Blood Pressure Lowering in Type 2 Diabetes: A Systematic Review and Meta-analysis

Importance  Lowering blood pressure (BP) is widely used to reduce vascular risk in individuals with diabetes.

Objective  To determine the associations between BP–lowering treatment and vascular disease in type 2 diabetes.

Data Sources and Study Selection  We searched MEDLINE for large-scale randomized controlled trials of BP–lowering treatment including patients with diabetes, published between January 1966 and October 2014.

Data Extraction and Synthesis  Two reviewers independently extracted study characteristics and vascular outcome data. Estimates were stratified by baseline BP and achieved BP, and pooled using fixed-effects meta-analysis.

Main Outcomes and Measures  All-cause mortality, cardiovascular events, coronary heart disease events, stroke, heart failure, retinopathy, new or worsening albuminuria, and renal failure.

Results  Forty trials judged to be of low risk of bias (100 354 participants) were included. Each 10–mm Hg lower systolic BP was associated with a significantly lower risk of mortality (relative risk [RR], 0.87; 95% CI, 0.78-0.96); absolute risk reduction (ARR) in events per 1000 patient-years (3.16; 95% CI, 0.90-5.22), cardiovascular events (RR, 0.89 [95% CI, 0.83-0.95]; ARR, 3.90 [95% CI, 1.57-6.06]), coronary heart disease (RR, 0.88 [95% CI, 0.80-0.98]; ARR, 1.81 [95% CI, 0.35-3.11]), stroke (RR, 0.73 [95% CI, 0.64-0.83]; ARR, 4.06 [95% CI, 2.53-5.40]), albuminuria (RR, 0.83 [95% CI, 0.79-0.87]; ARR, 9.33 [95% CI, 7.13-11.37]), and retinopathy (RR, 0.87 [95% CI, 0.76-0.99]; ARR, 2.23 [95% CI, 0.15-4.04]). When trials were stratified by mean baseline systolic BP at greater than or less than 140 mm Hg, RRs for outcomes other than stroke, retinopathy, and renal failure were lower in studies with greater baseline systolic BP (P interaction <0.1). The associations between BP-lowering treatments and outcomes were not significantly different, irrespective of drug class, except for stroke and heart failure. Estimates were similar when all trials, regardless of risk of bias, were included.

Conclusions and Relevance  Among patients with type 2 diabetes, BP lowering was associated with improved mortality and other clinical outcomes with lower RRs observed among those with baseline BP of 140 mm Hg and greater. These findings support the use of medications for BP lowering in these patients.

By 2030, it is estimated that there will be at least 400 million individuals with type 2 diabetes mellitus worldwide, with many of those affected being relatively young and living in low- or middle-income countries.¹Quiz Ref IDType 2 diabetes is associated with a substantially increased risk of macrovascular events such as myocardial infarction (MI) and stroke.² It is also now the leading cause of blindness in developed countries and is a leading cause of end-stage kidney disease through its effects on the microvasculature.^3,4 Blood pressure (BP) levels are on average higher among individuals with diabetes and increased BP is a well-established risk factor for people with diabetes.^5,6

In general adult populations, the association of BP with disease outcomes is continuous, with increasing risks of events occurring in parallel with increasing BP levels from as low as 115 mm Hg for systolic and 75 mm Hg for diastolic.^7,8 A similar association has been reported for BP and the risks of macrovascular⁹ and microvascular¹⁰ disease in patients with type 2 diabetes. Although data for every population subset and every outcome are not available, the nature of the association of BP appears consistent across large and diverse population groups including men and women, different ethnicities, older and younger people, and among individuals with and without established vascular disease.^7,11

Lowering BP in individuals with diabetes is an area of current controversy, with particular debate surrounding who should be offered therapy and the BP targets to be achieved. A number of recent guidelines (eTable 1 in the Supplement) have focused entirely on individuals with diabetes who have been diagnosed with hypertension and have established target levels for BP lowering that are less aggressive than previous recommendations. It is unclear whether the new guidelines have used the entire evidence base with respect to recommendations for BP lowering in patients with type 2 diabetes. To address this question, we conducted a comprehensive overview of the effects of BP-lowering treatment in patients with diabetes (regardless of the presence or absence of defined hypertension). We aimed to determine the extent to which BP-lowering treatment is associated with a lower risk of macrovascular outcomes and microvascular outcomes, with a specific focus on areas of current controversy using the totality of the applicable evidence.

