.gov website for randomised trials that were diabetes, blood pressure target, and blood pressure level at registered as completed but not yet published. baseline. We also explored potential heterogeneity using The updated literature search, data extraction, and quality univariate meta-regression. A two-sided p value less than assessment were done independently by two authors (XX 0·05 was regarded as statistically significant. Stata and AB) using a standardised approach (see appendix version 13.0 was used for all statistical analyses. pp 2–10, 21–25). All completed randomised controlled trials that compared more versus less intensive blood pressure Role of the funding source targets with pharmacological blood pressure-lowering The funders of the study had no role in study design; in the agents were eligible for inclusion, including those that collection, analysis, or interpretation of data; in the writing enrolled participants with hypertension, at high risk of of the report; or in the decision to submit for publication. cardiovascular or renal disease, or both. XX, EA, JL, and AR had full access to all the data, and XX

2 www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 Articles

and AR were responsible for the decision to submit for publication. 4300 articles found in database searches 1788 MEDLINE 2105 Embase Results 407 Cochrane Our search strategy found 4300 records. Once duplicates had been removed, 2979 abstracts were screened, and 1321 duplicates excluded 95 publications were selected for full-text review. This process yielded 21 publications from 19 trials (figure 1)6,8–10,13–26 (Mant J, personal communication). These 2979 abstracts reviewed trials included 44 989 participants with 2496 major cardiovascular events (14 studies), 1762 all-cause deaths 2887 excluded (19 studies), and 514 end-stage kidney disease events 1136 not original investigations 280 not randomised controlled trials (eight studies; appendix pp 2–8). All the trials had an open- 261 did not assess blood pressure lowering label design with very few patients lost to follow-up 1122 did not assess a different target or (0–4·9%). Mean follow-up was 3·8 years (range relevant outcome 18 trials of hypertension in pregnancy 1·0–8·4 years). The risk of bias varied substantially across 16 not human studies the studies (appendix pp 9–10). 3 more articles from other sources 54 other publications from the same trial Of the 19 trials, five (6960 participants) enrolled only 6,15–17,22 patients with diabetes, and six specifically recruited 95 full-text articles reviewed participants with chronic kidney disease (2809 participants).13,18–21,25 One of the studies was done in 74 excluded children with chronic kidney disease and hypertension 8 not randomised controlled trials (385 participants, with a mean age of 11·5 years).25 5 not original investigations Two trials (609 participants) recruited patients with 46 did not assess a different target or relevant outcome diabetes but without hypertension, who had a mean 13 other publications from the same trial baseline blood pressure of 136/84 mm Hg17 and 2 follow-up of <6 months 126/84 mm Hg.22 The other 17 trials (44 380 participants) recruited mostly patients with hypertension and vascular 21 publications (19 trials) included disease, kidney disease, diabetes, or other risk factors. The mean baseline blood pressure levels in the trials of Figure 1: Study selection adults were between 123 and 172 mm Hg for systolic The diagram summarises search results between Jan 1, 1950, and July 31, 2014. The search was updated on Nov 3, blood pressure and between 76 and 105 mm Hg for 2015, with results restricted to papers published from 2014 to that date. MEDLINE found 133 records, Embase 50 records, and Cochrane Central 68 records. All 251 records were screened for eligibility, but no new diastolic blood pressure, with an overall mean of studies meeting the inclusion criteria were identified. 159/92 mm Hg. The mean follow-up blood pressure levels in the less intensive blood pressure-lowering Data about the effects of intensive blood pressure regimen group were 140/81 mm Hg, compared with lowering on major cardiovascular events were available 133/76 mm Hg in the more intensive treatment group. from 14 trials including 43 483 participants and The blood pressure targets varied across the trials, and 2496 cardiovascular events6,8–10,14–17,19,21,23,24,26 (Mant J, some trials had change in blood pressure as the target personal communication; figure 2A). Overall, more endpoint, whereas others had a mixture of target level intensive blood pressure-lowering regimens reduced the and change in blood pressure. Several trials had risk of major cardiovascular events by 14% (95% CI 4–22) intensive group targets of lower than 140–150 mm Hg compared with less intensive regimens, without evidence systolic blood pressure and lower than 85–90 mm Hg of major heterogeneity in the size of effect across diastolic blood pressure,10,15,23,26 whereas in other studies, included studies (figure 2A). which tended to be more recent, systolic blood pressure Myocardial infarction was reported in 13 trials including targets in the intensive groups were were 20–30 mm Hg 42 389 participants, in whom 864 events were below these levels.6,9,18,21,24 Eight trials had diastolic blood recorded6,8–10,14–17,21,23,24,26 (Mant J, personal communication; pressure targets below 80 mm Hg.8,13,14,16–18,21,22 However, figure 2B). More intensive blood pressure-lowering therapy many trials did not achieve these targets for most reduced the risk of myocardial infarction by 13% (95% CI patients; as appendix pp 2–8 show, the mean follow-up 0–24) compared with less intensive regimens. 14 trials blood pressure was above the trial target in the more (43 483 participants) reported 1099 stroke events, and more intensive group for eight (42%) of the 19 trials and was intensive blood pressure-lowering regimens were above target in the less intensive group in two (16%) of associated with a 22% reduction (95% CI 10–32) in the risk the 19 trials (data were unavailable for one trial). Across of stroke compared with less intensive all trials, the weighted mean follow-up difference in regimens6,8–10,14–17,19,21,23,24,26 (Mant J, personal communication; blood pressure between the intensive versus less figure 2C). The magnitude of the risk reductions recorded intensively treated groups was 6·8/3·5 mm Hg. for stroke and myocardial infarction in this meta-analysis www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 3 Articles

