Hazard ratios for cardiovascular (A and C) and noncardiovascular mortality (B and D) in relation to on-treatment diastolic blood pressure (BP) (defined as low BP while receiving antihypertensive treatment) in the active treatment and placebo groups, adjusted for age, sex, smoking status, and previous antihypertensive treatment at baseline, and body weight, stroke, myocardial infarction, or diabetes at baseline and during follow-up. The hazard ratio gives the risk associated with a 5–mm Hg lower diastolic BP.
Hazard ratios for cerebrovascular events (A and C) and coronary heart disease (CHD) events (B and D) in relation to on-treatment diastolic blood pressure (BP) (defined as low BP while receiving antihypertensive treatment) in the active treatment and placebo groups, adjusted for age, sex, smoking status, and previous antihypertensive treatment at baseline, and body weight, stroke, myocardial infarction, or diabetes at baseline and during follow-up. The hazard ratio gives the risk associated with a 5–mm Hg lower diastolic BP.
Hazard ratios for cardiovascular events in relation to on-treatment diastolic blood pressure (BP) (defined as low BP while receiving antihypertensive treatment) in patients without (A and C) and with (B and D) coronary heart disease (CHD) at baseline in the active treatment group and in the placebo group, adjusted for age, sex, smoking status, and previous antihypertensive treatment at baseline, and body weight, stroke, myocardial infarction, or diabetes at baseline and during follow-up. The hazard ratio gives the risk associated with a 5–mm Hg lower diastolic BP.
Hazard ratios for cardiovascular events according to on-treatment diastolic blood pressure (BP) (defined as low BP while receiving antihypertensive treatment) by 5–mm Hg cutoff values in the active treatment (A) and placebo (B) groups, adjusted for age, sex, smoking status, and previous antihypertensive treatment at baseline; on-treatment systolic BP as a dichotomous variable; and body weight, stroke, myocardial infarction, or diabetes at baseline and during follow-up.
Fagard RH, Staessen JA, Thijs L, Celis H, Bulpitt CJ, de Leeuw PW, Leonetti G, Tuomilehto J, Yodfat Y. On-Treatment Diastolic Blood Pressure and Prognosis in Systolic Hypertension. Arch Intern Med. 2007;167(17):1884-1891. doi:10.1001/archinte.167.17.1884
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
It has been suggested that low diastolic blood pressure (BP) while receiving antihypertensive treatment (hereinafter called on-treatment BP) is harmful in older patients with systolic hypertension. We examined the association between on-treatment diastolic BP, mortality, and cardiovascular events in the prospective placebo-controlled Systolic Hypertension in Europe Trial.
Elderly patients with systolic hypertension were randomized into the double-blind first phase of the trial, after which all patients received active study drugs (phase 2). We assessed the relationship between outcome and on-treatment diastolic BP by use of multivariate Cox regression analysis during receipt of placebo (phase 1) and during active treatment (phases 1 and 2).
Rates of noncardiovascular mortality, cardiovascular mortality, and cardiovascular events were 11.1, 12.0, and 29.4, respectively, per 1000 patient-years with active treatment (n = 2358) and 11.9, 12.6, and 39.0, respectively, with placebo (n = 2225). Noncardiovascular mortality, but not cardiovascular mortality, increased with low diastolic BP with active treatment (P < .005) and with placebo (P < .05); for example, hazard ratios for lower diastolic BP, that is, 65 to 60 mm Hg, were, respectively, 1.15 (95% confidence interval, 1.00-1.31) and 1.28 (95% confidence interval, 1.03-1.59). Low diastolic BP with active treatment was associated with increased risk of cardiovascular events, but only in patients with coronary heart disease at baseline (P < .02; hazard ratio for BP 65-60 mm Hg, 1.17; 95% confidence interval, 0.98-1.38).
These findings support the hypothesis that antihypertensive treatment can be intensified to prevent cardiovascular events when systolic BP is not under control in older patients with systolic hypertension, at least until diastolic BP reaches 55 mm Hg. However, a prudent approach is warranted in patients with concomitant coronary heart disease, in whom diastolic BP should probably not be lowered to less than 70 mm Hg.
