Author + information
- Received November 11, 2009
- Revision received February 9, 2010
- Accepted February 15, 2010
- Published online June 29, 2010.
- Michael A. Weber, MD⁎,⁎ (, )
- George L. Bakris, MD†,
- Kenneth Jamerson, MD‡,
- Matthew Weir, MD§,
- Sverre E. Kjeldsen, MD∥,
- Richard B. Devereux, MD¶,
- Eric J. Velazquez, MD#,
- Björn Dahlöf, MD⁎⁎,
- Roxzana Y. Kelly, MS††,
- Tsushung A. Hua, PhD††,
- Allen Hester, PhD††,
- Bertram Pitt, MD‡,
- ACCOMPLISH Investigators
- ↵⁎Reprint requests and correspondence:
Dr. Michael A. Weber, SUNY Downstate College of Medicine, 450 Clarkson Avenue, Box 97, Brooklyn, New York 11203
Objectives The aim of this study was to determine which combination therapy in patients with hypertension and diabetes most effectively decreases cardiovascular events.
Background The ACCOMPLISH (Avoiding Cardiovascular Events Through COMbination Therapy in Patients Living With Systolic Hypertension) trial compared the outcomes effects of a renin-angiotensin system blocker, benazepril, combined with amlodipine (B+A) or hydrochlorothiazide (B+H). A separate analysis in diabetic patients was pre-specified.
Methods A total of 6,946 patients with diabetes were randomized to treatment with B+A or B+H. A subgroup of 2,842 diabetic patients at very high risk (previous cardiovascular or stroke events) was also analyzed, as were 4,559 patients without diabetes. The primary end point was a composite of cardiovascular death, myocardial infarction, stroke, hospitalization for angina, resuscitated arrest, and coronary revascularization.
Results In the full diabetes group, the mean achieved blood pressures in the B+A and B+H groups were 131.5/72.6 and 132.7/73.7 mm Hg; during 30 months, there were 307 (8.8%) and 383 (11.0%) primary events (hazard ratio [HR]: 0.79, 95% confidence interval [CI]: 0.68 to 0.92, p = 0.003). For the diabetic patients at very high risk, there were 195 (13.6%) and 244 (17.3%) primary events (HR: 0.77, 95% CI: 0.64 to 0.93, p = 0.007). In the nondiabetic patients, there were 245 (10.8%) and 296 (12.9%) primary events (HR: 0.82, 95% CI: 0.69 to 0.97, p = 0.020). In the diabetic patients, there were clear coronary benefits with B+A, including both acute clinical events (p = 0.013) and revascularizations (p = 0.024). There were no unexpected adverse events.
Conclusions In patients with diabetes and hypertension, combining a renin-angiotensin system blocker with amlodipine, compared with hydrochlorothiazide, was superior in reducing cardiovascular events and could influence future management of hypertension in patients with diabetes. (Avoiding Cardiovascular Events Through COMbination Therapy in Patients Living With Systolic Hypertension [ACCOMPLISH]; NCT00170950)
There are 2 critical parts to the treatment of hypertension in patients with diabetes mellitus. According to contemporary guidelines, blood pressure should be reduced to below 130/80 mm Hg and the primary antihypertensive drug should be a blocker of the renin-angiotensin system (RAS), either an angiotensin receptor blocker or an angiotensin-converting enzyme (ACE) inhibitor (1,2).
Blockers of the RAS are known to have cardiovascular and renal protective effects in diabetic patients (3–6). For many patients, however, a single antihypertensive agent is not sufficient to achieve blood pressure control, so a 2-drug combination is recommended as initial therapy if the baseline blood pressure in diabetic patients is 150/90 mm Hg or higher (1,2). Furthermore, guidelines suggest that a thiazide diuretic usually be combined with the angiotensin receptor blocker or ACE inhibitor (1). Diuretics not only add antihypertensive efficacy when combined with RAS blockers but can also provide cardiovascular protection (7,8).