A Blood pressure Relative risk Weight difference (mm Hg) (95% CI) (%)

HOT (1998)14 –2·9/–3·1 0·93 (0·80–1·09) 16·1 UKPDS-HDS (1998)15 –10·0/–5·0 0·69 (0·55–0·86) 11·8 ABCD (H) (2000)16 –6·0/–8·0 0·91 (0·60–1·37) 5·0 ABCD (N) (2001)17 –9·0/–6·0 0·97 (0·64–1·47) 5·0 REIN-2 (2005)21 –4·1/–2·8 0·80 (0·22–2·93) 0·6 JATOS (2008)23 –9·7/–3·3 1·05 (0·73–1·53) 5·9 Cardio-Sis (2009)24 –3·8/–1·5 0·53 (0·30–0·94) 2·9 VANLISH (2010)26 –5·4/–1·7 0·87 (0·54–1·39) 4·1 AASK (2010)19 –13·0/–7·0 1·09 (0·86–1·38) 11·2 ACCORD (2010)6 –14·2/–6·7 0·88 (0·74–1·05) 14·7 HOMED-BP (2012)8 –1·3/–0·8 1·04 (0·60–1·79) 3·1 SPS3 (2013)9 –11·0/NA 0·86 (0·71–1·05) 13·3 Wei et al (2013)10 –14·0/–5·9 0·59 (0·41–0·85) 6·1 PAST-BP (2015)* –2·0/–2·4 0·22 (0·03–1·84) 0·2 Overall –6·8/–3·5 0·86 (0·78–0·96), p=0·005 Events/population: 1090/18 767 vs 1406/24 716 I2=22·4%, p=0·21 Fixed-effects model 0·87 (95% CI 0·81–0·94), p<0·0001 0·25 1·0 2·5

