OUP user menu

Blood pressure targets of antihypertensive treatment: up and down the J-shaped curve

Alberto Zanchetti
DOI: http://dx.doi.org/10.1093/eurheartj/ehq281 2837-2840 First published online: 27 October 2010

This editorial refers to ‘Effects of a fixed combination of perindopril and indapamide in patients with type 2 diabetes and chronic kidney disease’, by H.J. Lambers Heerspink et al., on page 2888 and ‘J-curve revisited: an analysis of blood pressure and cardiovascular events in the Treating to New Targets (TNT) Trial’, by S. Bangalore et al., on page 2897

The solid evidence (mostly from placebo-controlled trials) that lowering blood pressure (BP) in hypertensives is accompanied by reduced incidence of cardiovascular outcomes generated the opinion that the lower the BP achieved the greater the outcome reduction. It has also been inferred that the direct relationship between BP and incident outcomes, down to at least 115 mmHg systolic BP (SBP) and 75 mmHg diastolic BP (DBP), described in uncomplicated individuals,1 can also be observed in treated hypertensives, and high-risk individuals might benefit from further BP lowering even if initially normotensives. These opinions and recommendations have recently been reappraised by a Task Force document of the European Society of Hypertension (ESH) on the basis of a critical analysis of available evidence from trials.2 It has been concluded that there is no solid evidence in favour of the recommendation that high-risk patients should require a SBP target below 130 mmHg and should receive antihypertensive agents even when their BP is in the high-normal range.2

Furthermore, several groups of authors have recently revisited the hypothesis of a J-shaped relationship between BP achieved by treatment and incident cardiovascular events. This hypothesis continues to be widely popular for several reasons: (i) common sense indicates a threshold BP must exist below which survival is impaired (at 0 mmHg all would be dead); (ii) physiology has shown there is a low (as well as high) BP threshold for organ blood flow autoregulation, and this threshold can be elevated when there is vascular disease; and (iii) there is a persisting heritage from an old conception of medicine seeing high BP as a compensatory mechanism with the task of preserving organ function.

Two interesting articles with impact on the issue of the safest BP target for antihypertensive treatment have recently been published.3,4 The first article3 is a subgroup analysis of ADVANCE, a large trial on diabetic patients in which additional administration of an angiotensin-converting enzyme (ACE) inhibitor–diuretic combination was compared with additional placebo.5 Current enthusiasm for lowering BP in diabetes was largely based on the results of early trials, such as HOT,6 UKPDS,7 SystEur,8 and SHEP,9 in which even small BP differences within the SBP range 140–150 mmHg were accompanied by large reductions in cardiovascular outcomes (25–40%).10 ADVANCE5 was able to show that benefits can also accompany SBP reductions within the 130–140 mmHg range, but the benefit is small (9% reduction in the composite endpoint of macrovascular plus microvascular outcomes, with a non-significant 8% reduction of macrovascular outcomes). The recent publication of the results of ACCORD11 has added the important information that no significant reduction in cardiovascular outcomes can be achieved in diabetics by reducing SBP to 119 rather than 133 mmHg. Seen all together (Figure 1), trials of antihypertensive treatment in diabetics appear to indicate that smaller benefits occur for SBP differences in the 130–140 mmHg range than in the 140–150 mmHg range, and suggest that no further benefit of antihypertensive therapy can be achieved in diabetics by lowering SBP below 130 mmHg.12

Figure 1

Achieved systolic blood pressure (SBP) values and reductions in cardiovascular (CV) events in trials of antihypertensive treatment in diabetics. Achieved SBP values are indicated in the histograms (yellow, less intensive treatment; brown, more intensive treatment) with ordinates on the left. % CV event (cardiovascular deaths, non-fatal myocardial infarctions, and strokes) reductions are indicated by the filled circles with ordinates on the right. Reductions are those reported in the original article, or calculated approximations when the combined cardiovascular outcome as specified above was not used in the original report. Data from the following trials are included: S.Eur.DM, Systolic Hypertension in Europe, diabetic subgroup;21 SHEP DM, Systolic Hypertension in the Elderly Program, diabetic subgroup;22 UKPDS, United Kingdom Prospective Diabetes Study;7 HOT DM, Hypertension Optimal Treatment, diabetic subgroup;6 HOPE, Microalbuminuria, cardiovascular and Renal Outcomes in the Heart Outcomes Prevention Evaluation;23 ADV = ADVANCE, Action in Diabetes and Vascular disease, Preterax and Diamicron-MR Controlled Evaluation;5 ABCD, Appropriate Blood pressure Control in Diabetes (HT, hypertensive subgroup;24 NT, normotensive subgroup25); ACRD = ACCORD, Action to Control Cardiovascular Risk in Diabetes.11 Modified from Zanchetti et al.12

In the further analyses,3 ADVANCE authors have investigated whether the relative benefits of SBP reduction down to the lower 130 mmHg range are greater in diabetic patients at higher risk. Both in ADVANCE5 and in ACCORD,11 cardiovascular event incidence was lower than expected (2.2 and 2.0% per year), but in the ADVANCE diabetics with chronic kidney disease ≥stage 3 studied by Lambers Heerspink et al.3 cardiovascular outcome incidence rose to 3.3% per year. Despite this higher risk, a greater relative benefit of more aggressive BP lowering could not be proved, although the absolute benefit was greater, leading to the prevention of 12 rather than six cardiovascular events per 1000 patients over 5 years.