However, more recent evidence suggests that calcium channel blockers, amlodipine in particular, can reduce cardiovascular events in patients with coronary disease or high-risk hypertension (9,10). Moreover, amlodipine may independently add vasoprotective effects when combined with agents such as ACE inhibitors (11,12). To test the relative merits of these 2 types of combination hypertension therapies, the ACCOMPLISH (Avoiding Cardiovascular Events Through COMbination Therapy in Patients Living With Systolic Hypertension) trial compared treatment outcomes of ACE inhibitor + amlodipine and ACE inhibitor + hydrochlorothiazide combinations. This trial was conducted in high-risk hypertensive patients and found that the amlodipine-based combination was more effective in reducing a composite of fatal and nonfatal cardiovascular events (13).
There was a pre-specified plan in the ACCOMPLISH trial to separately compare the outcomes effects of the 2 treatment strategies in patients with diabetes. This article reports the results of this comparison, both in the full diabetic patient cohort and additionally in very high risk diabetic patients defined as those with histories of cardiac events, stroke, or renal disease. To provide a clinical context for the diabetic patients, data from the nondiabetic hypertensive patients studied in the ACCOMPLISH trial are also analyzed.
The ACCOMPLISH trial was designed to test the hypothesis that a combination of an ACE inhibitor with amlodipine would be superior to the combination of the same ACE inhibitor with hydrochlorothiazide in reducing a composite of cardiac and stroke events in high-risk hypertensive patients. The full methods for this trial were described previously (13,14). A key pre-specified analysis for the study was the comparison of the effects on end points of the 2 treatment arms in the diabetic patients included in this trial. These results are the primary focus of this report.
The study was designed, supervised, analyzed, and interpreted by the academic authors of the present report. The roles of key supporting committees were described previously (13). The database of adjudicated end points was maintained at the Duke Clinical Research Institute.
The study was performed in hypertensive patients at high risk of cardiovascular and related events. The enhanced risk in these patients was established by the presence of previously established concomitant conditions, in particular diabetes mellitus, coronary events, myocardial infarction, revascularization, stroke, impaired renal function, peripheral arterial disease, and left ventricular hypertrophy. Details of these criteria were described previously (13). The diagnosis of diabetes in the study patients was made clinically by the individual investigators. Data on patients who satisfied the trial's entry criteria but did not have diabetes were also analyzed.
Immediately on establishing eligibility and entering the study, patients were randomly assigned to 1 of 2 treatment arms: either benazepril plus amlodipine (B+A) or benazepril plus hydrochlorothiazide (B+H). All previous antihypertensive therapies were immediately discontinued and replaced by one of the study's fixed combination therapies. The starting doses were benazepril 20 mg/day plus either amlodipine 5 mg/day or hydrochlorothiazide 12.5 mg/day. The study protocol then mandated an increase in the benazepril dose to 40 mg/day in both treatment arms. Thereafter, the amlodipine dose could be increased to 10 mg/day or the hydrochlorothiazide dose to 25 mg/day if required to achieve a target blood pressure goal of <140/90 mm Hg. For the diabetic patients (who represent the principal cohort of this report) or for patients with chronic kidney disease, a target blood pressure of <130/80 mm Hg was recommended, but not mandated.
If needed for blood pressure control, investigators could add other antihypertensive agents (except ACE inhibitors, angiotensin receptor blockers, calcium-channel blockers, or thiazide diuretics), including beta-blockers, clonidine, alpha-blockers, and spironolactone. Investigators could also use once-daily loop diuretics if in their clinical judgment these agents were required for volume management. After an initial 3-month period during which all the necessary blood pressure adjustments were made, patients returned for visits after an additional 3 months and then at 6-month intervals until the end of the trial. Blood pressure was measured by standard clinical trial methodology described previously (13).
The primary study end point was the time to the first recorded event. This was defined as the composite of the first occurrence of a cardiovascular event or death from cardiovascular causes. Death from cardiovascular causes was defined as sudden death from cardiac causes or death from myocardial infarction, stroke, coronary intervention, congestive heart failure, or other cardiovascular causes. Cardiovascular events were defined as myocardial infarction, stroke, hospitalization for unstable angina, coronary revascularization, or resuscitation after cardiac arrest. Only the first event in an individual patient was counted in the analysis of the primary end point.