The existence of a J-curve relationship between mortality or cardiovascular event and diastolic blood pressure (BP) during antihypertensive treatment (hereinafter called on-treatment BP) remains a matter of debate in patients with hypertension.1 Apart from intervention trials, in which patients are randomized to different target BP levels,2 placebo- or nontreatment-controlled trials are probably the best way to evaluate the J-curve. A large meta-analysis of individual patient data from 7 trials3- 10 showed a higher incidence of cardiovascular death with lower on-treatment diastolic BP, but this relationship was also observed for noncardiovascular death and similar relationships existed in the control patients; thus, the worse prognosis of lower diastolic BP could not be solely ascribed to antihypertensive treatment. A more recent meta-analysis11 involving 10 trials4- 10,12- 14 concluded that lowering diastolic BP to less than 70 mm Hg did not cause harm. However, these meta-analyses did not differentiate between systolic-diastolic hypertension and isolated systolic hypertension. Therefore, the results may not apply to patients with systolic hypertension who may be particularly vulnerable to low on-treatment diastolic BP, as shown in a post hoc analysis of the Systolic Hypertension in the Elderly Program (SHEP), in which a J-curve was observed for incident cardiovascular events during active treatment but not during administration of placebo.15 The SHEP investigators concluded that it could be harmful to intensify antihypertensive therapy when diastolic BP has reached 70 mm Hg. In addition, the SHEP analysis did not stratify for coronary heart disease (CHD) at baseline, whereas the risk of low on-treatment diastolic BP could be particularly high in patients with CHD.1,16 Because of the far-reaching consequences of the conclusions from the SHEP trial, we analyzed the relationship between on-treatment diastolic BP, mortality, and cardiovascular events in older patients with systolic hypertension randomized into the Systolic Hypertension in Europe (Syst-Eur) Trial12,17 and also accounted for CHD at baseline.
The protocol of the Syst-Eur Trial12 was approved by the ethics committees of the University of Leuven, Leuven, Belgium, and participating centers; all subjects in the 198 centers gave informed consent. Eligible patients had to be 60 years or older. During the run-in period with placebo, patients were seen at 3 baseline visits 1 month apart. Data about medical history were collected, including history of CHD, which comprised myocardial infarction and angina pectoris, based on supporting documents and examinations available to the local investigators. At each visit, BP was measured twice with a standard sphygmomanometer with the patient sitting. Patients were admitted to the double-blind phase of the trial when they had an average run-in period systolic BP of 160 to 219 mm Hg with diastolic BP less than 95 mm Hg. Patients were randomized to active medication or matched placebo groups. Active treatment consisted of nitrendipine (10-40 mg/d), which could be combined with or replaced by enalapril maleate (5-20 mg/d) or hydrochlorothiazide (12.5-25 mg/d) or both drugs, to reduce the sitting systolic BP by 20 mm Hg or more and to less than 150 mm Hg. At each 3-month visit, BP was measured twice with the patient sitting and the 2 BP measurements were averaged. When phase 1 of the Syst-Eur Trial was stopped on February 14, 1997, after a median follow-up of 2.0 years (range, 1 month to 8.1 years),12 patients in the control group were switched to the active study treatment regimen and patients in the active treatment group continued receiving active treatment (phase 2).17 Visits were scheduled every 6 months in the study centers and the same information was collected as during the double-blind phase of the trial. The extended follow-up ended on December 31, 2001. In the present analyses, follow-up data in the placebo group are from phase 1 of the trial and follow-up data in the active treatment group are from phases 1 and 2. For patients who withdrew from the study or who could not be followed up as planned, investigators collected data about vital status, occurrence of diseases, and use of antihypertensive medication via yearly telephone contact with patients, family members, or general practitioners.12,17