While studies such as ADVANCE3,5 and ACCORD11 raise the issue of whether, at least in diabetics, the benefit of BP lowering gradually flattens at lower SBP/DBP, the J-curve hypothesis goes even further and suggests risk may indeed increase. However, correct investigation of the J-curve requires comparison of three BP targets. So far, only the HOT study6 approached the issue in this way, but this trial was designed at a time when interest was concentrated on DBP and BP used to be reduced less aggressively; furthermore, HOT was conducted in hypertensive patients at low risk (∼1% cardiovascular events per year). Even the recent ACCORD11 compared only two treatment targets with a large SBP/DBP difference (14.2/6.1 mmHg) and cannot, therefore, answer the question of whether the most beneficial (=nadir) BP may be located between the two achieved targets.

Due to lack of direct evidence on the J-curve, recourse has been and is being made to an indirect approach: incident outcomes in randomized trials of antihypertensive treatment are plotted against achieved BP, independently of the randomized group, and the existence of a uniform direct relationship or a J-shaped relationship is explored. The method was inaugurated by analyses of EWPHE13 and HOT,6 and then extended to other trials. The results of a further important analysis of this type have now been published,4 adding new information to available evidence.

On the whole, the existence of a J-shaped curve has been investigated for both SBP and DBP, although more often for DBP, on the assumption that a critical level of DBP is particularly important for maintaining coronary blood flow. Some of the previous analyses have concluded that no J-curve exists (HOT,6 PROGRESS,14 and UKPDS15), while others have concluded in favour of its existence (INDANA meta-analysis,16 IDNT,17 INVEST,18 and ONTARGET19), although INDANA found it in both actively treated and placebo patients, and in SystEur a J-shaped curve for DBP was found only in placebo-treated patients.20 The data provided by Bangalore et al.4 are particularly interesting not only because they show once more a J-shaped curve, but also because they are derived from a trial of lipid lowering [Treating to New Targets (TNT)] rather than of antihypertensive treatment. Although 54% of the TNT patients were classified as hypertensive at baseline, antihypertensive therapy was not systematically changed during follow-up. Furthermore, as could be expected, the proportion of hypertensives was much lower in the lowest SBP group (26.5%). The finding in a trial such as TNT of a J-shaped curve relationship both for SBP and for DBP, and for all types of outcomes tested despite no substantial change in antihypertensive therapy, may be taken to support the conclusions of the INDANA meta-analysis16 and SystEur20 that the J-curve, if it exists, may not be the result of antihypertensive therapy.

Unfortunately, all attempts to investigate the J-curve by post-hoc analyses of trials have obvious limitations, correctly acknowledged by Bangalore et al.4 First, the major virtue of a trial, i.e. randomization, is lost. Secondly, the numbers of patients and events in the lowest SBP/DBP group are very small (4 and 1% of TNT patients for SBP and DBP, respectively). Thirdly, patients in the lowest BP groups differ markedly from those with higher BP, and often are at increased baseline cardiovascular risk (in TNT, patients with the lowest SBP had a higher previous incidence of myocardial infarction and heart failure). Although these baseline disparities are usually adjusted by sophisticated statistics, that statistics can entirely correct what nature has done can only be believed by those taking statistics as gospel rather than as a research instrument.

Finally, Figure 2 shows that the ‘nadir’ SBP/DBP values found in various studies are extremely different (SBP ranging from 169 mmHg in the INDANA controls11 to 112 mmHg in PROGRESS14; DBP from 94 to 72 mmHg). It does not appear that the different nadir BPs depend on a different baseline level of cardiovascular risk, which in most studies was high, or on average BP at randomization or average achieved BP (Figure 2). In many of the studied cohorts, independently of the BP ranges included, and of the presence or absence of antihypertensive treatment, those patients with BP in the lowest part of the range appear to be prone to a higher incidence of outcomes. Even in TNT,4 although antihypertensive treatment is unlikely to have been intensified in patients with a lower BP, these patients showed a small but consistent decrease in SBP/DBP during follow-up, whereas BP slightly increased in all other patients, suggesting worsened general health in patients with the lowest BP.

Figure 2

Nadir systolic (SBP) and diastolic blood pressure (DBP) values in trials investigating the J-curve. Risk of the various cohorts is indicated as % cardiovascular event (see definition in Figure 1) incidence in 5 years, except for INDANA, in which risk is of all-cause death. Randomized (Rand) and achieved SBP and DBP are also indicated. IND = INDANA meta-analysis16 including seven trials (C, control and A, actively treated patients); TNT, Treating to New Targets;4 HOT, Hypertension Optimal Treatment;6 ONT = ONTARGET, Ongoing Telmisartan Alone and in Combination with Ramipril Global Endpoint Trial;19 IDNT, Irbersartan Diabetic Nephropathy Trial;17 INV = INVEST, International Verapamil SR/Trandolapril study;18 UK = UKPDS, United Kingdom Prospective Diabetes Study;15 PR = PROGRESS, Perindopril Protection against Recurrent Stroke Study.14

Conclusions and practical applications

The number of studies investigating the J-curve are witness to the theoretical and practical importance of learning to what level SBP/DBP can be lowered to increase the overall benefit of the hypertensive patient without causing endangering hypoperfusion. Unfortunately, in the absence of correctly designed trials, recourse must be made to methods necessarily prone to bias, and all results are open to different and even contrasting interpretations. For the time being, it appears wise and safe to lower SBP/DBP to values within the range 130–139/80–85 mmHg, and possibly close to lower values in this range.2 However, wisdom should not be taken as evidence.12 and important issues such as goal BP and the J-curve should urgently be approached by a correctly designed randomized trial.

Conflict of interest: none declared.


This manuscript has been prepared in the context of the activities of the EC InGenious HyperCare Network of Excellence (contract LSHM-CT-2006-037093). The author is grateful to Mrs Donatella Mihalich for her valuable help.


  • The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.


View Abstract