Analysis of Recent Papers in Hypertension

DIFFERENCES IN CENTRAL AORTIC BLOOD PRESSURE EXPLAIN OUTCOME DIFFERENCES IN ASCOT: THE CAFE SUBSTUDY

Hypertension continues to be a major cause of stroke, coronary heart disease, heart failure, end‐stage renal disease, and death. Clinical trials in those with hypertension have shown that effective control of blood pressure (BP) results in an impressive reduction in cardiovascular (CV) morbidity and mortality. Most studies have utilized peripheral (brachial) BP to evaluate BP reduction in clinical hypertension trials, assuming that BP measured in the arm represents BP measured in the central aortic circulation. In fact, drugs with different mechanisms of action that have the same effect on brachial BP could have different effects on central aortic pressure, which may also be a more important determinant of clinical outcomes.

To investigate the significance of these issues, a prespecified substudy of the BP‐lowering arm of the Anglo‐Scandinavian Cardiac Outcomes Trial (ASCOT‐BPLA) ¹ was performed. The Conduit Artery Function Evaluation (CAFE) sub‐study recruited 2199 participants from five of the ASCOT‐BPLA (hereinafter referred to as ASCOT) centers located in the United Kingdom and Ireland from the original 19,257 patients with hypertension recruited into ASCOT. Patients volunteered for CAFE and were eligible for participation in this industry‐sponsored, independent, investigator‐initiated, investigator‐designed, and investigator‐led study if they were aged 40–79 years (mean age, 63 years) and free of coronary heart disease but had hypertension and at least three additional risk factors. Eligible patients (85% white, 79% male, body mass index of 29 kg/m²) had to have a baseline untreated systolic BP ≥160 mm Hg or diastolic BP ≥100 mm Hg or treated systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg. Ninety percent were on previous antihypertensive treatment. The additional risk factors included: male gender, smoker, age older than 55 years, left ventricular hypertrophy, electrocardiographic changes consistent with evidence of ischemia, type 2 diabetes, peripheral arterial disease, cerebrovascular disease, microalbuminuria or proteinuria, a ratio of plasma total cholesterol to high‐density lipoprotein cholesterol (HDL‐C) of six or more, or a family history of premature coronary heart disease. Exclusion criteria included previous myocardial infarction, treated angina at the time of randomization, a cerebrovascular event in the 3 months before randomization, fasting triglycerides >400 mg/dL, heart failure, uncontrolled arrhythmias, or any clinically important hematologic or biochemical abnormality on routine screening. Eligible patients were randomized using a prospective, randomized, open, blinded end point (PROBE) design and were randomized using a prespecified six‐step treatment algorithm to either the calcium channel blocker amlodipine, 5 mg initially, titrated up to a maximum of 10 mg; or to the β blocker atenolol, 50 mg initially, titrated up to 100 mg, both given once daily. If the patient was not controlled to the target BP of <140/90 mm Hg in those without diabetes and <130/80 mm Hg for those with diabetes, the angiotensin‐converting enzyme inhibitor perindopril, 4 mg initially, titrated up to a maximum of 8 mg daily, was added to the amlodipine‐based regimen, or the thiazide‐type diuretic bendroflumethiazide + potassium started at 1.25 mg, titrated up to 2.5 mg, was added to the atenolol‐based regimen. The α₁ antagonist doxazosin, 4 mg titrated up to 8 mg, was added, if necessary, as a third agent to each arm of the trial.

Recruitment into CAFE began in 2001, 1 year after randomization into ASCOT, when treatment regimens were stable. This was done to avoid the early BP fluctuations that often occur with initial up‐titration of the antihypertensive regimen. Recruitment continued over the remaining 4 years of ASCOT. Approximately 70% of ASCOT patients at each CAFE study center were recruited. Follow‐up visits took place at 6‐month intervals throughout the study.

The study used radial artery applanation tonometry and pulse wave analysis to calculate derived central BP and other parameters using the commercially available SphygmoCor system (SphygmoCor V7; Atcor Medical, West Ryde, New South Wales, Australia). All measurements were obtained by trained research nurses, technicians, or study physicians who maintained quality control. Measurements were obtained at scheduled ASCOT follow‐up visits. The objective was to obtain at least two measurements for each participant in the CAFE study. In fact, an average of 3.4 measurements per patient was recorded, and the number of measurements did not differ by treatment arm (3.3 measurements in the atenolol‐based arm; 3.5 measurements in the amlodipine‐based arm). Only 22% of patients had just one measurement. The mean follow‐up time after the initial tonometry measurement was 3 years (2.9 years in the atenolol‐based arm; 3 years in the amlodipine‐based arm). Within the first year, 36% of the CAFE cohort had undergone at least one CAFE study measurement, which increased to 67% by Year 2 and to 87% by Year 3. The CAFE applanation tonometry database was compiled by researchers blinded to treatment allocation, patient demographics, and clinical outcomes.

The primary objective of the CAFE study was a comparison of the two treatment regimens by central aortic pressures derived from applanation tonometry. As a secondary objective, the study investigators prespecified that they would examine the relationship between central aortic pressure and the primary and secondary outcomes of ASCOT. Post hoc, they defined a composite clinical outcome comprising all CV events and procedures and development of renal impairment for events as defined and validated by the ASCOT end points committee. Three Cox proportional‐hazards models were constructed to evaluate whether central aortic pressures/hemodynamic indexes measured during the CAFE study follow‐up were associated with the composite clinical outcome.