B

HOT (1998)14 –2·9/–3·1 0·82 (0·61–1·11) 20·0 UKPDS-HDS (1998)15 –10·0/–5·0 0·80 (0·60–1·05) 22·9 ABCD (H) (2000)16 –6·0/–8·0 1·12 (0·56–2·25) 3·7 ABCD (N) (2001)17 –9·0/–6·0 1·30 (0·68–2·49) 4·1 REIN-2 (2005)21 –4·1/–2·8 1·00 (0·14–7·02) 0·5 JATOS (2008)23 –9·7/–3·3 1·00 (0·32–3·09) 1·4 Cardio-Sis (2009)24 –3·8/–1·5 0·66 (0·19–2·33) 1·1 VANLISH (2010)26 –5·4/–1·7 1·25 (0·34–4·66) 1·0 ACCORD (2010)6 –14·2/–6·7 0·87 (0·69–1·09) 32·8 HOMED-BP (2012)8 –1·3/–0·8 0·71 (0·23–2·25) 1·3 SPS3 (2013)9 –11·0/NA 0·91 (0·58–1·42) 8·9 Wei et al (2013)10 –14·0/–5·9 0·99 (0·40–2·48) 2·1 PAST-BP (2015)* –2·0/–2·4 1·08 (0·07–17·18) 0·2 Overall –6·6/–3·4 0·87 (0·76–1·00), p=0·042 Events/population: 397/18 227 vs 467/24 162 I2=0·0%, p=0·99 Figure 2: Effects of intensive Fixed-effects model 0·87 (95% CI 0·76–1·00), p=0·042 0·25 1·0 2·5 blood pressure lowering on risk of cardiovascular C outcomes Forest plots showing the effects HOT (1998)14 –2·9/–3·1 0·87 (0·68–1·11) 19·8 of intensive versus less intensive UKPDS-HDS (1998)15 –10·0/–5·0 0·58 (0·37–0·90) 8·4 blood pressure-lowering ABCD (H) (2000)16 –6·0/–8·0 0·98 (0·40–2·43) 2·3 regimens on major 17 cardiovascular events (A), ABCD (N) (2001) –9·0/–6·0 0·32 (0·10–0·95) 1·6 myocardial infarction (B), and REIN-2 (2005)21 –4·1/–2·8 0·33 (0·04–3·17) 0·4 stroke (C). NA=not available. JATOS (2008)23 –9·7/–3·3 1·04 (0·69–1·59) 9·3 *Mant J, personal Cardio-Sis (2009)24 –3·8/–1·5 0·44 (0·14–1·42) 1·4 communication. A p value VANLISH (2010)26 –5·4/–1·7 0·70 (0·37–1·32) 4·5 <0·05 represents a significant AASK (2010)19 –13·0/–7·0 0·96 (0·64–1·44) 9·7 pooled point estimate of 6 relative risk. Boxes and ACCORD (2010) –14·2/–6·7 0·58 (0·39–0·88) 9·8 8 horizontal lines represent HOMED-BP (2012) –1·3/–0·8 1·25 (0·65–2·40) 4·3 relative risk and 95% CI for each SPS3 (2013)9 –11·0/NA 0·83 (0·66–1·04) 22·0 trial. The vertical dashed line on Wei et al (2013)10 –14·0/–5·9 0·58 (0·35–0·97) 6·5 each plot represents the point PAST-BP (2015)* –2·0/–2·4 0·00 (0·00–1·66×109) 0·0 estimate of overall relative risk. Overall –6·8/–3·5 0·78 (0·68–0·90), p=0·001 The size of each box is Events/population: 448/18 767 vs 651/24 716 proportional to weight of that I2=12·7%, p=0·32 trial result. Diamonds represent Fixed-effects model 0·79 (95% CI 0·71–0·90), p<0·0001 0·25 1·0 2·5 the 95% CI for pooled estimates Favours more intensive Favours less intensive of effect and are centred on blood pressure control blood pressure control pooled relative risk.

4 www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 Articles

Coronary heart disease Stroke Observed in current meta-analysis of trials of more vs less intensive blood pressure-lowering treatment 13% 22% Observed* in blood pressure difference trials28 16% 31% Observed* in trials of blood pressure lowering vs control, in non-hypertensive patients with cardiovascular disease29 20% 23% Expected* from cohort studies for people with systolic blood pressure >140 mm Hg27 19% 26% Expected* from cohort studies for people with systolic blood pressure 120–139 mm Hg27 19% 26%

*The associations observed in cohort studies for participants aged 55–69 years27 and the reductions shown in trials of blood pressure lowering vs control28 are shown, standardised to the 7 mm Hg systolic blood pressure difference recorded in the present meta-analysis (eg, previous blood pressure difference trials showed a relative risk for stroke of 0·59 with a 10 mm Hg systolic blood pressure reduction,28 so one would expect a 31% reduction for 7 mm Hg lower systolic blood pressure, because 0·597/10=0·69 and the relation between blood pressure change and relative risk is log-linear). The exception is for data from Thompson and colleagues’ study,29 in which the overall systolic blood pressure difference was not available.

Table 1: Comparison of expected and observed effects of a 7 mm Hg systolic blood pressure decrease on coronary heart disease and stroke outcomes

Trials Events (n)/patients (n) Mean blood pressure Relative risk (n) difference (mm Hg) (95% CI)

Other major vascular events and renal outcomes Heart failure 10 189/13 698 vs 221/19 608 –7·2/–4·0 0·85 (0·66–1·11) End-stage kidney disease 8 248/4533 vs 266/4157 –9·4/–5·1 0·90 (0·77–1·06) Albuminuria 3 926/2661 vs 988/2563 –10·1/–6·4 0·90 (0·84–0·97) Retinopathy 4 342/1421 vs 351/1244 –11·2/–6·3 0·81 (0·66–1·00) Mortality Cardiovascular death 13 354/18 209 vs 446/24 163 –6·9/–3·5 0·91 (0·74–1·11) Non-cardiovascular death 12 385/18 027 vs 476/23 966 –6·9/–3·6 0·98 (0·86–1·13 Overall mortality 19 794/19 537 vs 968/25 452 –6·8/–3·5 0·91 (0·81–1·03)

0·25 1·0 2·5

Favours more intensive blood pressure controlFavours less intensive blood pressure control

