Lowering blood pressure for cardiovascular risk reduction: moving the goalposts or spreading the net?

The concept of hypertension as a clinical diagnosis is under threat. Its definition is arbitrary. There is no diagnostic test other than blood pressure measurement. A multiplicity of underlying conditions (e.g. phaeochromocytoma, adenomatous primary aldosteronism and a range of renal disorders) exhibit elevated blood pressure but only as one feature of the disorder. Elevated blood pressure is a risk factor, but not a specific cause, for a range of vascular diseases. The genetic basis of most cases of hypertension is unknown; there is no single underlying mechanism of disease and no specific drug treatment. Hypertension therefore does not fit the disease definition that we associate with conditions such as cystic fibrosis, hepatitis C or lung cancer. Clinicians, patients, health authorities, actuaries and epidemiologists previously may have been comfortable with a traditional paradigm in which patients could be categorized as with or without hypertension; for example, witness the countless publications in which patients with ‘uncomplicated essential hypertension’ are compared or contrasted with ‘normal controls'. However, a rather different perspective is now emerging. Following the Pickering–Platt debates in the 1950s, it has become clear that blood pressure in the community exhibits a near normal distribution rather than a bimodal distribution. Blood pressure, not an arbitrary diagnosis of hypertension, is the key risk factor, and the vascular risks associated with blood pressure exhibit a continuous association with risk without a clear threshold. Law and Wald [1] have recently developed this approach by making an analogy with dose-dependent drug response relationships and demonstrating that blood pressure and serum cholesterol conform to a log-linear relationship in which an absolute change in the risk factor results in a proportional change in outcome. There is no discernible threshold, at least to systolic blood pressure as low as 110 mmHg. There is some uncertainty about whether the relationship with diastolic blood pressure and risk is entirely linear or whether there may be a J-shaped element [2], but this remains beyond the scope of the present discussion. A mass of evidence from prospective controlled trials attests to the value of blood pressure-lowering therapy in preventing all the complications of hypertension. Much discussion has centred on the possible differential efficacy of antihypertensive drugs in preventing stroke as opposed to myocardial infarction [3]. Even if it is accepted that there may be a relative shortfall in the effects of blood pressure lowering in preventing myocardial infarction, this does not detract from the overall beneficial effects of treatment. Treatment is effective in accelerated [4], severe [5], moderate [6] and mild [7] hypertension (accepting that these categories are arbitrary), in the elderly [8], in isolated systolic hypertension [9,10], in patients with and without underlying vascular disease and in all races. As a general rule, the benefits of treatment are proportional to the underlying risk. Accordingly, the emphasis has moved subtly from attempting to establish a diagnosis of hypertension to quantifying the risk associated with a particular level of blood pressure, and attempting to identify and deal with the modifiable risk factors [11]. This principle is incorporated in the current guidelines on hypertension [12–15]. However, along with advice to consider risk rather than blood pressure, these guidelines advocate a system of allocating patients to categories by level of blood pressure (Table 1).Table 1: Patient allocation to categories by level of blood pressureWhereas the risk assessment approach is gaining increasing acceptance, some doubts are being expressed about the appropriateness of this categorical approach. One obvious weakness is that patients do not remain in the same category, but progress to higher levels of blood pressure. Recent data from the Framingham study [16] show that 37% of patients with high normal blood pressure aged below 65 years and one-half of those over 65 years progressed to hypertension over a period of 4 years, with weight gain making an important contribution to this. In this issue of the journal, Yeo and Yeo [17] address the question of whether we should treat high-normal pressure. Their analysis used data from the 1998 Health Survey for England relating to 12 341 subjects with nurse-measured blood pressures. The Framingham equation [18] was used to estimate 10-year cardiovascular risk in the categories of mild hypertension, high-normal pressure and normal pressure. For the purposes of the analysis, a 10-year cardiovascular risk of 20% or greater was designated as the level that would justify drug treatment. The authors make several key observations. First, applying the Framingham equation to this English population resulted in the prediction of cardiovascular event rates that were close to the actual event rates observed in the Framingham study, allowing for the age difference of the populations. Second, no men or women aged 18–34 years had a predicted cardiovascular risk in excess of 20% over 10 years. This would suggest that patients in this age range do not need to receive antihypertensive treatment unless their blood pressure exceeds 160/100 mmHg, when current guidelines recommend treatment irrespective of other risk factors. Third, in each blood pressure category, a large proportion of men aged 61–80 years were at high risk of cardiovascular disease. These would benefit from blood pressure lowering to reduce cardiovascular risk, irrespective of their blood pressure category. Fourth, the much touted ‘threshold’ of 140/90 mmHg is neither a sensitive (71%) nor specific (54%) predictor of cardiovascular risk, confirming that this figure has little more than arbitrary value as a guide to treatment. Finally, 22.5% of the English population aged 18–80 years are either on antihypertensive drug therapy or would qualify for such therapy according to guidelines of the British Hypertension Society. In addition, if patients with ‘high-normal’ blood pressure are treated on the basis of their estimated 10-year cardiovascular risk, the authors estimate that an additional 2.6% of the English population may become eligible for blood pressure-lowering treatment. That is the nub of their study, which should give much food for thought to all clinicians dealing with elevated blood pressure. The main problem with the study by Yeo and Yeo [17] is that while