J-curves in hypertension: what do they tell us about treatment of high blood pressure?

This editorial refers to ‘Achieved diastolic blood pressure and pulse pressure at target systolic blood pressure (120–140 mmHg) and cardiovascular outcomes in high-risk patients: results from ONTARGET and TRANSCEND trials’†, by M. Böhm et al., on page 3105. The physiological determinants of diastolic blood pressure (DBP) and pulse pressure (PP) arise from a complex interaction between the heart and peripheral vascular structure and function. The dominant factors arising from the heart are the amount of stroke volume ejected to the time of peak pressure, as well as the rate of ventricular ejection (equating to aortic flow). Total arterial compliance (the inverse of aortic stiffness) is the major contributor from the vascular tree.1 Under resting conditions, roughly 85% of the perfusion of the left ventricle occurs during diastole.2 Thus, systemic DBP (minus diastolic pressure in the left ventricle) as the driving pressure is a major determinant of perfusion of the left ventricle. On the other hand, autoregulation of coronary blood flow serves to maintain a constant myocardial blood flow (related to metabolic needs) with changing perfusion pressures.3 In humans, the lower end of autoregulation (when DBP becomes too low to maintain myocardial perfusion) is not known and cannot be determined experimentally for obvious reasons. However, in dogs, a DBP as low as 32 mmHg has been observed as a critical pressure limit.4 In patients with coronary artery disease, a >70% cross-sectional area reduction in coronary arteries may be associated with preserved distal perfusion (due to dilatation of the microcirculation) but reduced maximum coronary blood flow. However, a cross-sectional area reduction of 85–90% is associated with impaired resting flow.5 With increasing age, systolic blood pressure (SBP) rises continuously, whereas DBP reaches a plateau at 50–60 years, and decreases afterwards, leading to an increase in PP.1 , 6 This reduction in DBP occurs in parallel with increasing aortic stiffness (which may be the underlying condition for this change in DBP).7 In young individuals, the aorta is compliant and each cardiac contraction not only transmits forward flow of the blood column, but also creates a pressure wave, which travels in the aortic and arterial wall downstream at a given speed (pulse wave velocity in young individuals is ∼5 m s–1). There is still contention regarding the exact mechanisms of arterial wave travel and how this may contribute to disease.8 However, a long-held theory is that at regions of impedance mismatch (e.g. vascular bifurcations), the energy is partially reflected, returns to the heart and the ascending aorta at the same speed, and merges with the antegrade wave during the same cardiac cycle, mainly during diastole in young people, resulting in augmentation of coronary perfusion. In contrast, increasing age is usually associated with aortic stiffening, increased pulse wave velocity, and occurrence of wave reflections during late systole, leading to increased SBP (and PP) and decreased DBP.9 Accordingly, aortic stiffness (and pulse wave velocity) is strongly and independently related to basal and hyperaemic coronary blood flow in humans following successful coronary interventions.10 When coronary disease comes into play, both aortic stiffening and coronary obstruction interact to decrease coronary perfusion. In patients with established coronary artery disease, measures of central arterial stiffness are independently and inversely related to ischaemic threshold.11 Additionally, in patients following coronary interventions, measures of central systolic loading, probably from increased wave reflections, are independent predictors of cardiovascular events.12 The study by Michael Boehm and colleagues in this issue of the journal13 examined the association between mean attained DBP and cardiovascular outcomes in patients who achieved an on-treatment SBP in the range of 120–140 mmHg (which was associated with the lowest cardiovascular risk) in the ONTARGET and TRANSCEND trials. Briefly, outcome data were analysed from high-risk patients aged 55 years or older with cardiovascular disease who were randomized to ramipril, telmisartan, and the combination. The composite outcome of cardiovascular death, myocardial infarction, stroke and hospital admission for heart failure, the individual components, and all-cause mortality were analysed according to mean on-treatment DBP as categorical (<70, 70–79, 80–89, and ≥90 mmHg) and continuous variables, as well as the change of DBP according to baseline DBP. PP was related to outcomes as a continuous variable. Based on their analyses, the authors conclude that in these patients with controlled SBP, a DBP of 70–80 mmHg was associated with the lowest event rate, whereas there was increased risk among those with both lower and higher DBP. Associations of DBP and PP with risk were similar. These data may suggest that despite optimal achieved SBP, risk may still be defined by low or high DBP13 (Take home figure). Indeed, a similar study among patients with isolated systolic hypertension in the Framingham Heart Study also found that the same cut-off point of ∼70 mmHg DBP was associated with increased events, but the risk was elevated among those with both low DBP and wide PP.14 Unfortunately, the authors missed the opportunity to perform such an analysis in the present study.13 A large sample size and a strict BP control and BP assessment in this landmark analysis are strengths of this study. However, the retrospective design makes the findings hypothesis generating rather than definite. Thus, the results have to be confirmed in a prospective randomized study. The authors correctly suggest that the findings of their study, and indeed all BP-focused studies, should be interpreted in light of the BP measurement method employed. In this current study, an automated oscillometric device was used to measure seated BP and, although the device has passed validation testing in the general population, recent data suggest that BP devices may be less accurate among people of higher cardiovascular disease risk.15 , 16 Furthermore, a compilation of data on the ability of BP devices to measure the true intra-arterial BP correctly showed systematic underestimation of intra-arterial brachial SBP (–6 mmHg) but overestimation of brachial DBP (+6 mmHg), such that PP was underestimated by 12 mmHg on average.14 Importantly, inaccuracy of cuff BP devices was most pronounced in the SBP zone of 120–160 mmHg,17 within that of all participants in the current study. Ultimately, it appears