We conducted a systematic review and meta-analysis, using the approach recommended by the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for meta-analyses of interventional studies.¹² Relevant studies were identified using the following search terms: anti-hypertensive agents or hypertension or diuretics, thiazide or angiotensin-convertingenzyme or receptors, angiotensin/antagonists & inhibitors or tetrazoles or calcium channel blockers or vasodilator agents or the names of all BP-lowering drugs listed in the British National Formulary as keywords or text words or the MeSH (Medical Subject Headings [of the US National Library of Medicine]) term blood pressure/drug effects.¹³ We used this existing strategy to identify BP-lowering trials published on MEDLINE, from January 1, 1966, to October 28, 2014, restricted to those published in MEDLINE-defined core clinical journals. The initial search was conducted by an experienced research librarian and no language restrictions were applied. Studies were restricted to clinical trials, controlled clinical trials, randomized controlled trials, or meta-analyses. Bibliographies of included studies and bibliographies of identified meta-analyses were searched by hand. We then manually examined whether each trial included patients with diabetes and searched for any reporting of results for the diabetic subgroup.

We included all randomized controlled trials of BP-lowering treatment in which the entire trial population comprised patients with diabetes or in which the results of a diabetic subgroup were able to be obtained. Randomized trials published between January 1966 and March 2014 were included. No trial was excluded due to the presence of comorbidities at baseline. Trials conducted in type 2 diabetic patients with heart failure and after MI were included. However, trials that were conducted in patients predominantly with type 1 diabetes were excluded. Studies that did not make a specific reference to diabetes type were included, assuming that the great majority of patients in these studies had type 2 diabetes. For inclusion, all trials were required to have greater than 1000 patient-years of follow-up in each randomized group to minimize risk of bias associated with small trials.¹⁴

Two researchers screened all abstracts identified in the initial search, excluding studies that violated inclusion criteria (presented in Eligibility). Following the initial abstract screen, full texts of all identified studies were acquired. Two researchers (C.E. and T.C.) then screened full-text studies in duplicate, with differences resolved by consensus. Once all eligible trials were identified, macrovascular and microvascular outcomes for diabetic subgroups were extracted independently by 2 researchers (C.E. and T.C.). Data regarding the patient population, drug or intervention in the treatment group, drug or intervention in the control group, sample size, duration, baseline BP, achieved BP in the treatment group, and mean reduction in BP (difference between achieved BP and baseline BP in the treated group minus the same difference in the control group) were extracted. If the baseline BP, each treatment group BP, or difference in BP levels between groups was unavailable for the subgroup with diabetes, we recorded the value for the entire trial.

In addition to all-cause mortality, data regarding 4 macrovascular outcomes were extracted: cardiovascular disease (CVD) events (defined as myocardial infarction, sudden cardiac death, revascularization, fatal and nonfatal stroke and fatal and nonfatal heart failure); major coronary heart disease (CHD) events (defined as fatal and nonfatal MI and sudden cardiac death, with silent MI excluded¹³); stroke (fatal and nonfatal); and heart failure (both new diagnosis of heart failure for those without heart failure at baseline and hospitalization for individuals with heart failure at baseline). For microvascular outcomes, 3 outcomes were extracted: retinopathy (defined as progress of 3 or more steps on the Early Treatment of Diabetic Retinopathy Scale [ETDRS] ¹⁵); renal failure (defined as end-stage renal disease requiring dialysis or transplantation or death due to renal disease); and albuminuria (development of microalbuminuria, as well as the composite of new or worsening albuminuria). For one trial, retinopathy was identified as participants requiring laser therapy for treatment of retinopathy rather than progression on the ETDRS; this was included in the meta-analysis.¹⁶ For 6 trials, peripheral artery disease was included in the reported cardiovascular composite; these were included in the meta-analysis.^17-22

For each outcome, both the total number of events and the summary statistic (either relative risk or hazard ratio with 95% CIs) were extracted if available. If a hazard ratio was provided, we utilized that ratio as a relative risk for the meta-analysis because hazard ratios avoid censoring associated with use of tabular data and have greater statistical power. However, if a hazard ratio was not provided, a relative risk ratio from the total number of events was calculated and used. If the total number of events was not provided, but a summary relative risk or hazard ratio was provided, the total number of events was derived from the reported relative risk or hazard ratio, to allow for the calculation of a pooled event rate.

