Simple‐Minded Antihypertensive Treatment: Of Assumptions, Potpourri, and Sausages
Year: 2009
Type: article
Abstract: For every complex problem there is a solution that is clear, simple … and wrong. —H. L. Mencken. In their provocative article published in the British Medical Journal (BMJ), Law and colleagues1 state: "Our results indicate the importance of lowering blood pressure (BP) in everyone over a certain age, rather than measuring it in everyone and treating it in some." The authors come to this simple conclusion from data derived from a potpourri of trials—147 randomized controlled trials enrolling patients with varying inclusion and exclusion criteria as young as 34 years to as old as 84 years, followed for <0.5 years to 8.4 years. The spectrum of patients in these trials varied from those with hypertension, myocardial infarction (MI), coronary artery disease without MI, and heart failure, to those treated with a variety of drugs, from placebo to different classes of antihypertensive agents, at varying doses and frequency, with different effects on BP, and with disparate definitions of end point, but all still resulting in one clear simple conclusion: BP measurement is useless! The authors go on to conclude that everyone beyond a certain age should be treated with an antihypertensive agent regardless of their baseline BP, shaking up the very foundation of risk stratification on which most of the principles of cardiovascular (CV) medicine is based on. Law and colleagues conclude by stating that their results indicate the importance of lowering BP in everyone over a certain age rather than measuring it in everyone and treating it in some (risk stratifying). Their conclusion is based on certain assumptions, derivations (from their analyses), and extrapolations: the 5 major drug classes all produced similar reduction in BP when taken at standard dose; for every 10-mm Hg systolic or 5-mm Hg diastolic reduction of pressure, the risk of coronary heart disease (CHD) decreased by at least 22% and stroke by at least 41%; all classes of BP-lowering drugs have a similar effect in reducing CHD events and stroke for a given reduction in BP, thereby excluding pleiotropic effects; and that the proportional reduction in risk was the same regardless of pretreatment BP and the presence or absence of existing CV disease. They, therefore, advocate a combination of three drugs at half standard dose for every one over certain age, regardless of their BP or other risk factors. One is left to wonder what the implications of such analyses and the conclusions would be for the practicing physician. A cursory look at Figure 2 in the article lends credence to the authors' conclusions that antihypertensive agents are efficacious at a wide range of BPs (systolic, 110 mm Hg to >170 mm Hg; diastolic, 70 mm Hg to >95 mm Hg). A closer look at just this analysis suggests a different story. While patients with higher BP (>140 mm Hg) were from hypertension trials, the trials with lower BP enrolled a completely different subgroup of patients. In the only two trials enrolling patients with pretreatment systolic pressures of 110 mm Hg to 119 mm Hg, both (the Survival and Ventricular Enlargement Trial [SAVE]2 and Prospective Randomized Amlodipine Survival Evaluation [PRAISE]3) were in patients with heart failure. Extrapolation of these results to healthy persons with systolic pressures of 110 mm Hg to 119 mm Hg and advocating antihypertensive agents for this cohort is inappropriate and potentially dangerous. The pathophysiology of hypertension is clearly different from that of heart failure, where neurohormonal activation leads to ventricular remodeling and consequent adverse sequelae. "Antihypertensive agents" in patients with heart failure block the neurohormonal activation, thereby causing reverse remodeling and consequently reduce the risk of future CV events. Clearly, this is not a BP phenomenon! In patients with depressed ejection fraction, neurohormonal blockade and consequent reverse remodeling, and improvement in left ventricular function can, in some cases, paradoxically lead to an increase in BP. In the Metoprolol CR/XL Randomized Intervention Trial in Congestive Heart Failure (MERIT-HF),4 systolic BP (baseline, 130 mm Hg) decreased less in the metoprolol CR/XL group than in the placebo group (−2.1 mm Hg vs 3.5 mm Hg; P=.013). The worse-than-placebo antihypertensive efficacy of metoprolol notwithstanding, the authors seemed to believe that a mere 2.1-mm Hg systolic BP decrease by metoprolol resulted in a stunning 44% relative reduction in the risk of all-cause mortality at just 1 year of treatment, necessitating early stoppage of the trial. The relative risk reduction with metoprolol therefore remains unexplained by difference in BP between the two arms. Clearly, the efficacy of β-blockers in patients with systolic heart failure is mostly a BP-independent effect. Law and colleagues estimate that for a 10-mm Hg systolic or 5-mm Hg diastolic reduction in pressure, the