that there may be distinctive ‘higher risk BP phenotypes’ that have higher intra-arterial BP at the central aorta, but which are unable to be discerned from standard cuff BP measurements18 as used in the study.13 The level to which these above factors could have some influence on the findings of Boehm and colleagues13 is unknown, but of future interest to understand more fully. In any case, findings are consistent with many studies over several decades showing that there is a J-curve relationship between DBP and cardiovascular disease mortality, particularly among the elderly1 and those with established coronary artery disease.19–22 The additional feature of this study is the consistency of this observation among people with high cardiovascular disease risk and history of hypertension, but with apparently well-controlled SBP. Considering the pathophysiological background with a close relationship between low DBP in elderly individuals and an increase in aortic stiffness and wave reflections (which are closely linked to coronary perfusion), it is no surprise that observational studies in the last 30 years have consistently shown a strong association between a lower DBP and an increase in cardiovascular events. In the Framingham population, in patients older than 60 years, SBP was directly, but DBP was inversely related to the risk of coronary artery disease (a lower DBP conferred a higher risk).23 One of the first publications investigating studies of treated hypertensives found the lowest incidence of myocardial infarction at a DBP between 80 and 90 mmHg.24 In the large INVEST study (International Verapamil–Trandolapril Study) in patients with coronary artery disease, the nadir of the curve indicating the lowest incidence occured at 83.4 mmHg.19 In that study, the J-curve was much more prominent in patients without coronary revascularization, further supporting the concept. More recently, in a large observational registry25 of patients with stable coronary artery disease, a DBP of <70 mmHg was associated with an increase in the composite endpoint of cardiovascular death, myocardial infarction, and stroke. Of note, mean changes in SBP and DBP during 5 years in this registry were roughly 2 and 2 mmHg, respectively. This suggests that the increased risk was mainly related to baseline DBP and not to treatment-induced DBP changes. In the study published in this issue of the journal,13 the authors observed an increased risk of cardiovascular events at on-treatment DBP <70 mmHg in well-treated study participants. Remarkably, in this group of patients, SBPs were roughly the same across all quartiles of attained DBP. Therefore, a higher PP as a surrogate of higher aortic stiffness was directly related to a lower DBP. Despite being an unlikely explanation, the J-curve is often attributed to reduced myocardial perfusion from critically low coronary filling pressure (occurring during diastole). However, brachial cuff DBP only gives a single point estimate of pressure during diastole and correlates poorly with subendocardial viability (derived from arterial pressure waveform analysis) irrespective of coronary artery disease status.26 Reverse causality is certainly a potential explanation because significant underlying disease of the heart (e.g. impaired left ventricular function) and vasculature (increased aortic stiffness/increased peripheral vascular resistance) causes both the low DBP and the increased risk for cardiovascular events.21 However, as previously discussed, inaccuracy of cuff BP measurement may limit the strength of this explanation. Altogether, the issue of the BP measurement method with the line of arguments discussed in the previous section would paint a very new perspective on J-curve associations. It may indeed be reverse causation, but for which measurement inaccuracy is an additional confounding factor has only just recently been recognized and is unable to be accounted for in the analysis by Michael Boehm and colleagues.13 In the HOT study,27 cardiovascular events occurred with similar rates in all three DBP target groups (≤90, ≤85, and ≤80 mm Hg). In addition, there was a clear benefit for the lowest target in the subgroup of diabetic patients. In a meta-analysis of 10 trials,28 not only the levels of BP, but also the effect of treatment (active treatment vs. placebo or no treatment) was taken into account. Using matched-pair analysis (actively treated patients were matched with untreated controls from the same trial by gender, age, smoking, previous cardiovascular events, and SBP and DBP at entry) and subdividing into quartiles according to the decrease in DBP in the actively treated patients (from –4 to –25 mmHg), the authors found that the benefit of antihypertensive treatment was not modified or attenuated by the decrease in DBP. In the matched-pair analysis, those patients with an achieved on-treatment DBP below the 10th percentile (124/62 mmHg) had huge reductions in cardiovascular events as compared with their matched untreated pairs (154/84 mmHg). Finally, a post-hoc analysis of the SPRINT study investigated whether the effects of the SBP intervention differed by baseline DBP.29 When only BP was analysed regardless of the randomized treatment, baseline DBP had a U-shaped association with the hazard of the primary cardiovascular disease outcome in accordance with previous studies. However, and most importantly, the benefit of SBP lowering did not differ by baseline DBP.29 First, observational studies such as the analysis by Boehm and colleagues published in this issue of the journal,13 are hypothesis generating only, and cannot replace prospective randomized controlled trials (RCTs) to influence treatment recommendations (Take home figure). Secondly, as discussed above, we have appropriate evidence from RCTs to guide therapeutic decisions. Finally, in practical terms, how does a clinician individually titrate therapy directed at DBP separate from SBP (or vice versa) to hit a ‘sweet spot zone’ of lowest risk SBP and DBP? In conclusion, in elderly individuals treated for hypertension, a lower DBP is a surrogate of an increase in aortic stiffening and premature wave reflections. This is the most likely explanation for the increase in cardiovascular events in these patients. Treatment decisions, however, should not be based on observational evidence. Studies taking treatment into consideration strongly suggest that the benefit of lowering SBP is not attenuated by low on-treatment DBP, at least down to 60 mmHg. Effects of diastolic blood pressure (DBP), pulse pressure (PP), and arterial stiffness on cardiovascular events in hypertensive patients. Conflict of interest: none declared.