Quiz Ref IDWe used the Cochrane risk of bias tool to evaluate this risk of methodological quality within included trials.²³ Selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and investigators), detection bias (blinding of evaluators), attrition bias (incomplete outcome data), and reporting bias (selective outcome reporting) were judged to be of either low, unclear or high risk. The trial, as a whole, was then judged to be of low, unclear, or high risk of bias, based on whether the level of bias in domains may have led to material bias in the outcomes of interest.

For all analyses, overall estimates of association were calculated using inverse variance weighted fixed-effects analysis with 95% CIs. Fixed-effects analysis was used because heterogeneity was low to moderate²⁴ and random-effects analysis, in certain circumstances, can give inappropriate weight to smaller studies.²⁵ Heterogeneity was characterized using the I² statistic. Tests of interaction between subgroups were performed using Cochran Q statistic. For our primary analysis, we restricted meta-analysis to trials judged to be of low risk of bias (total, 40).

Analyses were performed to determine: (1) whether BP-lowering treatment is associated with lower risks of all-cause mortality, macrovascular outcomes, and microvascular outcomes, and the magnitude of any such associations; (2) the associations between BP-lowering treatment and these outcomes stratified by initial BP level; (3) the associations between BP-lowering treatment and these outcomes stratified by achieved BP level; and (4) the associations between BP-lowering treatment and these outcomes based on different classes of BP-lowering medication used.

For the first objective, we included all trials that assigned individuals to a BP-lowering medication vs placebo or assigned individuals to higher or lower BP targets. For 2 trials that had 3 groups including a placebo, and for 1 trial that had 4 groups including a placebo, the active groups were combined for analysis by adding together all events and taking a weighted average of baseline BP, achieved BP, and BP reduction.^26-28

To determine the standardized associations of BP-lowering treatment for every 10–mm Hg reduction in systolic BP, our primary analysis, the log of the summary statistic of each trial (relative risk or hazard ratio) and the standard error of the log was multiplied by 10–mm Hg/systolic BP reduction. For example, if the log-hazard ratio was −0.1 and the BP reduction was 5 mm Hg, the standardized log-hazard ratio was calculated to be −0.1 (10 mm Hg/ 5 mm Hg) = −0.2. If a standardized association was significant, we additionally calculated the absolute risk reduction (ARR) for each significant outcome in events per 1000 patient-years of follow-up. We derived the ARR for each outcome using the formula ARR = (1-RR [relative risk per 10–mm Hg lower SBP])(ACR [assumed control risk]).²³ We used the pooled event rate (in events per 1000 patient-years) in the control groups of meta-analyzed trials for each outcome as the ACR. The number needed to treat over 10 years (NNT) was derived from the inverse of the ARR (per 10 patient-years) and its 95% CI. For trials that did not report the number of events (and only reported an RR), we derived the number of events from the reported RR to allow for calculation of ARR.

Because many trials conducted in patients with diabetes and heart failure did not report the magnitude of BP reduction, and therefore could not be standardized, we examined whether there was heterogeneity between heart failure trials and non–heart failure trials in nonstandardized analyses.

To determine if the associations between BP-lowering treatment and outcomes varied by initial systolic BP, we stratified trials into categories of 140 mm Hg and greater and less than 140 mm Hg, according to the mean baseline BP of participants. Similarly, to determine if the associations between BP-lowering treatment and outcomes varied by achieved BP, we stratified trials into categories of 130 mm Hg and greater and less than 130 mm Hg according to the mean systolic BP achieved in the intervention group. As a sensitivity analysis, we: (1) excluded trials (total, 12) from relevant analyses in which it was necessary to impute BP data (either baseline, follow-up, or difference between groups) for the diabetic subgroup from the entire population; (2) examined nonstandardized associations by baseline and achieved BP; and (3) included 5 trials that were judged to be of unclear or high risk of bias.

