This practice conflicts sharply with convincing evidence that the relationship between the height of the pressure and the incidence of stroke and heart attack is continuous [2, 3]. Popular practice, however, is driven by the diametrically opposite view that a specific threshold separates normotension from hypertension, distinguishing those who do from those who do not need treatment [4]. The notion that a cut-point value might define hypertension has been challenged from the beginning [5]. In the predrug era, the threshold was a frequently shifting number of only theoretic diagnostic convenience. With the appearance of effective oral antihypertensive drugs, however, the threshold concept became of immediate therapeutic significance.

The initial indication for hypotensive drugs was to treat malignant hypertension—a condition diagnosed not by a particular level of blood pressure but by specific clinical characteristics [6]. The idea that a particular blood pressure level could, by itself, guide therapeutic decisions became fashionable only later. It probably reflected eligibility criteria chosen for the Veterans Administration trial [7] rather than a well-considered analysis of the implications of its results. Since then, the National High Blood Pressure Education Program has generated consensus documents that have repeatedly recommended treatment decisions be based primarily on blood pressure recordings [4]. By contrast, more recently, the National Cholesterol Education Program attempted to avoid similarly unidimensional recommendations [8].

Present antihypertensive strategy could be improved substantially if treatment were better matched to actual risk for cardiovascular disease and the ability to prevent it. The evidence supports a two-pronged approach. First, the blood pressure of all persons should be lowered. A decrease of even a few points, if universally achieved, could substantially reduce cardiovascular morbidity and mortality rates [9]. In addition, treatment should be more aggressive for those relatively few persons—regardless of their blood pressure level—whose risk for disease is great enough to merit specific and potent drug therapy. The purpose of this commentary is to assess current clinical practice in the light of available experimental and epidemiologic data and to suggest modifications that will produce more effective and efficient individual therapeutic decisions.

Blood Pressure as a Risk Factor

A strong, consistent, graded, and independent association exists between blood pressure and the incidence of stroke and heart attack [10, 11]. Blood pressure is not, however, a disease. It is one of many biological, genetic, and behavioral characteristics whose presence and magnitude affects the relative risk for disease occurrence [12]. Neither blood pressure nor other risk factors are invariably present in those who have a cardiovascular event, nor are they always absent in those who do not. Despite increasing risk, even at the highest end of the blood pressure range, only some persons will have a stroke or heart attack. During 15 years of follow-up in the Framingham study, fewer than one third of those whose only risk factor was systolic blood pressure greater than 195 mm Hg (the highest category) had a stroke or heart attack [13]. Among the 325 348 MRFIT screenees, those with diastolic pressures of 90 mm Hg or greater were about twice as likely to have a heart attack or stroke as were those with diastolic pressures less than 90 mm Hg [14].

Similarly, a lower blood pressure does not confer total protection against a cardiovascular event. In fact, more than one half (57%) of all heart attacks and almost one half of all strokes in some population studies occur in persons with normal blood pressure [15, 16]. In short, blood pressure, like other risk factors, provides quantitative and not qualitative information. Although higher blood pressures are associated with more cardiovascular events, no threshold distinguishes those who will from those who will not have a heart attack or stroke.

Relative and Absolute Risk

Relative risk and absolute risk are different ways to express the likelihood of events for individuals or groups. The actual quantities are derived empirically from observation of defined populations such as in the Framingham Study. In practice, these measures are then extrapolated to predict risk for other individuals or groups with characteristics similar to the groups studied. Relative risk describes the increase (or decrease) in the likelihood of an event in one population compared with some other reference population. It is a ratio that imparts no information about the absolute expected incidence of events. Absolute risk, on the other hand, is a term used to describe the expected incidence of events. It is used to estimate the total number of events to be expected by a specific group or, by extension, the actual odds for a person to have an event.