Independent of the calculation of the primary end point, there were pre-specified analyses of the individual components of the primary and secondary end points. In these analyses, events were counted without censoring for previous occurrence of other end points. The secondary end point of the trial was a composite of cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. Other pre-specified end points included coronary revascularization procedures, unstable angina, hospitalization for heart failure, progression of renal disease, and all-cause mortality. In addition, this report provides data on acute clinical nonrevascularization coronary end points (myocardial infarctions, sudden cardiac deaths, and hospitalizations for unstable angina). In a post hoc analysis, this report describes renal events defined as serum creatinine concentrations that increased by at least 50% during the study (regardless of their study baseline values) and were above the normal laboratory-defined range.
The power and sample size of the study were originally calculated based on the entire study cohort, with the intent that the ACCOMPLISH trial would have 90% power to detect a 15% reduction in risk for the B+A group, based on the assumption of a 3.5% annual event rate for the B+H group. However, no such calculations were made purely for the patients with diabetes cohort in this trial. Nevertheless, because the diabetic patients represented approximately 60% of the total patients enrolled in the trial, it was considered appropriate to undertake the pre-specified analysis of this patient subgroup. In addition, it was decided to analyze data in patients with diabetes who additionally had histories of cardiovascular or stroke events (referred to as diabetic patients at very high risk). To provide a clinical context for these data, analysis was also performed in the patients without diabetes enrolled in the ACCOMPLISH trial.
All study outcomes were adjudicated according to standard criteria by a blinded clinical end points committee. Interim statistical analyses were performed at 6-month intervals for the data safety monitoring committee, which was able to determine whether a difference in outcomes existed between the 2 treatment arms, without knowing the identity of the 2 treatment arms. The description of this procedure, together with a detailed explanation of the specified stopping rules for the trial, was given previously (13). This committee did recommend early termination of the trial based on evidence that the criteria for satisfying the stopping rules had been met (13). All patients were included in analysis of the primary end point according to the intention-to-treat principle. The primary comparison of the treatment groups was based on a log rank test. Univariate Cox regression (which included only treatment in the model) was performed for the time to the first primary event to obtain the point estimate of the hazard ratio (HR) and its 95% confidence interval between the 2 treatment arms. Individual analyses were also performed for each component of the primary end point, without censoring for previous primary events. Secondary and other efficacy end points were analyzed with the use of a similar log-rank test and univariate Cox regression analyses. Comparisons among patient groups of clinical baseline and demographic data were performed using ttests for continuous variables and chi-square tests for categorical variables.
A total of 6,946 patients with hypertension and diabetes entered the trial; 3,478 were randomly assigned to the B+A group and 3,468 to the B+H group. There were 2,842 diabetic patients at very high risk (histories of cardiac, stroke, or renal events), of whom 1,432 were randomized to B+A and 1,410 to B+H. There were 4,559 patients without diabetes at baseline, of whom 2,266 were randomized to B+A and 2,293 to B+H. There were no meaningful differences in baseline clinical characteristics between the B+A and B+H treatment arms within any of the 3 patient groups. The clinical characteristics of the study patients are shown in Table 1.
For the full diabetes cohort, the mean duration of treatment in the B+A group was 29.7 and 29.5 months in the B+H group. Details of the final treatment regimens and drug doses in the diabetic patients are shown in Table 2.For the B+A group, the mean doses were 36.6 mg for benazepril and 7.9 mg for amlodipine; for the B+H group, the mean doses were 36.6 and 20.0 mg, respectively. The use of additional antihypertensive agents, as shown in Table 2, was similar in the 2 groups. Of the patients in this trial given additional drugs, selected at the discretion of the investigators, 65.2% received beta-blockers, 42.2% loop diuretics, 34.7% alpha-blockers, and 24.7% clonidine. The need for additional antidiabetes therapy during the trial also was similar in the 2 groups.
Almost all patients (97%) were already receiving antihypertensive drugs before entering the study. Their blood pressures while receiving this previous treatment are shown in Table 1and were similar for the 2 treatment arms. Similarly, the effects of the study drugs on blood pressure for the 2 treatment arms in the diabetic patients were similar throughout the trial and are shown in Figure 1.The mean value of blood pressures after the treatment adjustments were completed were 131.5/72.6 mm Hg in the B+A arm and 132.7/73.7 mm Hg in the B+H arm. The mean difference between the 2 arms across the study was −1.2/−1.1 mm Hg. The attainment of controlled blood pressure (<130/80 mm Hg in these diabetic patients, which was recommended to investigators but not mandated) was 45.9% in the B+A arm and 43.7% in the B+H arm.
