Meta‐analyses and blood pressure goals
Editorial Staffan Björck, Karin Manhem, Annika Rosengren, Samuel Adamsson, Soffia Gudbjörnsdottrir, Stefan Franzén
published June 26 2018
Läs i full text hela artikeln utan lösenord, alla tabeller, figurer och referencer
In summary without tables and figures and references
The new 2017 blood pressure guidelines from the American Heart Association, the American College of Cardiology (ACC/AHA), and 9 other health organizations recommend that patients with an elevated cardiovascular risk should be treated with antihypertensive medications if blood pressure is higher than 130/80 mm Hg.
Patients with diabetes or kidney disease are automatically placed in this higher risk category.
Contrary to this, the still valid 2013 European Society of Hypertension and European Society of Cardiology guidelines for hypertension raised the recommendation for patients with diabetes to a systolic blood pressure (SBP) goal of <140 mm Hg from <130 mm Hg.
This decision was mainly based on the J‐shaped relationship between blood pressure and cardiovascular risk in observational studies, inconclusive randomized controlled studies, and the failure of the Action to Control Cardiovascular Risk in Diabetes study (ACCORD)3 to show a significant effect of intensive treatment on the main outcome.
The background material for the 2017 AHA/ACC guidelines that support lower blood pressure goals is specified in the systematic review data supplement that lists 15 randomized controlled trials and 9 meta‐analyses, that compare relative risk reductions for lower versus standard blood pressure targets. All but 1 of the 9 meta‐analyses are published between 2016‐2017. The only meta‐analysis specifically focusing on diabetes showed an increased cardiovascular mortality associated with intensive antihypertensive treatment. This alarming finding, that does not support the new guidelines is not commented upon in the systematic review. We have analyzed this meta‐analysis further.
Many articles have been published on methods and interpretations of meta‐analyses. They are sometimes described as the highest form of evidence, but they have also been called misleading. The quality of meta‐analyses in the field of hypertension have recently been thoroughly reviewed from a statistical perspective.
Overall, there is an exponential growth in the number of meta‐analyses with a current rate of publications estimated to be around 8 per month. It seems inevitable that the same underlying studies are analyzed over and over again.
Seven of the meta‐analyses used for the AHA/ACC guidelines are based on 4 to 19 original studies selected from an underlying set of 28 original publications, meaning they all largely overlap. One of the studies is a network meta‐analysis. No meta‐analysis is based on individual data.
Results are presented differently, but the conclusions are similar, indicating better cardiovascular outcome with intensive treatment but with limited support for the lowest blood pressure levels. The network meta‐analysis supports the concept “the lower, the better” for cardiovascular risk.
Two of the meta‐analyses differ from the others because they are based on many more studies. The largest meta‐analysis of 112 studies, contains a mixture of blood pressure lowering treatments across various baseline blood pressure levels and comorbidities, and include trials of anti‐hypertensive drugs even for indications other than hypertension. The other study is a meta‐analysis of 49 studies of patients with diabetes only, and a similar broad recruitment strategy that we will particularly focus on.
We have analyzed the main conclusion in the above mentioned large diabetes meta‐analysis, that intensive antihypertensive treatment is harmful, since it might have an impact on future guidelines and would like to highlight some issues that have implications for the assessment of meta‐analyses in general.
We have evaluated this meta‐analysis by looking at the included studies and what they tell us from a clinical perspective instead of analyzing statistical terms such as estimates, heterogeneity and confidence intervals.
This diabetes meta‐analysis presents risk estimates for complications associated with antihypertensive treatment in patients with high, intermediate or low baseline blood pressure and 6 different cardiovascular endpoints. The main conclusion is that if SBP is less than 140 mm Hg, further blood pressure lowering treatment is associated with an increased risk of cardiovascular death (relative risk 1.15, 95% confidence interval 1.00 to 1.32), with no further benefit.
The first problem with this type of meta‐analysis is that it is not based on results for patients who have specified initial blood pressure. Each study is assigned to a particular group regardless of the individual patient's blood pressure. The result is a large overlap between individual blood pressure levels in the various groups. For example, the Perindopril Substudy in Coronary Artery Disease and Diabetes (PERSUADE) study classified treatment as beneficial for all patients with blood pressure above 140 mm Hg even though almost half of them were below 140 mm Hg, simply because the average at baseline had been 140.1 mm Hg. Furthermore, the patients who received active treatment in this study averaged baseline blood pressure of 139.8 mm Hg. If the study for that reason had been classified as an examination of treatment for patients with low blood pressure (<140 mm Hg), the main conclusion would have faltered.