For blood pressure, the relative risk for heart attack or stroke increases continuously with rising systolic and diastolic blood pressure. Specifically, Figure 1 shows that each 7.5 mm Hg increase in diastolic blood pressure is associated with 46% more strokes (relative risk, 1.46) [17]. This number gives no indication of the actual number of persons in any particular group who will have a stroke. Instead, it expresses the relative expectation of an event in persons in two groups that differ only in level of blood pressure. It is worth noting that most myocardial infarctions occur among patients with "normal" diastolic blood pressure.

View larger version (13K): [in this window] [in a new window]   Figure 1. Relative risk for stroke and coronary heart disease by usual diastolic blood pressure. Stroke data were compiled from 7 prospective observational studies with 843 events, and coronary heart disease data were compiled from 9 prospective, observational studies with 4856 events. The 5 categories of diastolic blood pressure (DBP) are defined by baseline diastolic blood pressure. Estimates of the usual diastolic blood pressure in each category are taken from mean diastolic blood pressure values 4 years after baseline in the Framingham Study.

This observational association neatly matches experimental experience. In 14 intervention trials, a 5- to 6-mm Hg decrease in diastolic blood pressure was associated with a 42% decrease in cerebrovascular events [5]. Furthermore, this benefit was approximately the same whether study participants had mild, moderate, or severe hypertension. For coronary events, the protection associated with the same blood pressure reduction was less than predicted by observational studies. Nevertheless, roughly the same percentage reduction in events occurred for a given blood pressure decrease, regardless of initial pressure.

The practical implication of continuously increasing risk associated with rising blood pressure is clear. Reduction of blood pressure should prevent cardiovascular disease regardless of level of initial pressure [18, 19]. For example, among 50- to 59-year-old persons, a 5-mm Hg reduction in diastolic pressure would prevent about 16% of all heart attack deaths [9]. By comparison, the alternative approach of treating all the roughly 5% whose diastolic pressures exceed 100 mm Hg to attain a diastolic pressure of 85 mm Hg would prevent about 9% of such events. These two approaches, however, are in no way incompatible. In fact, for the few individuals whose cardiovascular disease risk (absolute risk) is great, more substantial blood pressure reduction can offer even greater benefit.

In contrast to relative risk, absolute risk quantifies the probability of an event occurring in a population. In cardiovascular practice, it reflects the sum of an individual's entire constellation of clinical characteristics. It can be estimated with considerable precision by measuring the levels of conventional risk factors and by detecting preclinical disease.

The process of defining absolute risk exposes the great prognostic differences between patients, even those with identical blood pressure. For example, a hyperlipidemic, glucose-intolerant, 55-year-old male smoker with left ventricular hypertrophy has a distinctly greater risk than does a 55-year-old male nonsmoker without cardiac enlargement—at every level of blood pressure (Figure 2). The increasing relative risk associated with increasing blood pressure persists among persons with higher or lower absolute risk. By contrast, the actual hazard at any given blood pressure level can vary dramatically depending on overall risk. In this example, persons at lower absolute risk, even those with systolic pressures greater than 195 mm Hg, are only one tenth as likely to have an event as are those of the same age and sex with systolic blood pressures of 105 mm Hg but at high absolute risk (46 versus 372/1000) [20].

View larger version (18K): [in this window] [in a new window]   Figure 2. Absolute and relative risk for a Cardiovascular disease event in a high- and low-risk 55-year-old man by systolic blood pressure. High Risk: left ventricular hypertrophy, cigarette smoker, glucose intolerance, cholesterol = 8.02 mmol/L.

Experimental evidence supports predictions based on these observations. Participants in the Medical Research Council (MRC) trial, all of whom had the same blood pressure levels, could be divided into high- and low-risk groups (age, 65 versus 45; serum cholesterol in mmol/L, 8 versus 5; body mass index, 30 versus 27; ischemic electrocardiogram, positive versus negative; cigarettes/d, 40 versus 0). Untreated, they experienced 149 and 3.7 events, respectively, per 1000 patient-years [21]. Diuretic-based therapy generated equal blood pressure reduction in both groups and produced roughly the same percentage (35% and 22%) cardioprotection. In the high-risk group, however, 52 events per 1000 were prevented, whereas in the low-risk group, the benefit was less than one [1]. Thus, it was possible to distinguish within a single blood-pressure stratum, in advance and using tools less powerful than those now available, different levels of risk and potential for benefit.