Figure 3: Effect of intensive versus less intensive blood pressure lowering on the risk of other major vascular events, renal outcomes, and mortality Weights are from random-effects analysis. Diamonds represent the 95% CI for pooled estimates of effect. were similar to those anticipated from large cohort studies27 across the included studies (I²=63·7%, p=0·041, appendix and with those in a meta-analysis of a broader set of all p 17) which is mostly attributable to the result of the blood pressure difference trials (ie, treatment vs control ACCORD study.6 A sensitivity analysis excluding ACCORD and more intensive vs less intensive;28 see table 1). resulted in a risk reduction of 25% (RR 0·75 [95% CI Ten trials (33 306 participants) reported 410 occurrences of 0·65–0·86], p<0·0001) with a much reduced I² value of heart failure,6,10,14–17,19,21,23,24 with a non-significant reduction in 18·1%. Notably, there were significant imbalances in several this outcome in patients allocated to intensive blood of the baseline characteristics between randomised groups pressure-lowering regimens compared with less intensive in this substudy of ACCORD.6 regimens (p=0·24; figure 3, appendix p 14). Eight trials Our analysis showed no clear effect of more intensive including 8690 participants recorded 514 end-stage kidney blood pressure lowering on the risk of cardiovascular death disease outcomes.6,13,15,18–21,25 Compared with less intensive (p=0·36; figure 3, appendix p 18),6,8–10,14–17,19,21,23,26 (Mant J, blood pressure lowering, the more intensive regimen did personal communication), non-cardiovascular death not significantly affect the risk of end-stage kidney disease (p=0·802; figure 3, appendix p 19),6,8–10,14–17,19,21,23,26 or all-cause (p=0·22; figure 3, appendix p 15). Three trials6,15,16 reported death (p=0·135; figure 3, appendix p 20)6,8–10,13–26 (Mant J, data about progression of albuminuria (5224 participants personal communication) as compared with less intensive and 1924 events) and showed that more intensive blood blood pressure control, with confidence intervals that were pressure control reduced the risk of albuminuria compatible with modest effects in either direction. progression by 10% (95% CI 3–16) compared with less No evidence suggests that the observed effects of more intensive control (p=0·004; figure 3) with no evidence of intensive blood pressure-lowering regimens on major heterogeneity (I²=0·0%, p=0·65; appendix p 16). vascular events differed across trial subgroups defined Progression of retinopathy was reported by four trials with according to a broad range of baseline characteristics 2665 participants and 693 events.6,15–17 A borderline (p values for heterogeneity all greater than 0·05; figure 4). significant reduction in retinopathy occurred with more In particular, there was no clear evidence that the benefits intensive blood pressure lowering (p=0·048, figure 3) but of more intensive blood pressure lowering varied by the substantial heterogeneity in the magnitude of the effect starting mean baseline blood pressure of the trial www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 5 Articles

Subgroup of major Trials (n) Relative risk p value for cardiovascular events (95% CI) heterogeneity

Sample size (n) <1200 8 0·78 (0·63–0·97) 0·16 ≥1200 6 0·91 (0·83–1·00)

Mean age (years) <62 7 0·91 (0·79–1·04) 0·31 ≥62 7 0·81 (0·69–0·96) Follow-up (years) <4 8 0·92 (0·81–1·04) 0·20 ≥4 6 0·81 (0·69–0·94) Cardiovascular event rate <2·2% 7 0·92 (0·82–1·02) 0·18 ≥2·2% 7 0·81 (0·67–0·97) Diabetes mellitus Yes 5 0·83 (0·71–0·96) 0·76 No or mixture 9 0·88 (0·76–1·01) Chronic kidney disease Yes 2 1·08 (0·86–1·36) 0·13 No or mixture 12 0·84 (0·76–0·93) Baseline systolic blood pressure (mm Hg) 120–139 3 0·89 (0·76–1·05) 140–160 7 0·83 (0·68–1·00) 0·60 >160 4 0·89 (0·73–1·09) Target blood pressure (mm Hg) Systolic blood pressure <130 and/or diastolic blood pressure <80 8 0·89 (0·80–0·99) 0·15 Systolic blood pressure ≥130 or diastolic blood pressure ≥80 6 0·81 (0·61–1·08) Blood pressure targets met by most patients Yes 8 0·82 (0·73–0·93) 0·17 No 5 0·94 (0·80–1·11) Target systolic blood pressure in intensive group <140 or <150 mm Hg 4 0·76 (0·60–0·97) 0·06 <120–<130 mm Hg or diastolic target 10 0·91 (0·84–1·00) Mean follow-up systolic blood pressure (mm Hg) ≥140 5 0·91 (0·76–1·10) 0·10 130–139 8 0·82 (0·73–0·92) Systolic blood pressure difference (mm Hg) <6 6 0·87 (0·73–1·05) 0·66 ≥6 8 0·86 (0·76–0·99) Jadad score <3 7 0·89 (0·79–1·00) 0·73 ≥3 7 0·85 (0·73–1·01)