The second problem is that subgroup analyses in some of the largest studies in the meta‐analysis of the same SBP interval (140 to 150 mm Hg) contradict the main message in the meta‐analysis that treatment, when SBP is less than 140 mm Hg, might be harmful. These underlying studies have separately shown that risk reduction through antihypertensive treatment is independent of baseline SBP. The ADVANCE trial, the largest study that intentionally recruited patients with a broad range of baseline blood pressures, assessed the effects on vascular complications of a fixed combination of an ACE inhibitor and a diuretic in patients with type 2 diabetes. The authors could not identify any evidence that the effects of the study treatment differed according to baseline blood pressure. Two other large studies that were covered in the meta‐analysis, in the 140 to 150 mm Hg range, reported similar findings.These 3 studies, which account for two‐thirds of all subjects in the 140 to 150 mm Hg interval, all conclude that beneficial effects of antihypertensive treatment on the primary outcome are independent of baseline blood pressure. Furthermore, the same finding was noted by 1 study in the low blood pressure range. The design of the meta‐analysis obscures this descriptive information from individual studies. Risk ratios for these 3 studies are presented in Figure 2.
A third problem is the clinical heterogeneity of the studies that examined antihypertensive treatment when SBP was below 140 mm Hg. In these studies, antihypertensives are often prescribed for reasons other than lowering blood pressure. For example, the largest included study terminated dual blockade of the renin‐angiotensin system with aliskiren and an ACE inhibitor early because of a tendency to increased cardiovascular morbidity. The treatment did not lower blood pressure from baseline and showed a difference in SBP of only 1.3 mm Hg between subjects and controls. Our opinion is that this trial, representing 38% of included patients in the low blood pressure range, should not be included in the meta‐analysis since the results reflect safety data for dual blockade of the renin‐angiotensin system rather than any effect of antihypertensive treatment. Of the remaining 9 studies in the low blood pressure range, 2 other small ones should be excluded because they report no cardiovascular deaths.
The fourth problem is that the most important study of the lowest blood pressure range is the randomized ACCORD trial, the only 1 aimed at SBP below 120 mm Hg.3 In this trial, there were 60 cardiovascular deaths in the group receiving intensive antihypertensive treatment and 58 deaths among those receiving conventional treatment. Follow‐up data have been presented for almost 9 years, covering 84% of the 4733 patients in the study.13 During the extended follow‐up period, 121 cardiovascular deaths were reported in the intensive treatment group and 129 in the control group. A favorable effect was identified for the primary composite cardiovascular endpoint of intensive antihypertensive treatment among patients not assigned to tight glycemic control. No indication was found that combined cardiovascular events were also more frequent among patients with the lowest baseline blood pressure (Figure 2). Although the treatment protocol was scrapped, blood pressure in the intensive treatment group was 130 mm Hg which is lower than the control group, during the extended follow‐up period.
The ACCORD trial is unique for the meta‐analysis in that it is the only randomized trial that compared conventional and intensive antihypertensive treatment for baseline blood pressure below 140 mm Hg. It is also unique in that it demonstrated a major reduction (14 mm Hg) from a low level. Contrary to the main conclusion of the meta‐analysis, the ACCORD follow‐up trial alone shows that even aggressive antihypertensive treatment can be safe regarding cardiovascular mortality and may afford protection against cardiovascular morbidity.
From a statistical point of view, the meta‐analysis addresses the effects of lowering blood pressure; however, the primary goal of several studies in the lower range was not to further reduce blood pressure and only 1 study addresses this point. Since exchangeability is a fundamental assumption in a meta‐analysis, the model is inappropriate. Exchangeability requires that the expected effects of various studies are similar and that the observed effects differ only as a result of random variability. This was not the case.
The authors of the diabetes meta‐analysis claim that the results are immune to confounding because the analysis is based on randomized studies. The assertion would have been true if the studies had been randomized between blood pressure levels at either baseline or endpoint. The studies are, on the contrary, randomized between different types of interventions. Thus, confounding variables, including concomitant diseases associated with baseline and final blood pressure and the risk of death from other outcomes, could have arisen. As a result, this meta‐analysis has the same scientific value as an observational study of the association between blood pressure and mortality in which no effort has been made to adjust for confounding.
To address the question raised by Brunström et al, meta‐analysis would be better if based on individual or data, stratified according to relevant groups of specific blood pressure levels, and derived from suitable studies. One such study, based on 32 trials, including 201 566 patients for whom individual data or data stratified by initial blood pressure had been obtained,17 concluded that there was no evidence of differences in the risk reductions achieved with various antihypertensive regimens across groups defined according to higher or lower levels of baseline systolic blood pressure.
In conclusion, meta‐analyses may conceal important information from underlying studies that can best be understood descriptively. Information from individual major studies included in the meta‐analysis by Brunström et al suggests that relative risk reduction through antihypertensive treatment is independent of baseline blood pressure, and supports the new US guidelines by means of a more proactive approach to antihypertensive treatment for patients with elevated risk
CONFLICTS OF INTEREST
The authors report no conflicts of interest.
Läs i full text hela artikeln utan lösenord, alla tabeller, figurer och referencer
www red DiabetologNytt