Current Clinical Practice

Current practice ignores these clinical factors. Many physicians rely on a blood pressure threshold alone to diagnose and guide treatment. By missing the crucial distinction between blood pressure as a quantity (risk factor) rather than a quality (disease), many hypertensive patients, despite little hope of benefit measured by reduction of risk for stroke or myocardial infarction, are subjected to life-long drug therapy. It is also possible, although not proved, that longer-term trials would reveal the ability of blood-pressure reduction to prevent disease progression. At the same time, many high-risk normotensive persons, who will incur more than one half of all heart attacks, are excluded from the cardioprotective potential of hypotensive therapy.

In the MRC trial, where entry to treatment was based on a threshold blood pressure, 850 persons were treated for a year to prevent one stroke [22]. The inclusion of many low-risk persons in the MRC trial undoubtedly contributed to the high treated-to-benefitted ratio, but that is, of course, just the point. Nevertheless, if this modest benefit could be realized at no or some known small cost, as in childhood vaccination, then the effort might be justified. Unfortunately, treatment carries risk. Every labeled and treated person is exposed to possible adverse consequences [23]. Electrolyte and metabolic disturbances are common, and impotence is frequent [24]. Furthermore, it has been postulated, but not proved, that too great a decrease in pressure (the J-shaped curve) may actually induce the myocardial infarction that treatment was prescribed to prevent [25, 26]. All of these proven and postulated consequences of treatment have led some authorities to argue that the equivocal results of antihypertensive trials in regard to prevention of myocardial infarction reflect the net result of heart attacks avoided, less others caused by overzealous treatment [27].

But what about the failure to treat those many "normotensive" persons who are at great absolute risk? Figure 3 shows that a given systolic blood pressure difference predicts roughly similar (30%) cardiovascular protection in three groups with different pressures [28, 29]. The potential benefit thus varies with absolute risk. Group A men, with seemingly normal systolic pressures (135 mm Hg) have a greater probability of disease than do members of groups B and C despite their higher starting pressures (165 and 195 mm Hg, respectively). A decrease from 135 to 105 mm Hg could be expected to prevent 103 events in group A, whereas the same 30-mm Hg decline would prevent only 45 and 30 events, respectively, in the lower-risk groups, despite their initially higher pressures. Regrettably, no clinical trial has tested this hypothesis. However, because it has been shown repeatedly that the same pressure reduction yields the same benefit in severe, moderate, and mild hypertensive persons, it is hard to imagine that, at least in the middle range of blood pressure (80 to 100 mm Hg), benefit would disappear simply because some arbitrary threshold had been breached.

View larger version (16K): [in this window] [in a new window]   Figure 3. Cardiovascular disease events by systolic blood pressure change for high-, medium-, and low-risk men. High risk: systolic blood pressure = 135 mm Hg, cholesterol = 8.02 mmol/L, glucose intolerance, left ventricular hypertrophy. Medium risk: systolic blood pressure = 165 mm Hg, cholesterol = 6.72 mmol/L, glucose intolerance, no left ventricular hypertrophy. Low risk: systolic blood pressure = 195 mm Hg, cholesterol = 4.78 mmol/L, no glucose intolerance, no left ventricular hypertrophy.