Possible variation in the associations between BP-lowering treatment and outcomes by class of medication was examined by comparing all trials that tested a given class of medication (angiotensin-converting enzyme [ACE] inhibitor, angiotensin II receptor blocker [ARB], β-blocker, diuretic, calcium channel blocker [CCB]) against another class. One trial comparing classes of medication had 3 groups, but only provided summary statistics (RR reductions) for macrovascular outcomes.²² Because inclusion of both comparisons would have led to duplicate counting of participants, we randomly selected a group for inclusion in the analysis. We did not include microvascular outcomes in the analysis by class of medication because there were not enough trials to allow for meta-analysis.

Analyses were performed using STATA statistical software version 12 (StataCorp) and R version 3.0 (R Project).

In total, 10 598 abstracts were screened and 10 251 excluded (Figure 1). Next, 347 studies were screened in a full-text review and of these, an additional 213 were excluded. Of the 134 randomized trials identified, 89 trials were excluded, the majority because they did not provide separate results for patients with diabetes.

Thirty-three were considered trials that tested BP lowering, by either comparing a BP-lowering drug against a placebo (26 trials) or comparing BP lowering to different target levels (7 trials) (Table). Seventeen trials compared different classes of drugs against each other. At minimum, all 45 trials included a report about all-cause mortality or 1 prespecified macrovascular outcome, and 16 trials included a report about 1 of the 3 prespecified microvascular outcomes. Twelve BP-lowering trials reported only a baseline BP or achieved BP for the entire study population, but not separately for the diabetic subgroup. One trial was judged to be of high risk of bias due to evidence of poor allocation concealment, while 4 trials were judged to be of unclear risk of bias, and 40 trials of low risk of bias (eTable 2 in the Supplement). Therefore, 40 trials involving 100 354 participants formed the primary analysis, with an additional 5 trials (at unclear or high risk of bias) involving 4232 participants included in secondary analyses.

A 10–mm Hg reduction in systolic BP was associated with a significantly lower risk of all-cause mortality (RR, 0.87 [95% CI, 0.78-0.96]), CVD events (RR, 0.89 [95% CI, 0.83-0.95]), CHD events (RR, 0.88 [95% CI, 0.80-0.98]), and stroke events (RR, 0.73 [95% CI, 0.64-0.83]). The associations for heart failure events (RR, 0.86 [95% CI, 0.74-1.00]) and renal failure (RR, 0.91 [95% CI, 0.74-1.12]) were not significant (Figure 2, eFigures 1-8 in the Supplement). For microvascular outcomes, a 10 mm Hg-lower systolic BP was associated with a lower risk of retinopathy (RR, 0.87 [95% CI, 0.76-0.99]) and albuminuria (RR, 0.83 [95% CI, 0.79-0.87]) (Figure 2). The corresponding ARRs, in events per 1000 patient-years of follow-up, were 3.16 (95% CI, 0.90-5.22) for all-cause mortality (number needed to treat [NNT] over 10 years = 32 [95% CI, 19-111]); ARR was 3.90 for CVD events (95% CI, 1.57-6.06; NNT = 26 [95% CI, 17-64]); ARR was 1.81 for CHD events (95% CI, 0.35-3.11; NNT = 55 [95% CI, 32-284]); ARR was 4.06 for stroke events (95% CI, 2.53-5.40; NNT = 25 [95% CI, 19-40]); ARR was 2.23 for retinopathy events (95% CI, 0.15-4.04; NNT = 45 [95% CI, 25-654]); and ARR was 9.33 for albuminuria (95% CI, 7.13-11.37; NNT = 11 [95% CI, 9-14]).