A subgroup of 353 patients in the diabetes group underwent ambulatory blood pressure monitoring after 2 years of study treatment. The data for systolic blood pressure are shown in Table 3and indicate that there were no significant differences in 24-h, daytime or nighttime values between the B+A and B+H treatment arms.
Study end points
The overall ACCOMPLISH trial was stopped early when the independent data safety monitoring committee observed a difference in the incidence of the primary end point between the 2 study arms that exceeded the pre-specified stopping point and recommended termination of the trial. Details of this decision were described previously (13).
For the diabetic patients as a whole, the time to the first primary end point in each treatment arm is shown in Figure 2.This end point occurred in 307 (8.8%) patients in the B+A group and 383 patients (11.0%) in the B+H group, representing an absolute risk reduction of 2.2% and a hazard reduction of 21% (HR: 0.79; p = 0.003). The number needed to treat for 30 months (to save 1 primary end point) was 46. Data for other key end points for diabetic patients in this trial are shown in Table 4.
High-risk diabetic patients
Because of the frequency of prevalent cardiovascular and renal disease in diabetic patients, further analyses were performed in those hypertensive diabetic patients who had histories of coronary disease, revascularization, stroke, peripheral arterial disease, or left ventricular hypertrophy (13).
Of the 2,842 such patients, 1,432 were randomized to B+A treatment and 1,410 to B+H treatment. The baseline characteristics of these patients are shown in Table 1. For the high-risk diabetic patients randomized to B+A, the achieved mean doses of these agents were 36.6 mg and 7.8 mg, respectively, and 98% received at least 1 additional antihypertensive agent. For the B+H arm, the mean doses were 36.4 mg and 19.9 mg, respectively. The effects of treatment on blood pressure in these 2 study arms are shown in Figure 1. The mean achieved blood pressures in the B+A arm and B+H arm were 131.2/71.9 mm Hg and 132.4/73.3 mm Hg, respectively; the mean difference was −1.2/−1.4 mm Hg.
The time to the first primary end points in the 2 treatment arms is shown in Figure 2. There were 195 events (13.6%) in the B+A arm and 244 events (17.3%) in the B+H group, representing an absolute risk reduction of 3.7% and a hazard reduction of 23% (HR: 0.77, p = 0.007). The number needed to treat for 30 months to prevent 1 primary end point was 28. Although this number needed to treat in the high-risk diabetic patients was lower than that for the non–high-risk diabetic patients and the nondiabetic patients, the interactions for treatment effect between these groups was not statistically significant. The other end points of interest are shown in Table 5.
Of the 4,559 nondiabetic patients; 2,266 were randomized to B+A and 2,292 to B+H. Baseline characteristics are shown in Table 1. Blood pressure data are shown in Figure 1; the achieved mean blood pressure values in the 2 treatment arms were 131.8/74.3 and 132.3/75.4 mm Hg, respectively. The end point data for the 2 treatment arms are shown in Figure 2(primary end point) and Table 6(all end points).
The analysis reported here indicates that in diabetic patients with hypertension the combination of an ACE inhibitor with amlodipine is superior to the combination of the same ACE inhibitor with hydrochlorothiazide in reducing cardiovascular outcomes. We observed that the size of this effect is similar to that in nondiabetic patients.
A further analysis in patients classified as those with high-risk diabetes, defined as patients with histories of cardiovascular, stroke, and renal events in addition to their diabetes, demonstrated a similar advantage for the combination of amlodipine with blockade of the RAS in reducing the primary study end point. In fact, the number needed to treat (to prevent 1 primary event over 30 months in the amlodipine arm compared with the hydrochlorothiazide arm), which was 48 in the nondiabetic patients and 46 in the overall diabetic cohort, was 28 in the high-risk diabetic patients. Tests for the interactions between high-risk diabetic patients and either the non–high-risk diabetic patients or the nondiabetic patients with treatment effect were not statistically significant. It was shown previously that high-risk diabetic patients have an exaggerated probability of cardiovascular events (15), so it is useful to note that the benefit of the B+A treatment in these patients was at least as strong as in the other study groups.