risk of CHD reduces by 25% and stroke by 36%. Based on these statistics, a systolic pressure reduction of at least 15 mm Hg would be required to produce the above 44% reduction in all-cause mortality seen in the MERIT-HF trial. Law and colleagues may argue that the results are due to special effects of β-blockers in patients with a history of MI (within the first 2–3 years postinfarction only). In the MERIT-HF trial, only 48% of patients had a history of MI, half of which were more than 5 years prior to study enrollment. The special effects of β-blockers over and above that due to BP can only be attributed to <20% of patients enrolled in the MERIT-HF trial, and therefore cannot explain the impossible benefits of metoprolol in this trial. Law and colleagues wishfully assume that all antihypertensive drugs reduce BP to the same extent at a standard dose. This data is derived from their own meta-analysis of 354 randomized double-blind trials published in 2003.5 Of note, this analysis excluded trials recruiting patients with acute MI, heart failure, or other CV disease. They, however, extrapolate this result (of randomized controlled trials excluding MI, heart failure, or CV disease) to their recent meta-analysis wherein they included MI, heart failure, and other CV disease trials, excluded in the prior analyses. This assumption is proven wrong by the simple fact that amlodipine in the PRAISE trial3 of patients with heart failure lowered systolic pressure by 2 mm Hg, whereas in patients with hypertension, there was a 6-fold greater decrease in BP (12/6 mm Hg).6 Similarly, metoprolol CR (up to 200 mg/d) in the MERIT-HF trial4 (systolic heart failure patients) reduced systolic BP by a mere 2.1 mm Hg, whereas a comparable dose in the African American Study of Kidney Disease (AASK) trial7 (hypertensive renal disease patients) reduced it by 15 mm Hg. The dictum of Law and colleagues that all antihypertensive agents reduce BP to the same extent at a standard dose, while possibly true in patients with hypertension, will not hold true across a spectrum of disease states. The reduction in BP depends not only on the medication and its dosage but also on underlying condition for which it is used. Clearly, the notion that one size fits all is erroneous. In making a generalized conclusion about prescribing antihypertensive medications to everyone regardless of their BP, the authors seem to purposefully ignore the adverse effects of these medications. Even in their own analyses, 25% of patients, ie, 1 of every 4 patients enrolled, stopped study medications for various reasons.1 Most likely, the majority of these withdrawals were due to adverse effects of medications. Considering that this withdrawal rate was derived from randomized trials, where every possible effort is made to ensure patient compliance, it can only be assumed that this withdrawal rate would be much higher if we were to treat the general population (with the present definition of a normal range of BP) with antihypertensive agents. The Medical Research Council (MRC) study allows us to calculate that for every MI or stroke prevented, 3 patients treated with atenolol withdrew from the study secondary to impotence and another 7 withdrew because of fatigue.8 Moreover, in a "network meta-analysis" of 22 clinical trials9 with 143,153 participants who did not have diabetes at randomization, the risk of new-onset diabetes was most pronounced with diuretics and β-blockers, more so than with placebo or other classes of antihypertensive agents, implying a negative metabolic effect of these medications. Treating a healthy 60+-year-old with a systolic of 110 mm Hg with antihypertensive agents that are known to increase the risk of new-onset diabetes, such as β-blockers and diuretics, hardly seems justifiable. We have previously shown that in an analysis of 30,842 patients with hypertension from randomized trials, thiazide diuretics increased risk of new-onset diabetes by 32% when compared with placebo or non–β-blocker antihypertensive agents.10 In an analysis of 55,675 patients with hypertension from randomized trials, β-blockers also conferred a 32% increased risk of new-onset diabetes when compared with placebo or nondiuretic antihypertensive agents.10 The results of our analyses allow us to calculate that treatment of 1000 hypertensive patients with β-blockers for 4.4 years will result in 14 excess cases of diabetes mellitus.11 Given that 65 million Americans have hypertension, this would account for 910,000 new cases of diabetes over the same period.11 The global consequences of treating asymptomatic normotensive patients with metabolically averse medications (such as diuretics and β-blockers) can be devastating. The prospective study collaboration analysis12 has been used, overused, and abused to demonstrate the benefit of antihypertensive therapy. Not surprisingly, Law and colleagues also used this study to estimate reduction of events for a given fall in BP. In doing so, they seemed to