Outcome measures were as follows: (1) fatal and nonfatal cerebrovascular events including stroke and transient ischemic attack; (2) fatal and nonfatal CHD events including sudden death, myocardial infarction, and coronary revascularization (percutaneous coronary interventions and coronary artery bypass surgery); (3) an aggregate of all cardiovascular events (cardiovascular death, myocardial infarction, stroke, heart failure, coronary revascularization, aortic aneurysm, and transient ischemic attack); (4) cardiovascular mortality including all fatal cardiovascular events; and (5) noncardiovascular death. Events were corroborated by the Syst-Eur Trial End Point Committee.12,17
Database management and statistical analyses were performed using SAS software (version 8.2; SAS Institute Inc, Cary, North Carolina). Data are reported as mean ± SD or as percentage. The prognostic value of diastolic BP during follow-up was analyzed in patients with at least 1 follow-up visit. In the first analysis, diastolic BP (linear and quadratic term) was entered as a time-dependent covariate in multivariate Cox regression models.18 In this analysis, the likelihood of an event that occurred at time t depends on the value of the last available diastolic BP before time t for all subjects still in follow-up at time t. We adjusted for variables with potential influence on BP and outcome, that is, age, sex, smoking status, and previous antihypertensive treatment at baseline, and weight, stroke, myocardial infarction, and diabetes mellitus both at baseline and as time-dependent covariates during follow-up. The analyses were performed with and without systolic BP as a continuous time-dependent covariate. We plotted the adjusted hazard ratio (HR) vs on-treatment diastolic BP. The (adjusted HR − 1) × 100 is the percent increase in risk associated with the 5–mm Hg lower on-treatment diastolic BP. In the second analysis, we calculated HRs in 5–mm Hg decrements of on-treatment diastolic BP, as in the SHEP study.15 For each cutoff value, outcome was compared between patients with diastolic BP below that level and patients with diastolic BP at or above that level. Adjustment was performed as described, with inclusion of on-treatment systolic BP as a dichotomized variable (cutoff value, 140 mm Hg), as in the SHEP trial.15 These analyses were repeated with on-treatment systolic BP as a continuous variable and without systolic BP. All analyses were performed on the basis of intention to treat. A 2-tailed P ≤ .05 was considered statistically significant.
Among the 4695 patients randomized in phase 1 of the Syst-Eur Trial, 2225 patients in the placebo group and 2358 patients in the active treatment group had at least 1 follow-up visit. Patient characteristics at baseline were not significantly different between the groups (Table 1). Mean ± SD age of the total study population was 70.2 ± 6.7 years; 66.9% were women. Mean ± SD BP was 173.8 ± 9.9/85.5 ± 5.8 mm Hg; 46.5% of patients had been treated with antihypertensive drugs in the last 6 months before inclusion in the trial, 29.9% had previous cardiovascular complications, and 14.5% had a history of CHD.
In the placebo group, 52 patients (2.3%) were lost to follow-up by the end of phase 1; after longer follow-up, 284 patients (12.0%) in the active treatment group were lost to follow-up by the end of phase 2. Lost to follow-up was defined as no information available for more than 1 year. At the last follow-up visit, 2070 patients in the active treatment group and 1782 patients in the placebo group received active study medication or placebo tablets; the study medication was combined with open-label antihypertensive drugs in 220 actively treated patients and 51 patients in the placebo group. One hundred fifty-seven patients in the active treatment group and 241 patients in the placebo group received open-label antihypertensive drugs only. The remaining 131 patients in the active treatment group and 202 patients in the placebo group were untreated. Mean systolic BP, that is, the mean systolic BP of all postbaseline results for each patient, was 149.0 ± 10.3 mm Hg (range,116-206 mm Hg) during active treatment and 161.8 ± 13.6 mm Hg (range, 114-238 mm Hg) during placebo administration. These values were, respectively, 78.8 ± 6.0 mm Hg (range, 50-100 mm Hg) and 83.6 ± 6.4 mm Hg (range, 50-110 mm Hg) for diastolic BP. Table 2 gives the distribution of mean diastolic BP and of all diastolic BP measurements in the 2 groups. The percentage of patients with at least 1 diastolic BP measurement less than 55 mm Hg during follow-up was 5.7% in the active treatment group and 1.1% in the placebo group. The rates of events per 1000 patient-years for cardiovascular events, cardiovascular mortality, and noncardiovascular mortality were 39.0, 12.6, and 11.9, respectively, during 5219 patient-years of follow-up during placebo treatment and 29.4, 12.0, and 11.1 during 14 511 patient-years during active treatment (Table 3).