appearing to offer a radical approach, it also abides by the arbitrary definitions of blood pressure that it purports to dismantle. The logical extrapolation of the data presented is to eschew the terms ‘hypertension', ‘mild hypertension', ‘high normal blood pressure’ and ‘normotension’ and to consider blood pressure not as a discrete but as a continuous variable. Therapeutic purists will demand evidence from controlled trials, although it has been argued that this is no longer necessary [1]. Some evidence from published trials would indeed support the proposal that blood pressure should be reduced in high-risk patients irrespective of initial blood pressures. The PROGRESS trial [19] was specifically designed to test this hypothesis in patients with preceding stroke or transient ischaemic attack and conclusively demonstrated the beneficial effects of active treatment (perindopril/indapamide) for the prevention of stroke and coronary heart disease, both in patients who were labelled as ‘hypertensive’ and ‘normotensive’ at baseline. Similarly, in the HOPE study [20], substantial benefits were achieved by treating high-risk vascular patients with ramipril. Admittedly, some interesting questions have been raised as to whether the benefits in both of these studies were linked entirely to blood pressure lowering or to some other property of the active drugs. There is certainly room for continuing debate and further research on this topic and it will undoubtedly provide much academic mileage for years to come. Recent meta-analyses suggest little difference between blood pressure-lowering drug classes in respect of cardiovascular outcomes [21,22]. One of the criticisms levelled at the MRC mild hypertension trial [5] was that many of the patients included were not significantly hypertensive. Indeed, in the placebo group, approximately one-third of patients had diastolic blood pressure less than 90 mmHg at each annual measurement, while 70% had diastolic blood pressure less than 90 mmHg at one of the first three annual visits. In spite of this, the active treatments prevented stroke. The number needed to treat (NNT) was relatively high (just over 70 patients to prevent one cardiovascular event in 10 years). By contrast, in patients whose 10-year cardiovascular risk exceeded 20%, the NNT over 10 years for a modest 10/5 mmHg reduction in blood pressure is 25 or less [15]. The reason for the relatively high NNT in the MRC mild hypertension trial is not so much that the study included patients with high-normal blood pressure, but almost certainly that it included patients at relatively low cardiovascular risk. This also explains why it required such large numbers (approximately 90 000 patient years) to demonstrate a modest, but by no means negligible, effect. This is probably the nearest we will come to a very large prospective trial in patients with ‘high-normal’ blood pressure. An almost exact parallel for the strategy of blood pressure lowering as opposed to treating hypertension now exists in respect of lipid lowering with statins. As little as 10 years ago, it was considered acceptable to treat hyperlipidaemia with these drugs. For certain patients with inherited dyslipidaemias, that may still be appropriate, but the goal posts have moved dramatically. First, it was established that lowering cholesterol was beneficial as secondary prevention in patients with previous coronary disease and elevated cholesterol [23]; subsequently, it was shown that statins were useful as primary prevention in men with high cholesterol at increased risk [24]. Further developments saw the value of statin therapy extended to patients with established coronary heart disease but with a broader range of cholesterol levels, including those considered to be ‘average’ [25,26]. The recently reported MRC Heart Protection Study [27] was designed on the premise that the relationship between serum cholesterol and cardiovascular risk was continuous and that there was no threshold below which cholesterol lowering should not be beneficial. This study has convincingly demonstrated, in patients at increased risk either because of preceding disease or because of underlying risk factors, that lowering cholesterol is safe and beneficial in patients with a serum cholesterol down to 3.5 mmol/l. Thus, it can be confidently recommended that adults in the UK are offered statin therapy on the basis of risk rather than cholesterol level, and with a view to lowering cholesterol rather than treating an arbitrarily defined hyperlipidaemia. What the study by Yeo and Yeo [17] is proposing is a broadly similar strategy for blood pressure lowering. Whether clinicians will be persuaded on the basis of their analysis and without a specific prospective outcome study is difficult to judge at this stage. In the UK, prescribers are encouraged to adopt an evidence-based approach to medical treatment, with an explicit recognition that the strongest levels of evidence are meta-analyses and randomized trials. The analysis by Yeo and Yeo [17] falls some way short of those guidelines and, even when supported by the recent paper from Law and Wald [1], would be classified as evidence level III with a possible grade B recommendation [28]. It may take some time to catch on. It must also be recognized that a relatively small proportion (2.6%) of the population translates into a substantial total number of patients, with all its attendant implications. A problem for one sector – the Health Services – can, of course, be seen an opportunity for another – the Pharmaceutical Industry. The case to date is largely epidemiological, but several major studies in progress (ADVANCE, ONTARGET) will add to the prospective evidence base in due course. ONTARGET (http://www.ontarget.cardio.on.ca) will study a similar population to that studied in HOPE and will examine the effects of ramipril and/or telmisartan. ADVANCE (http://www.ctru.auckland.ac.nz/advance/) will examine not only the effects of blood pressure lowering in type 2 diabetes, but also the effects of blood glucose lowering. The principle can be applied to a number of continuous numeric risk factors and is not specific for blood pressure. All this may make good sense, both epidemiologically and therapeutically, but many familiar problems will remain. How accurate is the measurement of blood pressure? How can we identify the patient at greatest risk? How can we ensure that patients identified are optimally treated? Which drugs shall we use? What is the clinical effectiveness of the risk-targeting strategy? Can we afford it? ‘Plus ça change, plus c'est la meme chose!'