As noted above, treatment does carry risk. The notion of treating "normotensive" persons is, at first glance, threatening. However, as noted above, the risk for disease determines the absolute potential for benefit, and because the hazard of treatment remains relatively stable, concentrating treatment on high-risk patients (even with a diastolic pressure of less than 90 mm Hg) seems the most prudent way to maximize the benefit-to-cost ratio. In fact, results from the recently reported SOLVD study [30] are reassuring in that regard. In that study, blood pressure reduction in high-risk normotensive persons reduced the incidence of myocardial infarction. In the SOLVD study, although involving a very special group of patients from whom wide extrapolation is unjustified, a treatment-induced decrease in diastolic pressure from 76 to 72 mm Hg, compared with no change in the control group, was associated with a 28% reduction in heart attacks [30]. Consistent with that experience is the recently completed "Trial of Treatment for Isolated Systolic Hypertension in the Elderly (SHEP)," in which reduction of normal diastolic blood pressure was associated with a decline in cardiovascular morbidity and death [31]. All SHEP patients did have elevated systolic pressure, which was also reduced. Reduction of systolic or diastolic pressure, or both, appears useful, and this seems to be true on both sides of the imaginary threshold conventionally used to separate normal from abnormal blood pressure.

An Alternative Clinical Strategy

Provision of hypotensive drugs should depend on absolute risk and potential for benefit rather than on blood pressure level alone. As noted above, various tools are available—and soon more will be—to measure absolute risk in individual patients. The known risk factors themselves provide a good start. They tend to cluster, probably because of their common metabolic influences, and thus multiply the danger [32]. Smokers with an unfavorable family history, high cholesterol, and high fasting blood sugar are at considerable risk for a cardiovascular event, even when their diastolic blood pressure is 85 mm Hg. Other factors, such as plasma renin [33] and insulin [34], promise to increase the list of factors that can together provide a closer approximation of absolute risk. Detection of vascular disease in its preclinical state improves identification of patients at high risk. It is now possible to detect traces of urinary albumin [35] and early narrowing of peripheral vessels [36], and, by echocardiography, to precisely describe cardiac size [37]. Risk factors and preclinical disease together form the basis for identifying persons whose potential for benefit justifies blood pressure reduction, even when the pressure is normal. Indeed, formal strategies to quantify total risks are being developed and applied [38, 39].

The American Heart Association, for example, has produced a scheme that uses the Framingham experience as the basis for assigning specific numeric values to various demographic characteristics, as well as to the presence and extent of risk factors [40-42]. The sum of these numbers Figure 4 can be used to provide an index of absolute probability of an event for each individual. In the examples presented in Figure 4, it can be seen that different 55-year-old white men (panels B and C) with the same absolute risk can have either clearly normal (panel B) or clearly hypertensive (panel C) levels of systolic pressure. By the same token, at either pressure (136 or 170 mm Hg), the probability of events can be lower or higher than average for all persons of that age and sex. Because a given reduction in blood pressure can be expected to produce a roughly equivalent percentage decrease in risk for events, it is probable that with the same degree of blood pressure reduction, more persons at greater absolute risk, such as those in panels A and D, would benefit than would those in panels B and C. For every patient, quantification of risk, albeit imperfect, can provide useful clues to the baseline risk and the potential for benefit that therapy could yield. This approach is readily amenable to computer analysis. Thus, the implications of different therapeutic options could be readily explored.

View larger version (19K): [in this window] [in a new window]   Figure 4. Ten-year probability of stroke by systolic blood pressure for four 55-year-old white men with different risk profiles. Patient A = previously treated, history of cardiovascular disease and diabetes, smokes cigarettes, and has left ventricular hypertrophy; patient B = smokes cigarettes; patient C = no other risk factors; patient D = history of diabetes, smokes cigarettes, and has left ventricular hypertrophy; patient E = average risk.

Replacement of an antihypertensive treatment policy dictated by threshold with one driven by a risk–benefit analysis will certainly complicate the lives of physicians and patients. Without simple guidelines, it is probable that those relatively few patients with the highest levels of diastolic blood pressure (>104 mm Hg) will continue to receive antihypertensive drug therapy regardless of the constellation of other factors. Similarly, those at the lowest levels (<80 mm Hg), regardless of absolute risk, should probably not be treated with drugs.