Figure 1 shows the results for cardiovascular and noncardiovascular mortality without inclusion of on-treatment systolic BP in the regression models. On-treatment diastolic BP did not affect cardiovascular mortality, either during active treatment or during placebo administration. In contrast, noncardiovascular mortality was significantly and curvilinearly associated with diastolic BP, both with active treatment (P < .005) and with placebo (P < .05). Low on-treatment diastolic BP was associated with higher noncardiovascular mortality. For example, the HR for noncardiovascular mortality for 5–mm Hg lower diastolic BP, from 65 to 60 mm Hg, was 1.15 (95% confidence interval [CI], 1.00-1.31) with active therapy and 1.28 (95% CI, 1.03-1.59) with placebo. Results were similar for cancer mortality, for which the HRs were, respectively, 1.26 (95% CI, 1.01-1.58) and 1.36 (95% CI, 0.98-1.88).
Figure 2 shows the results for cerebrovascular events and CHD events. In the placebo group, the association between on-treatment diastolic BP and the risk of a cerebrovascular event was curvilinear, indicating an increased risk at low diastolic BP levels (P < .01); for example, the HR for 5–mm Hg lower diastolic BP, from 65 to 60 mm Hg, was 1.25 (95% CI, 1.01-1.54). This was not the case with active treatment. The CHD risk remained unchanged at lower diastolic BP with both active therapy and placebo. No significant association was found between the risk of all cardiovascular events and on-treatment diastolic BP during either active treatment or placebo administration (data not shown).
Figure 3 shows the relationship between on-treatment diastolic BP and the aggregate of all cardiovascular events according to the history of CHD at baseline. During active treatment, there were 25.6 cardiovascular events per 1000 patient-years in the 2022 patients without CHD at baseline and 56.7 events per 1000 patient-years in the 336 patients with CHD. In the placebo group, there were 24.6 events per 1000 patient-years in the 1897 patients without CHD and 73.3 events per 1000 patient-years in the 328 patients with CHD. Low diastolic BP during active treatment was not associated with increased risk in patients free of CHD at baseline, whereas the risk increased when on-treatment diastolic BP was lower than about 75 mm Hg in patients with CHD (P < .02), with an HR of about 1.1 at 70 mm Hg. However, the interaction between CHD status at baseline and on-treatment diastolic BP was not significant (P = .19). In the placebo group, there was no support for an influence of low diastolic BP on the incidence of cardiovascular events in patients with CHD at baseline, but low diastolic BP was associated with increased risk of cardiovascular events in patients without CHD (P < .01). However, the interaction between CHD status at baseline and on-treatment diastolic BP was not significant (P = .13). Results were similar when on-treatment systolic BP was included as a continuous time-dependent covariate in the analyses (data not shown).
All analyses were repeated using 5–mm Hg cutoff categories for on-treatment diastolic BP, as in the SHEP trial.15 These analyses yielded results similar to those when diastolic BP was used as a continuous variable. Similar to Figure 4 in Somes et al,15 our Figure 4 shows the results for the aggregate of all cardiovascular events, with inclusion of on-treatment systolic BP as a dichotomized variable in the model. There was no consistent association between on-treatment diastolic BP and the adjusted HR for these events with either active treatment or placebo. Results were similar when on-treatment systolic BP was used as a continuous variable or was removed from the models (data not shown).
