β-blocker-based therapy has been a routine treatment for hypertension for many years. More recently, the effectiveness of β-blocker-based treatment for hypertension has been questioned [1-4]. In this issue of the journal, Blackburn et al. [5] use a retrospective, observational cohort study to evaluate the use of atenolol compared to other blood pressure (BP)-lowering drug classes and have concluded that they can detect no signal of harm associated with atenolol relative to other drugs when used for the routine treatment of uncomplicated hypertension. Before considering this information, it is worth reflecting on the source of original concerns about β-blocker use for the treatment of hypertension.

Evidence from recent clinical outcome trials demonstrates that, compared to other classes of BP-lowering treatments, β-blockade (atenolol) was no more effective at preventing major cardiovascular events, including myocardial infarction [1-4,6,7]. Moreover, β-blockers are significantly less effective at preventing stroke and increase the risk of developing diabetes compared to angiotensin-converting enzyme inhibitor, angiotensin receptor blocker or calcium channel blocker (CCB)-based treatment, especially when combined with thiazide-type diuretics [1-4,6-8]. These concerns were consolidated when the UK-based National Institute for health and Clinical Excellence (NICE), working in collaboration with the British Hypertension Society (BHS), recommended in its most recent review of the treatment of hypertension, that β-blockers should no longer be used as a preferred routine initial therapy for individuals with hypertension [9]. This rigorous guideline review process had considered evidence from all major clinical outcome trials for the treatment of hypertension and used meta-analysis to evaluate the effectiveness of drugs at reducing major cardiovascular outcomes.  . From this analysis, it was clear that, in hypertension trials, β-blocker-based therapy was no more effective than other classes of BP-lowering treatment at preventing myocardial infarction, unstable angina, heart failure or death, and less effective at preventing stroke. The NICE/BHS guideline also undertook a formal cost-effectiveness analysis [9]. It is important to emphasize that this differs from the approach adopted by other guidelines, which have simplistically concluded that the cheapest drug is the most cost-effective. Cost-effectiveness analyses first consider the relative effectiveness of different drugs on clinical outcomes and then consider costs. The costs are more than those associated with the acquisition of drugs. Costs also include the cost benefit of reducing clinical events, hospitalization and death, and the cost of harm associated with treatment (e.g. the development of diabetes). This cost-effectiveness analysis was damning for β-blockers. Put simply, when used for treating hypertension, β-blocker-based therapy was; (i) the least effective treatment at preventing major cardiovascular events, especially stroke; (ii) more likely to increase the risk of developing diabetes (especially when combined with a thiazide); and (iii) was dominated in the cost-effectiveness analysis by all other treatment options and was the least cost-effective treatment option  [9].

Importantly, this does not mean that β-blocker-based therapy is an ineffective treatment for hypertension. It simply means that we have better and more cost-effective treatment options available. Nevertheless, there are patients who will require the benefits of β-blocker-based therapy beyond their impact on BP, notably those with symptomatic angina, post-myocardial infarction and chronic stable heart failure, and the NICE/BHS guideline recommends that such patients should continue to receive this medication. β-blocker-based therapy is also a treatment option for patients who have tried other drugs but cannot tolerate them, or those who require additional drug therapy for resistant hypertension, or as an option for women of child-bearing potential in whom other drugs may be less desirable.

There has been concern expressed about whether the downgrading of β-blockers should be applied to all β-blockers, or whether it should only apply to atenolol, which has been the drug most often tested in outcome trials in hypertension. This can be addressed in two ways. First, in my view, the guidelines should whereever possible base their recommendations on evidence from outcome trials and, because the other β-blockers have not been tested in this arena, it would be wrong to assume that the other β-blockers would have performed better than atenolol. Although there may be theoretical reasons why other β-blockers with different pharmacological properties may be more effective than atenolol in individuals with hypertension, it is surely necessary to show that such alternatives are (i) better than atenolol and (ii) at least as effective as other classes of BP-lowering drugs before they can be recommended as a routine initial treatment for hypertension.

So how to reconcile the findings of Blackburn et al. [5] with those of major clinical outcome trials? First and foremost, it cannot be over-emphasized that retrospective case-controlled studies are no substitute for rigorous large-scale outcome trials and should never be the basis for treatment recommendations: these studies are not randomized. The choice of therapy by doctors is not usually randomly allocated in clinical practice and there is always considerable potential for confounding by indication. Although Blackburn et al. [5] point out that they have taken all reasonable steps to adjust for potential confounding, there is still evidence that it exists (e.g. the higher nitrate use and lower inhaled corticosteroid use in those receiving atenolol). No amount of statistical adjustment can adequately compensate for such confounding, and never will. Indeed, given the number of variables adjusted for and the low rate of the primary outcome in the study by Blackburn et al. [5], there is also the danger of 'overfitting' the modelling, thereby diminishing its ability to show real differences. Blackburn et al. [5] address this by stating that the proportional hazards assumption was not violated. In addition, the study is of relatively short duration and has few events, thereby undermining its potential to show real differences between drugs. Moreover, there are no data provided or available with regard to BP control between treatment groups.

Let us not forget the concerns generated by similar case-controlled analyses, which demonized CCBs at the turn of this century [10-12]. According to these analyses/reviews, calcium channel blockers were reported to be less effective at preventing myocardial infarction, and increase the risk of gastrointestinal bleeding and cancer! Subsequent large-scale randomized clinical outcome trials firmly debunked these concerns [7,13,14]. So comprehensive was the turn-around that the NICE/BHS review of the most recent data identified calcium channel blockers as the most cost-effective treatment option for hypertension [9].

Blackburn et al. [5], referring to their retrospective case-controlled study, conclude; 'the low event rates for all cohorts suggest atenolol has not been associated with a significant burden of cardiovascular morbidity or mortality in its traditional role for uncomplicated hypertension. Further study is needed to identify the specific types of patients that should avoid atenolol as an antihypertensive agent' [5]. This conclusion is contentious for at least two reasons. First, atenolol has never been tested in outcome trials of uncomplicated hypertensive patients and, where it has been tested in hypertensive patients, it was the least cost-effective treatment option [9]. Second, why should we place a burden on practitioners to try and identify patients in whom β-blockers should be avoided when we have a range of more cost-effective and more metabolically friendly treatment options available with which to initiate therapy in all patients? In my view, it was bold and appropriate to relegate β-blockade from its position as a preferred initial therapy option for the routine treatment of hypertension [15]. Many of the myths about the putative benefits of β-blockade for the hypertensive patient (e.g. better primary prevention of myocardial infarction, better prevention of heart failure, better prevention of atrial fibrillation) have been blown away by recent outcome trials [6,7]. Unfortunately, it has been more difficult to remove the physicians' comfort blanket founded on the notion that β-blockade will 'protect and stabilise the heart' in hypertensive patients. The case study of the aforementioned calcium channel blocker debacle illustrates that, although retrospective case-controlled studies can be used to generate debate, they must never be used to generate clinical practice guidelines for the treatment of patients: the randomized clinical trial remains the standard of proof.