Analyses not standardized to a 10–mm Hg BP reduction showed broadly similar results to standardized analyses (eFigures 9-16 in the Supplement) except that associations between BP-lowering treatment and the risk of heart failure (RR, 0.81 [95% CI, 0.76-0.86]) and renal failure (RR, 0.88 [95% CI, 0.79-0.99]) were statistically significant. Estimates were also similar in trials that were conducted in heart failure/left ventricular systolic dysfunction (HF/LVSD) patients and non-HF/LVSD patients, with the exception of heart failure as the outcome, where the RR was significantly lower in patients with HF/LVSD (eFigure 17 in the Supplement).

Trials were stratified by mean initial systolic BP level (range, ≥140 mm Hg to <140 mm Hg) and the associations of outcomes with a 10-mm Hg systolic BP lowering compared between strata (Figure 3). Significant interactions were observed for mortality, CHD, CVD, and heart failure (all P < .1), with lower relative risks observed among those trials with mean baseline systolic BP of 140 mm Hg or greater and no significant associations among the group with baseline systolic BP of less than 140 mm Hg. BP-lowering treatment was associated with lower risks of stroke and albuminuria, regardless of initial systolic BP. Excluding trials with imputed BP values did not alter the main findings (eFigure 18 in the Supplement). Conclusions were broadly similar for the nonstandardized analyses (eFigure 19 in the Supplement), although tests for interactions varied for some outcomes. Nonstandardized estimates were similar when 1 trial at unclear risk of bias was included (eFigure 20 in the Supplement).

Trials were stratified by the systolic BP achieved in the treatment group (≥130 or <130 mm Hg) and the associations of a 10–mm Hg systolic BP lowering with risk compared between strata (Figure 4). Significant interactions were observed for mortality, CHD, CVD, heart failure, and albuminuria (all P < .1), with lower relative risks in the 130–mm Hg or greater stratum than the lower than 130–mm Hg stratum. No significant interaction was observed for stroke or retinopathy. The associations between BP-lowering treatment and the risks of stroke and albuminuria were significant for both strata. Standardized estimates made excluding trials with imputed BP values were not substantively different (eFigure 21 in the Supplement). For nonstandardized estimates, there was heterogeneity for CHD and stroke (eFigure 22 in the Supplement).

Few differences were observed in the associations between BP-lowering treatment and outcomes for regimens based on different classes of medication used (Figure 5). The key exception was heart failure, in which diuretics were associated with a significantly lower RR (0.83 [95% CI, 0.72-0.95]) than all other forms of medication, driven largely by the results of ALLHAT. ARBs also appeared to be particularly associated with a lower RR of heart failure, although data were only available from 2 trials and the CIs were wide. By contrast, CCBs were associated with a higher RR of heart failure when compared with all other classes of medications. There was also some evidence that CCBs were associated with a lower risk of stroke (RR, 0.86 [95% CI, 0.77-0.97]), while β-blockers were associated with a higher risk (Figure 5). ARBs were associated with a lower risk of mortality relative to other classes of medications, an association was driven by the LIFE study, which compared an ARB against a β-blocker. Estimates were similar when 1 trial judged to be of high risk of bias and 3 trials judged to be of unclear risk of bias were included (eFigure 23 in the Supplement).

Quiz Ref IDIn this meta-analysis, BP-lowering treatment was associated with a significantly lower risk of all-cause mortality, CVD events, CHD events, and stroke events in patients with diabetes. Additionally, BP-lowering treatment was associated with a lower risk of retinopathy and albuminuria. Although the associations between BP-lowering treatment and risk of vascular outcomes were largely not significantly different across classes of medication, our results provide evidence of differential associations between classes for heart failure and stroke, as has been previously observed in the general population.¹³

Although we performed both standardized and nonstandardized analyses of the data, we elected to primarily focus on the standardized analyses, which are less influenced by any effects that may not be mediated through BP lowering. However, although there are some differences, the conclusions are broadly similar with and without standardization. Our analyses show evidence that BP-lowering treatment was associated with lower risks of outcomes in trials that included patients with an initial mean BP level of 140 mm Hg or greater, compared with patients with an initial mean BP level of less than 140 mm Hg, with the exception of stroke, albuminuria, and retinopathy. Similarly, when trials were stratified by achieved systolic BP, treatment was associated with lower risks only in the less than 130-mm Hg stratum for stroke and albuminuria in standardized analyses.