The major findings of the present study in older patients with isolated systolic hypertension are that lowering of diastolic BP with active antihypertensive treatment, to as low as about 55 mm Hg, does not seem to increase cardiovascular mortality, whereas low diastolic BP is associated with higher noncardiovascular mortality. In patients with evidence of CHD at baseline, low on-treatment diastolic BP is associated with increased risk of cardiovascular events, which is not the case in patients without a history of CHD. Inclusion of on-treatment systolic BP in the regression models did not affect the relationships between on-treatment diastolic BP and outcome, which indicates that the findings were not confounded by the achieved systolic BP.
Many observational cohort studies reported on the shape of the relationship between BP and risk in various populations, including patients with hypertension receiving BP-lowering treatment.1 The results have not always been consistent. In addition, data from untreated subjects are unlikely to be applicable to treated hypertensive patients, and observational studies on treated hypertensive patients are difficult to interpret in the absence of a control group. A recent secondary analysis of the International Verapamil-Trandolapril Study emphasized the existence of a J-curve in actively treated patients with hypertension with stable CHD.16 The finding of a J-curve may be the result of confounding factors and ill health, associated with low BP and high risk for events (reverse causation). This has been confirmed in a large Finnish observational study in treated hypertensive patients,19 in whom the observed J-curve relationship between diastolic BP and cardiovascular and noncardiovascular mortality could be ascribed to preexisting cardiovascular disease, resulting in low diastolic BP. Placebo- or nontreatment-controlled trials are more appropriate for assessment of the existence of a J-curve for on-treatment BP because results in treated patients can be compared with those in a control group. In a meta-analysis of individual patient data from 7 randomized clinical trials involving 40 233 persons with primarily systolic-diastolic hypertension,4- 10 a J-curve relationship was observed between diastolic BP and the risk of cardiovascular and noncardiovascular mortality in both treated and untreated patients.3 The authors concluded that the increased risk observed in patients with low BP was not related to antihypertensive treatment and could probably be explained by poor health conditions. This conclusion is supported by the results from the European Working Party on High Blood Pressure in the Elderly trial in which patients with low on-treatment BP had lower values for body mass index and hemoglobin, which can be considered indicators of poor health.20
When older patients with isolated systolic hypertension are treated with antihypertensive drugs, not only systolic but also diastolic BPs are reduced.8,12,17,21 It has been argued that patients with isolated systolic hypertension and, thus, by definition having low diastolic BP, would be particularly vulnerable to the consequences of treatment-induced further lowering of diastolic BP.1 This is an important issue of clinical relevance because diastolic BP may reach very low values in these patients before systolic BP is normalized. The SHEP investigators concluded that lower achieved diastolic BP was associated with increased risk of CHD, stroke, and cardiovascular disease in actively treated patients; thus, some patients may have been overtreated.15 The relative risk for the combined cardiovascular events became significant for diastolic BP less than 70 mm Hg and approached a 2-fold increase in risk for diastolic BP less than 55 mm Hg. This adverse relationship was not observed in the control group. Our overall results differ from those of the SHEP trial. In the active treatment group, we have not observed significant relationships between on-treatment diastolic BP and, respectively, cardiovascular mortality, cerebrovascular events, CHD events, and an aggregate of cardiovascular events, also when applying the same statistical methods as in the SHEP trial.15
Several differences between the SHEP trail and the Syst-Eur Trial should be considered. The cutoff value for diastolic BP required for randomization into the trial was 90 mm Hg in the SHEP trial and 95 mm Hg in the Syst-Eur Trial. The SHEP trial investigators could analyze on-treatment diastolic BP down to 25 mm Hg, whereas diastolic BP was rarely less than 55 mm Hg in the Syst-Eur Trial. First-line treatment in the SHEP trial was a diuretic (chlorthalidone), with the possible addition of the β-blocker atenolol or reserpine for better BP control; in the Syst-Eur Trial, treatment was started with the dihydropyridine calcium channel blocker nitrendipine, to which the converting enzyme inhibitor enalapril and hydrochlorothiazide could be added. Patients with CHD at baseline were not analyzed separately in the SHEP trail. About 5% of the randomized patients had a history of myocardial infarction8; thus, it is unlikely that the percentage of patients with CHD exceeded the prevalence of CHD at baseline in the Syst-Eur Trial.