In practice, important changes in thinking are needed for the vast number of persons whose pressures are at or near (on both sides) the current arbitrary threshold. At this juncture, it is worth emphasizing the imprecision with which pressure is measured. Even with the most laborious and time-consuming approaches, the assignment of a specific blood pressure value to an individual patient is an inexact science. The middle zone (80 to 104 mm Hg diastolic or 140 to 180 mm Hg systolic) includes about 70% of all adults [43]. Within this range, the decision to administer drugs for each individual should be based on absolute risk.

Clearly, present knowledge cannot guarantee that all those at high risk will be identified. However, existing tools can discriminate better than an arbitrary blood pressure cut-point between those who might benefit and those whose risk is so slight that treatment offers little promise of help. Comprehensive patient assessment will identify some patients in the ranges previously described as normotensive who, by virtue of family history, glucose intolerance, minimal renal dysfunction, left ventricular hypertrophy, and a high serum cholesterol level, are at increased cardiovascular risk. Reduction of blood pressure should produce real benefit for these patients. The other modifiable risk factors should, of course, also be vigorously pursued. In fact, it may sometimes be possible, by eliminating or reducing other risk factors, to convert a high-risk into a low-risk person and thus reduce the need for antihypertensive drugs. Studies to quantify the precision and test the validity of this approach should be high on the research agenda of preventive cardiologists.

Other patients who do not smoke, have low cholesterol and fasting blood sugar levels, a low renin profile, a normal echocardiogram, and no evidence of renal dysfunction will have diastolic pressure levels of about 100 mm Hg; these patients should be spared drug therapy. The experience of controls in the Australian National Trial of Antihypertensive Treatment provides reassurance about the safety of withholding drug therapy [44]. Fully 78% of untreated persons whose initial diastolic pressure was 100 to 104 mm Hg had pressures at some time during the study of less than 100 mm Hg [45]. Most importantly, that majority whose pressure tended to decrease spontaneously had event rates indistinguishable from those who were treated to reach the same pressure. Excess morbidity was limited to the 22% of controls whose pressure remained the same or increased during the study. Presumably, in patients with mild hypertension without other risk factors or evidence of preclinical disease, an increase in blood pressure would precede any adverse pressure-related cardiovascular outcome. Thus, regular patient monitoring could provide the appropriate signal for timely intervention.

This new paradigm must be effectively communicated to patients. Options must be presented so that absolute risk and the potential for benefit are clear. Then, informed by these quantitative data, each patient's unique way of weighing risk and benefit can lead to rational decision making. Because personal values and preferences differ, individual therapeutic decisions in the same circumstances can legitimately be expected to vary.

Summary

Top Summary Author & Article Info References

The issue is not whether blood pressure is a risk factor for cardiovascular disease or whether its reduction saves lives. That has been convincingly demonstrated. The issue is how best to apply that knowledge. Logic argues for displacing the whole community's pressure distribution to the left. If only a small change could be feasibly, acceptably, and safely achieved, it would prevent more disease than any conceivable clinical strategy. Regrettably, knowledge about what to do to achieve this goal and how to do it leaves much to be desired. Investment in fundamental and applied research to improve dietary and behavioral approaches to attaining these goals is needed.

But physicians do have powerful and varied tools to dramatically lower blood pressure in individual patients. How and when to use those tools wisely is the challenge. Blood pressure must be seen for what it is: a reflection of relative risk, a risk factor, and only one of the many factors that together determine absolute risk. Blood pressure reduction does not cure (because high blood pressure is not a disease) but rather lowers the odds of developing real disease (heart attack, stroke, and renal failure). The need for hypotensive therapy should therefore be determined by the absolute likelihood of a stroke or heart attack and the associated opportunity for their prevention and not by any particular blood pressure level. There must be some proportionality between potential for benefit and burden of intervention. The demanding task of clinical medicine is to define those parameters; to assist the patient in assessing their implications; and, in implementing the therapeutic choice that follows, to provide the best possible care.

Because the relation of blood pressure to cardiovascular disease is continuous, reliance on a threshold value to determine therapeutic behavior is not valid. Instead, effective and efficient clinical management should be based on quantitative assessment of each person's absolute risk and potential for benefit.