Quiz Ref IDThe failure to identify an association between BP-lowering treatment and lower risk of the cardiovascular composite with lower achieved BP levels is contrary to the findings of a large meta-analysis of 147 randomized trials performed in a broader population, which found evidence of a lower risk of CHD and stroke events at much lower achieved BP levels.¹³ A key difference between the previous meta-analysis and the current study is the inclusion of large numbers of patients with heart failure or recent MI in the prior overview, among whom mechanisms of action independent of BP lowering have been hypothesized. It is also likely that the inclusion of a number of recent trials of dual renin angiotensin system blockade (particularly the ALTITUDE study³¹) has attenuated the anticipated benefits of achieving lower BP levels. These results also contrast with a previous meta-analysis, which did not observe an association between BP lowering to a target of level of 130 mm Hg or less and risk of retinopathy or an association between BP lowering to any target (≤135 or ≤130 mm Hg) and risk of MI.⁸² Our differing results are likely due to the greater power of our meta-analysis (45 trials, 104 586 participants), relative to the previous review, which examined 13 trials enrolling 37 736 participants.

Our results lead to potentially different recommendations from those made in several recent guidelines (eTable 1 in the Supplement). For example, the recent JNC 8 (eighth Joint National Committee) guidelines relaxed the threshold for initiation of BP-lowering treatment from 130 mm Hg to 140 mm Hg in individuals with diabetes.⁸³ The ACCORD trial, which compared a target of lower than 120 mm Hg to lower than 140 mm Hg, was cited in support of this decision, as no significant reduction in the composite outcome of cardiovascular death, nonfatal MI, and nonfatal stroke was observed.⁵⁹ Although we also observed that BP lowering was not associated with a lower risk of CVD or CHD events at a baseline systolic BP of lower than 140 mm Hg, we did observe lower risks of stroke, retinopathy, and progression of albuminuria. Thus, in contrast with the recommendations of the JNC 8 guidelines, for individuals at high risk of these outcomes (eg, individuals with a history of cerebrovascular disease or individuals with mild nonproliferative diabetic retinopathy), the commencement of BP-lowering therapy below an initial systolic BP level of 140 mm Hg and treatment to a systolic BP level below 130 mm Hg should be considered.

When deciding whether to commence BP-lowering treatment in individuals with systolic BP below 140 mm Hg or whether to aim for systolic BP levels below 130 mm Hg, the risk of adverse events associated with BP-lowering treatments needs to be considered. Quiz Ref IDThis requires individualized assessment of the likely absolute benefits and risks with shared decision making between patients and clinicians. Although many trials reported select adverse events, these data were too disparate to allow formal meta-analysis. In ACCORD, the rate of serious adverse events attributed to BP lowering in the intensive treatment group (which reached an achieved BP of 119 mm Hg) was 2.5 times that of the control group (an achieved BP of 133 mm Hg), however, the absolute rate of these adverse events in the intensive group was low, and was substantially lower than that of the primary outcome (0.70% per year vs 1.87% per year).⁵⁹ It is possible that there is a higher risk of some important adverse events with large reductions at lower BP levels. Additionally, because many trial participants will have been treated with multiple classes of BP-lowering medications, differences between classes, with regard to efficacy and adverse events, may have been obscured. An individual patient data meta-analysis of efficacy and adverse events, stratified by patient characteristics, baseline BP, and class of medication, is necessary to provide the reliable evidence required to fully evaluate the overall balance of risks and benefits.