In contrast to the findings on cardiovascular mortality, we observed increases in noncardiovascular and cancer mortality at low diastolic BP, both with active treatment and with placebo. These findings are compatible with the notion that the J-curve may be explained by ill health and reverse causation. These data could also suggest that one should be careful in treating patients at high risk of noncardiovascular death.
An unexpected observation in the placebo group was that low diastolic BP was associated with a higher incidence of cerebrovascular events and not of CHD. However, the former finding agrees with the pooled analysis of 9 epidemiologic studies in patients with isolated systolic hypertension in which an inverse relationship between stroke mortality and diastolic BP was observed.22 That lower diastolic BP was not associated with increased risk of cerebrovascular events during active treatment in the Syst-Eur Trial is compatible with a greater reduction in stroke risk23 and a greater decrease in carotid intima-media thickness24 with regimens based on calcium channel blockers than with other regimens.
Some limitations of the present post hoc analysis must be considered. All patients received active study treatment after the end of the double-blind phase of the trial, which increased the number of observations in the active treatment group. The number of observations and, consequently, the power of the statistical analyses are smaller in the placebo group, which explains the larger CIs. The risk of low diastolic BP could be greatest for CHD events, particularly in patients with evidence of CHD at baseline. Whereas we observed an increased risk for an aggregate of cardiovascular events at low diastolic BP in patients with evidence of CHD at baseline, a similar but nonsignificant trend was observed for incident CHD (data not shown), related to the small number of events and the limited power of the analysis.
In conclusion, these findings support the hypothesis that antihypertensive treatment can be intensified for the prevention of cardiovascular events when systolic BP is not under control in older patients with isolated systolic hypertension, at least until diastolic BP reaches about 55 mm Hg. However, a prudent approach is warranted in patients with isolated systolic hypertension and concomitant CHD, in whom diastolic BP should probably not be lowered to less than 70 mm Hg, when the relative risk reaches about 1.1.
Correspondence: Robert H. Fagard, MD, Hypertension and Cardiovascular Rehabilitation Unit, U Z Gasthuisberg-Hypertensie, Herestraat 49, B-3000 Leuven, Belgium (email@example.com).
Accepted for Publication: May 20, 2007.
Author Contributions: Dr Fagard had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the study analysis. Study concept and design: Fagard, Bulpitt, and de Leeuw. Acquisition of data: Fagard, Staessen, Celis, Bulpitt, de Leeuw, Tuomilehto, and Yodfat. Analysis and interpretation of data: Fagard, Staessen, Thijs, Bulpitt, de Leeuw, and Leonetti. Drafting of the manuscript: Fagard, de Leeuw, and Yodfat. Critical revision of the manuscript for important intellectual content: Staessen, Thijs, Celis, Bulpitt, de Leeuw, Leonetti, and Tuomilehto. Statistical analysis: Fagard, Thijs, and Bulpitt. Obtained funding: Fagard, Staessen, Bulpitt, and Tuomilehto. Administrative, technical, and material support: Staessen, Celis, and de Leeuw. Study supervision: Fagard, Staessen, Celis, and de Leeuw.
Financial Disclosure: None reported.
Funding/Support: This study was supported by Bayer AG (Wuppertal, Germany) and the National Fund for Scientific Research (Brussels, Belgium). Study medication was donated by Bayer AG and Merck & Co Inc (Whitehouse Station, New Jersey). The trial was carried out in consultation with the World Health Organization, the International Society of Hypertension, the European Society of Hypertension, and the World Hypertension League.
Role of the Sponsors: Investigators at the University of Leuven initiated and designed the Syst-Eur Trial before seeking sponsorship. The funders had no role in the design and conduct of the trial; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Previous Presentation: This study was presented at the 15th European Meeting on Hypertension; June 18, 2005; Milan, Italy.
Additional Contributions: Nicole Ausseloos provided secretarial assistance. The Syst-Eur trial was a concerted action of the BIOMED Research Program sponsored by the European Union.