intentionally ignore (as did many pharmaceutical companies before them) the fact that this study does not deal with antihypertensive therapy or its evaluation. The prospective study collaboration clearly shows that usual BP was strongly and directly related to vascular and overall mortality, ie, that a 10-mm Hg lower usual systolic BP would, in the long-term, be associated with about a 40% lowering risk of stroke or death or about a 30% lower risk of death from CHD. However, this study did not show that reducing systolic BP by 10 mm Hg confers similar benefit. The authors then go on to suggest that all classes of BP-lowering drugs based on the above (wrong) assumptions have a similar effect in reducing CHD events and stroke for a given reduction in BP. Perhaps this may apply to some drugs, which were evaluated in their meta-analysis. However, short-acting calcium channel blockers such as sublingual nifedipine are powerful antihypertensive agents but are known to increase the risk of MI and stroke.13 Recent well-conducted large randomized trials seem to suggest that it is not only lowering BP that is important, but how you lower it as well. In the Avoiding Cardiovascular Events in Combination Therapy in Patients Living With Systolic Hypertension (ACCOMPLISH) trial,14 for similar reduction in BP, a benazepril-amlodipine combination was superior to benazepril-hydrochlorothiazide for reduction of CV outcomes. In the Losartan Intervention For Endpoint Reduction (LIFE) study, BP was lowered to the same extent by atenolol-based therapy and losartan-based therapy, yet there was a highly significant difference in the primary outcome favoring the losartan arm. Indeed, data from multiple meta-analyses performed in patients with hypertension suggest the inferiority of β-blockers for CV risk reduction in hypertension,15-18 resulting in withdrawal of endorsement for these medications as first-line therapy for hypertension by major national and international guidelines.19, 20 Even in the authors' own estimate (online Web Figure 3: http://www.wolfson.qmul.ac.uk/epm/publications/web/bptrial/figs/figure3.pdf) in drug comparison trials (again a potpourri of disease conditions), β-blockers were not efficacious at preventing CHD events (relative risk, 0.89; 95% confidence interval, 0.78–1.02) and resulted in a 18% increase in the risk of stroke compared with other antihypertensive agents (online Web Figure 4). Law and colleagues suggest that similar to that for low-density lipoprotein cholesterol, lower is better for BP. The linear theory may be true for the usual BP in untreated individuals, such as that shown by the prospective collaboration analysis12 however, this linear theory has been challenged for nearly 3 decades (especially for diastolic pressure) in patients with preexisting coronary artery disease.21-24 In patients with stable coronary artery disease, the relationship between BP and CV outcomes has been shown, in some studies, to follow a J- or U-shaped curve, with higher event rates at very low and very high BP.21, 23, 25-30 Physiologically, a J curve would be expected to exist as extremely low or high pressures are incompatible with life. While some may argue that such extreme low pressure is not possible due to the body's own reflex mechanisms, this may be true in the untreated patient. Antihypertensive therapy certainly can lower BP inappropriately to life-threatening levels, as has been documented with sublingual nifedipine. While the concept of J curve is controversial, global treatment of BP regardless of pretreatment levels should take into consideration the remote possibility of a J curve (in patients with preexisting CAD) and, hence, increase in mortality in this cohort. The editor of BMJ, Fiona Godlee, obviously impressed by the "enormous amount of work" performed by Law and colleagues, stated in the accompanying editorial that the authors looked at data from 147 trials of antihypertensive treatment published between 1966 and 2007.31 Unfortunately, many of these trials had nothing to do with hypertension but rather dealt with patients with MI or heart failure. One is left to wonder whether the degree of BP reduction and CHD protection, derived from a potpourri of trials with different assumptions and extrapolation, is simply a result of regression toward the mean. Of note, two of the authors (Law and Wald) proposed the "polypill" and have filed for a patent on this therapeutic approach, and, therefore, their research may not be entirely free of bias. As Fiona Godlee emphasized, "In their new paper, they find some indirect support for the polypill concept." Clearly the study of Law and colleagues is a prime example of our dictum: A meta-analysis is like a sausage, only God and the butcher knows what goes in it, neither would ever eat any.
Source: Journal of Clinical Hypertension
Authors: Sripal Bangalore, Franz H. Messerli
Language: English
Subfield: Cardiology and Cardiovascular Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W2000125396