Overall 14 0·86 (0·78–0·96)

0·5 1·0 2·0

Favours more intensive blood pressure controlFavours less intensive blood pressure control

Figure 4: Effects of intensive blood pressure lowering on the risk of major cardiovascular events in subgroups of trials Weights are from random-effects analysis. Boxes and horizontal lines represent relative risk and 95% CI for each trial. The vertical dashed line represents the overall point estimate relative risk according to the horizontal axis. The size of each box is proportional to weight of that trial result. The diamond represents the 95% CI for pooled estimates of effect and is centred on pooled relative risk.

participants, the absolute level of the systolic or diastolic pressure lowering on major cardiovascular outcomes target set for the intensive group, or the mean follow-up according to the baseline characteristics also showed no blood pressure in the control group. After excluding the evidence of heterogeneity (appendix p 10). As a result of the four trials with relatively high systolic blood pressure consistent proportional reductions, absolute benefits were targets in the intensive group (150 mm Hg in one trial15 and proportional to absolute risk. For trials in which all patients 140 mm Hg in three trials10,23,26), the reduction in major had vascular disease, renal disease, or diabetes at baseline, vascular events was still evident (RR 0·91 [95% CI the average control group rate of major vascular events was 0·84–1·00]). Univariate meta-regression of intensive blood 2·9% per year compared with 0·9% per year in other trials,

6 www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 Articles

Trials Participants Total events, n (event rate per Relative risk (95% CI) p value (n) (n) person-year, %) More intensive Less intensive regimen regimen Serious adverse events associated with blood 6 12 265 280 (1·2%) 211 (0·9%) 1·35 (0·93–1·97) 0·112 pressure lowering* Severe hypotension 5 10 089 61 (0·3%) 23 (0·1%) 2·68 (1·21–5·89) 0·015 Dizziness 5 9778 589 (3·1%) 536 (2·8%) 1·11 (0·99–1·23) 0·111 Adverse events leading to discontinuation of 4 9665 169 (1·0%) 174 (1·0%) 0·96 (0·72–1·28) 0·787 treatment

*For three trials, total serious adverse events are reported, whereas for the other three trials, serious adverse events associated with blood pressure lowering were reported.

Table 2: Adverse events occurring in more intensive versus less intensive blood pressure-lowering trials and the numbers needed to treat were 94 (95% CI 44–782) events, such as fewer than 1000 myocardial infarction in these trials versus 186 (107–708) in all other trials. outcomes. To maximise power, we assessed major Data for adverse effects potentially associated with cardiovascular events as a whole, but the makeup of this treatment were reported inconsistently by individual trials. composite outcome can vary by factors like patient group Table 2 summarises the data that were available. and blood pressure level. However, our review was Six trials6,9,21,25,26 (Mant J, personal communication) reported comprehensive and increased the numbers of events data for severe adverse events associated with blood available for analysis by 29% compared with our previous pressure lowering (12 265 participants and 491 events), review7 by including four new trials (Mant J, personal showing no clear effect of more intensive versus less communication).8–10 A key issue in this topic is that few intensive blood pressure lowering on severe adverse events trials have simultaneously achieved large blood pressure associated with blood pressure lowering (table 2). separations between randomised groups and recorded Five trials6,9,20,24,25 reported severe hypotension outcomes large numbers of events. In such a setting, equivocal (10 089 participants, with 61 events in the more intensive results can be expected from individual trials as a result of blood pressure-lowering treatment regimen vs 23 in the low statistical power and even meta-analyses if they are less intensive regimen) with more intensive blood pressure based on only a subset of all trials.30 This situation is control almost tripling the risk of hypotension (p=0·015). especially true for outcomes such as myocardial infarction, More intensive blood pressure control led to a trend to an which is less strongly associated with blood pressure than increased risk of dizziness (five trials with 9778 participants is stroke,27 and also for composite outcomes such as all and 1125 events [Mant J, personal communication];6,9,24,25 cardiovascular events because these will comprise quite a table 2). Finally, the rate of drug discontinuation did not high proportion of coronary heart disease events at lower differ between the more intensively and less intensively blood pressure levels. However, this review showed clearly treated groups in the four trials6,22,23,25 that reported data that the relative risk reductions for stroke and coronary (9665 participants and 343 events; table 2). heart disease were consistent with those expected for blood pressure differences of about 7 mm Hg systolic.28 Discussion Prospective observational studies in people without Our updated systematic review and meta-analysis shows major illness at baseline consistently show that that the reductions in major cardiovascular events from more association of blood pressure with risk for cardiovascular intensive blood pressure-lowering treatment regimens events is direct and continuous at blood pressure levels of aiming for lower blood pressure targets. These beneficial 115/75 mm Hg or higher.27,31,32 However, several publications effects were consistent across major patient subgroups in recent years, derived from non-randomised studies in and types of interventions and significant gains could be people with vascular disease, have suggested a J-curve achieved from further lowering of systolic blood pressure association between blood pressure and outcome.33 This to lower than 140 mm Hg. Although an increase in finding has led to concern that excessive lowering of blood hypotension occurred as a result of more intensive blood pressure could increase the risk of cardiovascular events. pressure lowering, including serious hypotensive events, The results of the present meta-analysis support the idea there was no suggestion that these adverse effects would that those observations were probably a result of outweigh the benefits of treatment in high-risk patient uncontrolled confounding, and especially reverse populations. causation, whereby major illness is itself a cause of low One of the main limitations of this review is the lack of blood pressure, rather than a consequence. Similar individual participant data, which would have allowed a concerns were voiced in the 1980s and 1990s about the more reliable assessment of treatment effects in different associations of low cholesterol levels with adverse patient groups. Additionally, there were few cause-specific outcomes,34 but these were dispelled when large www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 7 Articles