It should be noted that there were baseline differences, other than the presence or absence of diabetes, between the diabetic and nondiabetic patients in the ACCOMPLISH trial. The diabetic cohort included a higher proportion of women and black patients than the nondiabetic cohort; the diabetic patients also had a lower prevalence of previous coronary and stroke events and thus were less likely to be receiving therapies such as lipid-lowering agents, beta-blockers, and antiplatelet drugs. However, as would be expected, the high-risk diabetic subgroup had the same previous event profiles and adjunctive therapies as the nondiabetic patients. As well, blood pressures and antihypertensive therapies in the diabetic and nondiabetic groups were very similar to each other at baseline.
Secondary end points
The ACCOMPLISH trial used a broad composite primary end point that incorporated fatal and nonfatal cardiovascular and stroke end points, including coronary revascularization. The trial was terminated early after an interim data analysis revealed that there was a difference in event rates for the primary end point between the B+A and B+H treatment arms. This early termination of the trial limited its power to test fully the differing effects, if any, of the 2 treatment regimens on some of the important individual end points recorded in the trial.
Nevertheless, there were some noteworthy findings for secondary end points. In the total diabetes cohort, there were significantly fewer coronary outcomes in the B+A arm. This applied to both acute clinical events (myocardial infarction, sudden cardiac death, and hospitalized unstable angina), which were reduced by 27%, and coronary revascularization procedures, which were reduced by 20%. On the other hand, heart failure rates were not different between the B+A and B+H treatment arms in either the diabetic or nondiabetic cohorts. Previously, it had been claimed that amlodipine, used as a monotherapy, may not be an optimal therapy for heart failure prevention (8), but this does not appear to be an issue when this drug is combined with a blocker of the RAS.
Because virtually all patients were receiving blockers of the RAS before entering the study, the creatinine-increasing effects of these agents were already incorporated into their baseline values, so that any further increases likely reflected intrinsic changes in renal function. We performed a post-hoc analysis based on the criteria of serum creatinine concentrations that increased by at least 50% during the trial (regardless of their baseline values) and were also above the laboratory-defined normal range. This clinically relevant index of renal disease progression was 47% lower in the B+A than the B+H arm in the diabetic cohort. It was similarly lower in the high-risk diabetic subgroup (38%) and in the nondiabetic patients (62%). These findings, although significant and clinically informative, were not produced by a pre-specified trial analysis and so should be regarded as exploratory rather than definitive. Still, they deserve further study in future trials.
Control of blood pressure
In designing this trial, hydrochlorothiazide was selected as the diuretic in the comparator combination because it is overwhelmingly the most widely used thiazide agent in clinical practice. It could be noted, however, that chlorthalidone was the diuretic used in some previous hypertension trials (7,8) and is a more potent and long-acting agent than hydrochlorothiazide (16). Even so, in the doses used in this trial, hydrochlorothiazide combined with an ACE inhibitor provided powerful blood pressure control that was similar to that observed in the ACE inhibitor–amlodipine treatment arm. Furthermore, there was no difference between the 2 treatment arms in the number of dose titrations or add-on drugs required to achieve these blood pressures.
These findings were confirmed by an ambulatory blood-pressure monitoring substudy performed after 2 years of study treatment. The hydrochlorothiazide combination was at least as efficacious in reducing systolic blood pressure as the amlodipine combination, both during the nighttime as well as the daytime hours, thus indicating similar durations of action in the 2 treatment arms.
Metabolic and safety findings in this trial reflected the known properties of the drugs. There was no difference between the treatment groups in ACE inhibitor–related cough, but there was a greater incidence of peripheral edema in the amlodipine combination and a slightly greater incidence of hypokalemia in the hydrochlorothiazide combination. Despite these apparently minimal differences, we cannot exclude the possibility that some as-yet unidentified effect of the thiazide may have made it less effective in reducing clinical end points than the calcium channel blocker.
There may be inadequate patient numbers in subgroups of clinical trials to allow adequate analysis of outcomes, but in the present report the cohort of patients with diabetes was sufficiently large to explore the relevant events. It should be noted, however, that the analysis of the high-risk diabetic patients was not pre-specified, but was undertaken to test whether the outcomes in these vulnerable patients were similar to the full diabetes group.