With data from 104 586 patients enrolled in 45 large trials, these analyses included substantially more information than previous published meta-analyses addressing this question (eTable 3 in the Supplement). This was, in large part, due to our efforts to identify all relevant data, by including not just trials performed in patients with diabetes, but also the results reported for the diabetic subgroups in large trials of mixed populations. For the associations with treatment based on initial BP level and target BP levels, we base our interpretation of the data primarily on the presence or absence of statistical heterogeneity across subgroups rather than the significance of the results in each individual subgroup. This approach minimizes the effect of the wide CIs obtained for each individual subgroup consequent on division of the data, and provides for a more robust interpretation. The uncertainty that ensues from the subgroup analyses emphasizes the need for more clinical trial data, particularly in relation to systolic BP targets of less than 130 mm Hg. Further trials that evaluate BP-lowering treatment into the 120- to 130-mm Hg range among hypertensive and nonhypertensive diabetic individuals would clarify whether lowering systolic BP to a target of less than 130/80 mm Hg would further reduce vascular risk relative to a target of less than 140/90 mm Hg, because the reliability of this meta-analysis is limited by the scarcity of large trials with achieved BP levels in the 120- to 130-mm Hg range. Additionally, the relatively short follow-up of included trials may also have prevented associations of BP-lowering treatment with vascular outcomes from being observed, particularly for outcomes such as heart failure and renal failure, which are often a consequence of MI and albuminuria, respectively.

Among patients with type 2 diabetes, BP lowering was associated with improved mortality and other clinical outcomes. These findings support the use of medications for BP lowering in these patients. Although proportional associations of BP-lowering treatment for most outcomes studied were attenuated below a systolic BP level of 140 mm Hg, data indicate that further reduction below 130 mm Hg is associated with a lower risk of stroke, retinopathy, and albuminuria, potentially leading to net benefits for many individuals at high risk for those outcomes.

Corresponding Author: Kazem Rahimi, DM, MSc, The George Institute for Global Health, Oxford Martin School, University of Oxford, 34 Broad St, Oxford OX1 3DB, United Kingdom (kazem.rahimi@georgeinstitute.ox.ac.uk).

Author Contributions: Drs Emdin and Rahimi had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Emdin, Rahimi, Neal, Patel.

Acquisition, analysis, or interpretation of data: Emdin, Rahimi, Neal, Callender, Perkovic, Patel.

Drafting of the manuscript: Emdin, Rahimi.

Critical revision of the manuscript for important intellectual content: Emdin, Rahimi, Neal, Callender, Perkovic, Patel.

Statistical analysis: Emdin.

Obtained funding: Rahimi.

Administrative, technical, or material support: Neal, Callender.

Study supervision: Rahimi, Neal, Patel.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Rahimi reports receipt of grants from the National Institute for Health Research (NIHR) and the Oxford Martin School during the conduct of this study. Drs Neal and Perkovic report receipt of payments to the The George Institute for Global Health from several corporations for research and for speaking about diabetes, blood pressure and cardiovascular disease. Dr Perkovic reports receipt of grants and personal fees from Abbvie, personal fees and other from Boehringer Ingleheim, other from Astellas, grants and other from Janssen, other from Novartis, GlaxoSmithKline, personal fees from Servier, AstraZeneca, Roche, and Merck, grants and personal fees from the National Health and Medical Research Council (NHMRC), grants from Baxter, and grants and nonfinancial support from Pfizer outside the submitted work. Dr Patel reports receipt of payment to her institution for her participation in an industry advisory board, and funding to her institution from Servier Laboratories and the NHMRC to conduct (as study director) the ADVANCE trial, which focused on blood pressure lowering in people with diabetes. No other disclosures were reported.

Funding/Support: The George Institute for Global Health has received grants from several pharmaceutical companies to conduct clinical trials in patients with diabetes. Dr Neal is supported by an Australian Research Council Future Fellowship. Drs Neal, Perkovic, and Patel are supported by NHMRC of Australia Senior Research Fellowships and hold an NHMRC program grant. Dr Rahimi is supported by the NIHR Oxford Biomedical Research Centre and NIHR Career Development Fellowship. Mr Emdin is supported by the Rhodes Trust. The work of the George Institute is supported by the Oxford Martin School.

Role of the Sponsors: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank Jun Cao, MRes, Imperial College London, for sorting abstracts in duplicate, and Tatjana Petrinic, MSc, University of Oxford, for helping to design the search strategy. Neither was compensated for their contributions.