randomised trials of intensive cholesterol lowering refuted Grant, during the conduct of the study, and grants and personal fees from increases in outcomes such as cancer.35 The higher potency Servier International, outside the submitted work. AS has received grants from the National Health and Medical Research Council of Australia, of statins compared with blood pressure-lowering drugs is during the conduct of the study. KR has received grants from NIHR probably the main reason why this issue was resolved Oxford Biomedical Research Centre, Oxford Martin School, and NIHR several decades earlier for lipids than for blood pressure. Career Development Fellowship, during the conduct of the study. VP has Contrary to some expectations, this meta-analysis received grants from Abbvie, Baxter, and Janssen; has received personal fees from Merck and Servier; has been on the steering committee for provides evidence that intensive blood pressure lowering Astellas; has been on the steering commiteee or advisory board for GSK, reduces cardiovascular events, although it also shows Bristol-Myers Squibb, and Janssen; has received honoraria for speaking at that an intensive strategy does increase the risk of some science symposia from Eli Lilly, Pfizer, and Astra Zeneca; and has been on adverse effects. A definite increase in hypotension the steering committee and received honoraria for scientific presentations from Boehringer Ingleheim, outside the submitted work. AR has received occurred in patients receiving the more intensive blood grants from Servier outside the submitted work and also receives salary pressure-lowering regimen, including severe episodes of support in part from The George Institute for Global Health. George hypotension. Side-effects and quality-of-life data were Health Enterprises, the social enterprise section of The George Institute, incompletely and inconsistently collected across the has received investment for the development of fixed-dose combination therapy containing statin, aspirin, and blood pressure-lowering drugs. contributing studies, which makes interpretation of XX, JL, AB, TN, SM, FT, GSH, and JM declare no competing interests. these findings challenging. However, no evidence Acknowledgments suggested an increase in fatal or life-threatening events, This work was funded by National Health and Medical Research Council or other adverse events that were of similar severity to of Australia Program Grant APP1052555. JL was supported by the grant major cardiorenal events. Although the data about falls from the National Natural Science Foundation of China (no. 81270795), and fractures were scarce, the incidence rate would have Program for New Century Excellent Talents in University from the Ministry of Education of China grant NCET-12-0011, and Excellent Young been low in these generally less elderly patient Scientists Fund of the National Natural Science Foundation of China populations, and reassuring data are available from (no. 813220065). AR was supported by a National Health and Medical more vulnerable, elderly patients.36,37 In patient Research Council of Australia Principal Research Fellowship APP1066280. populations at high risk of vascular disease, the absolute We are very grateful to Giuseppe Remuzzi and the REIN study group for providing useful additional information for this meta-analysis. benefits of treatment will comfortably surpass the References absolute harms. 