Our use of 2 types of composite coronary end points, namely nonrevascularization events (sudden cardiac death, myocardial infarction, and unstable angina) or revascularization procedures, was also not pre-specified. Our intention in establishing these categories was to examine the rate of hard clinical end points independently of the more elective coronary interventions. Likewise, as discussed earlier, the use of the renal end point (increase in creatinine) was not originally planned. Still, we believe that this exploratory finding is of clinical interest and serves to support the observation—shown in our table of metabolic and renal measurements—of the divergent effects of amlodipine and hydrochlorothiazide therapies on renal function and albuminuria.
The combination of an ACE inhibitor with amlodipine, compared with its combination with hydrochlorothiazide, was more effective in preventing fatal and nonfatal cardiovascular outcomes in hypertensive diabetic patients. This finding was particularly evident in a subgroup of high-risk diabetic patients with histories of cardiovascular, stroke, and renal events and appeared to be independent of hemodynamic effects. Patients receiving the amlodipine combination were more likely to experience peripheral edema, but other safety and metabolic findings were not meaningfully different between the treatments. For the large proportion of diabetic patients whose blood pressure can be controlled with a 2-drug combination, the use of amlodipine with a blocker of the RAS should now be considered.
This study was supported financially by Novartis, who also provided logistical and statistical support and expertise (based on the analysis plans of the academic authors). Dr. Weber receives lecture fees from Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi Sankyo, Forest Pharmaceuticals, GlaxoSmithKline, Novartis, and Sanofi-Aventis; and consultant fees from Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi Sankyo, Forest Pharmaceuticals, GlaxoSmithKline, Gilead, Novartis, and Takeda Pharmaceuticals. Dr. Bakris receives consulting fees from Merck, Takeda, Boehringer-Ingelheim, Abbott Laboratories, Walgreen's, BMS/Sanofi, Gilead, Novartis, Forest Pharmaceuticals, GlaxoSmithKline, and Daiichi Sankyo; lecture fees from Novartis and Forest Pharmaceuticals; and grant support from GlaxoSmithKline, Juvenile Diabetes Research Foundation, Forest Labs, and CVRx. Dr. Jamerson receives consulting fees from Novartis, Merck, and Daiichi Sankyo; lecture fees from Novartis, Abbott Laboratories, Bristol-Myers Squibb, GlaxoSmithKline, and Merck; and research support from NIH, NIDDK, NHLBI, Novartis, and King Pharmaceuticals. Dr. Weir serves as scientific advisor for Amgen, Nicox, Novartis, Boehringer Ingelheim, Sanofi, Bristol-Myers Squibb, Daiichi Sankyo, and Johnson & Johnson. Dr. Kjeldsen receives lecture fees from Novartis, AstraZeneca, Bayer, Boehringer-Ingelheim, Menarini, Merck, Sankyo, Sanofi, Servier, and Takeda. Dr. Devereux serves on an advisory board for Merck and has received lecture fees from Merck within the past 2 years; during the same time period, he has served as a consultant for Novartis, Sanofi-Aventis, Novo Nordisk Pharmaceuticals, and Neurosearch. Dr. Velazquez receives consulting and lecture fees from Novartis, Cardio-Kinetix, NovaCardia, and Medtronic Foundation. Dr. Dahlöf receives consulting fees from Novartis, Daiichi Sankyo, and Boehringer Ingelheim and lecture fees from Novartis, Boehringer Ingelheim, Pfizer, and Merck. Roxzana Kelly, Tsushung Hua, and Allen Hester are employed by Novartis Pharmaceuticals. Dr. Pitt receives consulting fees from Novartis, Pfizer, Merck, Takeda, Bayer, Forest Laboratories, Sankyo, Ono, and AstraZeneca; has stock options from Relypsa, BG-Medicine, and Nile Therapeutics; and has received grants from Medtronic, Bayer, Novartis, and Abbott Laboratories.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- benazepril plus amlodipine
- benazepril plus hydrochlorothiazide
- hazard ratio
- renin-angiotensin system
- Received November 11, 2009.
- Revision received February 9, 2010.
- Accepted February 15, 2010.
- American College of Cardiology Foundation
- Turnbull F.,
- Neal B.,
- Algert C.,
- et al.
- The ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group
- Dahlof B.,
- Sever P.S.,
- Poulter N.R.,
- et al.,
- ASCOT Investigators
- Neutel J.M.,
- Smith D.H.,
- Weber M.A.
- Weber M.A.,
- Bakris G.L.,
- Dahlof B.,
- et al.