1 James PA, Oparil S, Carter BL, et al. 2014 evidence-based guideline Several research questions arise from this work, for the management of high blood pressure in adults: report from perhaps the most important of which is how best to the panel members appointed to the Eighth Joint National Committee (JNC 8). JAMA 2014; 311: 507–20. achieve and maintain greater reductions in blood 2 Mancia G, Fagard R, Narkiewicz K, et al. 2013 ESH/ESC guidelines pressure given the relatively modest blood pressure for the management of arterial hypertension: the Task Force for the differences seen in these trials. Combination therapy will Management of Arterial Hypertension of the European Society of 38–40 Hypertension (ESH) and of the European Society of Cardiology be an important part of this solution, but other (ESC). Blood Press 2013; 22: 193–278. approaches to improve treatment rates and adherence 3 National Institute for Health and Care Excellence. The clinical will be needed. The application of non-pharmacological management of primary hypertension in adults: clinical guideline approaches, such as weight loss and dietary sodium 127. London: NICE, 2011. 4 Rodgers A, Ezzati M, Vander Hoorn S, et al. Distribution of major restriction, might help but will also be insufficient for health risks: findings from the Global Burden of Disease Study. most patients. PLoS Med 2004; 1: e27. In conclusion, this review and meta-analysis provides 5 Kerr AJ, Broad J, Wells S, Riddell T, Jackson R. Should the first priority in cardiovascular risk management be those with prior clear evidence of the benefits of more intensive blood cardiovascular disease? Heart 2009; 95: 125–29. pressure lowering, including in high-risk patients whose 6 ACCORD Study Group. Effects of intensive blood-pressure control systolic blood pressure is lower than 140 mm Hg. Existing in type 2 diabetes mellitus. N Engl J Med 2010; 362: 1575. 1–3 7 Lv JC, Neal B, Ehteshami P, et al. Effects of intensive blood pressure clinical guidelines should be revised accordingly, to lowering on cardiovascular and renal outcomes: a systematic review recommend more intensive blood pressure-lowering and meta-analysis. PLoS Med 2012; 9: e1001293. treatment in high-risk patient groups. 8 Asayama K, Ohkubo T, Metoki H, et al. Cardiovascular outcomes in the first trial of antihypertensive therapy guided by self-measured Contributors home blood pressure. Hypertens Res 2012; 35: 1102–10. AR, XX, JL, and VP conceived and designed the study. XX, AB, EA, JL, 9 Benavente OR, Coffey CS, Conwit R, et al. Blood-pressure targets in and AR did the study. XX analysed the data. AR wrote the first draft of patients with recent lacunar stroke: the SPS3 randomised trial. the report, and AR, XX, EA, and JL helped to write the final version. All Lancet 2013; 382: 507–15. authors read and met the ICMJE criteria for authorship. All authors 10 Wei Y, Jin Z, Shen G, et al. Effects of intensive antihypertensive agree with the results and conclusions of the report. treatment on Chinese hypertensive patients older than 70 years. Declaration of interests J Clin Hypertens 2013; 15: 420–27. EA has received grants from the National Health and Medical Research 11 Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that Council, during the conduct of the study. BN has received grants from evaluate health care interventions: explanation and elaboration. Jannsen, Abbvie, Dr Reddy’s Laboratories, Merck Schering Plough, and Ann Intern Med 2009; 151: W65–94. Roche, and honoraria or travel reimbursement from Abbott, Novartis, 12 Woodward M. : design and data analysis, 2nd edn. Pfizer, Servier, Roche, and Janssen, outside the submitted work. MW has Boca Raton, FL: Chapman and Hall/CRC Press, 2005. received personal fees from Amgen and Novartis, and grants and personal 13 Toto RD, Mitchell HC, Smith RD, et al. “Strict” blood pressure fees from Roche, outside the submitted work. JC has received grants from control and progression of renal disease in hypertensive the National Health and Medical Research Council of Australia Program nephrosclerosis. Kidney Int 1995; 48: 851–59.

8 www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 Articles

14 Hansson L, Zanchetti A, Carruthers SG, et al. Effects of intensive 28 Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs blood-pressure lowering and low-dose aspirin in patients with in the prevention of cardiovascular disease: meta-analysis of hypertension: principal results of the Hypertension Optimal 147 randomised trials in the context of expectations from Treatment (HOT) randomised trial. HOT Study Group. Lancet 1998; prospective epidemiological studies. BMJ 2009; 338: b1665. 351: 1755–62. 29 Thompson AM, Hu T, Eshelbrenner CL, et al. Antihypertensive 15 UK Prospective Diabetes Study Group. Tight blood pressure control treatment and secondary prevention of cardiovascular disease and risk of macrovascular and microvascular complications in events among persons without hypertension: a meta-analysis. type 2 diabetes: UKPDS 38. BMJ 1998; 317: 703–13. JAMA 2011; 305: 913–22. 16 Estacio RO, Jeffers BW, Gifford N, Schrier RW. Effect of blood 30 Arguedas JA, Perez MI, Wright JM. Treatment blood pressure pressure control on diabetic microvascular complications in targets for hypertension. Cochrane Database Syst Rev 2009; patients with hypertension and type 2 diabetes. Diabetes Care 2000; 3: CD004349. 23 (suppl 2): B54–64. 31 MacMahon S, Peto R, Cutler J, et al. Blood pressure, stroke, and 17 Schrier RW, Estacio RO, Esler A, Mehler P. Effects of aggressive coronary heart disease. Part 1, prolonged differences in blood blood pressure control in normotensive type 2 diabetic patients pressure: prospective observational studies corrected for the on albuminuria, retinopathy and strokes. Kidney Int 2002; regression dilution bias. Lancet 1990; 335: 765–74. 61: 1086–97. 32 Lawes CM, Rodgers A, Bennett DA, et al. Blood pressure and 18 Schrier R, McFann K, Johnson A, et al. Cardiac and renal effects cardiovascular disease in the Asia Pacific region.J Hypertens 2003; of standard versus rigorous blood pressure control in 21: 707–16. autosomal-dominant polycystic kidney disease: results of a seven-year 33 Chalmers J. Is a blood pressure target of <130/80 mm Hg still prospective randomized study. J Am Soc Nephrol 2002; 13: 1733–39. appropriate for high-risk patients? Circulation 2011; 124: 1700–02. 19 Appel LJ, Wright JT Jr, Greene T, et al. Intensive blood-pressure 34 Jacobs D, Blackburn H, Higgins M, et al. Report of the Conference control in hypertensive chronic kidney disease. N Engl J Med 2010; on Low Blood Cholesterol: mortality associations. Circulation 1992; 363: 918–29. 86: 1046–60. 20 Sarnak MJ, Greene T, Wang X, et al. The effect of a lower target 35 Cholesterol Treatment Trialists’ (CTT) Collaboration. Efficacy and blood pressure on the progression of kidney disease: long-term safety of cholesterol-lowering treatment: prospective meta-analysis follow-up of the modification of diet in renal disease study. of data from 90 056 participants in 14 randomised trials of statins. Ann Intern Med 2005; 142: 342–51. Lancet 2005; 366: 1267–78. 21 Ruggenenti P, Perna A, Loriga G, et al. Blood-pressure control for 36 Warwick J, Falaschetti E, Rockwood K, et al. No evidence that frailty renoprotection in patients with non-diabetic chronic renal disease modifies the positive impact of antihypertensive treatment in very (REIN-2): multicentre, randomised controlled trial. Lancet 2005; elderly people: an investigation of the impact of frailty upon 365: 939–46. treatment effect in the HYpertension in the Very Elderly Trial 22 Estacio RO, Coll JR, Tran ZV, Schrier RW. Effect of intensive blood (HYVET) study, a double-blind, placebo-controlled study of pressure control with valsartan on urinary albumin excretion in antihypertensives in people with hypertension aged 80 and over. normotensive patients with type 2 diabetes. Am J Hypertens 2006; BMC Med 2015; 13: 78. 19: 1241–48. 37 Peters R, Beckett N, Burch L, et al. The effect of treatment based on 23 JATOS Study Group. Principal results of the Japanese trial to assess a diuretic (indapamide)±ACE inhibitor (perindopril) on fractures in optimal systolic blood pressure in elderly hypertensive patients the Hypertension in the Very Elderly Trial (HYVET). (JATOS). Hypertens Res 2008; 31: 2115–27. Age Ageing 2010; 39: 609–16. 24 Verdecchia P, Staessen JA, Angeli F, et al. Usual versus tight control 38 Wald DS, Law M, Morris JK, Bestwick JP, Wald NJ. Combination of systolic blood pressure in non-diabetic patients with therapy versus monotherapy in reducing blood pressure: hypertension (Cardio-Sis): an open-label randomised trial. Lancet meta-analysis on 11,000 participants from 42 trials. Am J Med 2009; 2009; 374: 525–33. 122: 290–300. 25 Wuhl E, Trivelli A, Picca S, et al. Strict blood-pressure control and 39 Law MR, Wald NJ, Morris JK, Jordan RE. Value of low dose progression of renal failure in children. N Engl J Med 2009; combination treatment with blood pressure lowering drugs: 361: 1639–50. analysis of 354 randomised trials. BMJ 2003; 326: 1427. 26 Ogihara T, Saruta T, Rakugi H, et al. Target blood pressure for 40 Feldman RD, Zou GY, Vandervoort MK, et al. A simplified approach treatment of isolated systolic hypertension in the elderly: valsartan to the treatment of uncomplicated hypertension: a cluster in elderly isolated systolic hypertension study. Hypertension 2010; randomized, controlled trial. Hypertension 2009; 53: 646–53. 56: 196–202. 27 Lewington S, Clarke R, Qizilbash N, Peto R, Collins R. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002; 360: 1903–13.

www.thelancet.com Published online November 6, 2015 http://dx.doi.org/10.1016/S0140-6736(15)00805-3 9