Before data analyses, 126 patients of the 2199 recruited into the study were excluded because their radial arterial pressure waveforms were of insufficient quality because of abnormal heart rhythms or low‐amplitude pulses. The remaining 2073 patients, well matched between the two treatment arms for baseline characteristics, were included in the intention‐to‐treat analysis. With a similar baseline seated brachial BP of 160/93 mm Hg in the two treatment arms, which was slightly lower than in the total ASCOT population, brachial BP fell −26/−13.8 mm Hg for the atenolol‐based vs. −27.8/−15.7 mm Hg for the amlodipine‐based arms, respectively. Most patients (95%) were taking at least two BP‐lowering drugs, with 56% and 60% receiving the predefined combination therapy of amlodipine ± perindopril or atenolol ± thiazide, respectively. Only 3.5% of atenolol and 7.0% of amlodipine patients remained on monotherapy throughout the study.

Brachial systolic pressures and derived central aortic pressures as well as the BP load for each treatment arm presented as the mean area under the curve (AUC) were analyzed. Despite insignificant differences in brachial systolic Bps throughout the CAFE study (AUC difference, 0.7 mm Hg; 95% confidence interval [CI], −0.4 to 1.7 mm Hg; p=0.2), derived central aortic systolic Bps were substantially lower with amlodipine‐based therapy compared with the atenolol‐based regimen (AUC difference, 4.3 mm Hg; 95% CI, 3.3–5.4 mm Hg; p<0.0001) with similar but smaller differences in central aortic diastolic Bps (AUC difference, 1.4 mm Hg; 95% CI, 0.6–2.1 mm Hg; p<0.001), suggesting that an important difference between the two treatment regimens was the impact on central aortic systolic pressure. Central aortic pulse pressure was also lower throughout the CAFE study in the amlodipine ± perindopril‐based therapeutic arm compared with atenolol ± thiazide‐based therapy (AUC difference, 3.0 mm Hg; 95% CI, 2.1–3.9 mm Hg; p<0.0001) despite a slightly higher brachial pulse pressure in the amlodipine‐based arm. Although the CAFE study was predominantly male, similar differential effects of the two treatment arms on central pressures were observed in women, with no interaction on central pressure seen based on gender, age, or treatment arm. The atenolol‐based regimen was associated with an increase in both central aortic systolic pressure wave augmentation (AUC difference, 3.8 mm Hg; 95% CI, 3.3–4.4 mm Hg; p<0.0001) and the augmentation index, the percentage that the systolic pressure wave is attributable to wave reflection (AUC difference, 6.5%; 95% CI, 5.8–7.3%; p<0.0001).

The predefined secondary analysis used Cox regression modeling to determine whether central hemodynamic parameters measured in the CAFE cohort were associated with clinical outcomes. All three constructed Cox regression models revealed that central aortic pulse pressure was significantly associated with hazard for the post hoc‐defined clinical composite end point of total CV events and procedures and the development of renal impairment both before and after adjustment for age and baseline risk factors (unadjusted p<0.0001; p<0.05 after adjustment for baseline variables).

CAFE is the largest prospective evaluation of the effects of CV drugs on derived central aortic pressures and hemodynamic indices. The results in 2073 ASCOT participants followed for up to 4 years showed substantial and consistent differences in central aortic pressures and hemodynamics in favor of the amlodipine ± perindopril‐based therapy vs. atenolol ± thiazide‐based therapy despite conventional brachial cuff measurement showing similar systolic BP between treatment arms. Brachial BP is not always a good surrogate for the effects of antihypertensive drugs on arterial hemodynamics. With the atenolol ± thiazide‐based treatment, much less effective at lowering central aortic pressures, the authors believe that the CAFE study provides a plausible mechanism to explain, at least in part, the more favorable clinical outcome for patients treated with amlodipine ± perindoprilbased therapy in ASCOT‐BPLA. How this translates into differences found in other BP‐lowering treatment trials and whether the technology used in this trial will be incorporated into other clinical hypertension trials to better understand the differing mechanistic benefits of antihypertensive drug therapy remains unclear. These authors believe their findings suggest a mechanism to support the recent meta‐analysis challenging the recommendation that β‐blocker therapy be used as initial therapy for uncomplicated hypertension.—Williams B, Lacy PS, Thom SM, et al., for the CAFE Investigators for the Anglo‐Scandinavian Cardiac Outcomes Trial (ASCOT) Investigators; CAFE Steering Committee and Writing Committee. Differential impact of blood pressure‐lowering drugs on central aortic pressure and clinical outcomes: principal results of the Conduit Artery Function Evaluation (CAFE) study. Circulation. 2006;113:1213–1225.

COMMENT

Most experts believe it is the achieved reduction in BP, regardless of the antihypertensive agent selected, that translates into long‐term clinical outcome improvement in CV events. In the original ASCOT‐BPLA study, however, patients treated with an amlodipine‐based regimen had a 16% relative risk reduction in the incidence of total CV events and procedures compared with an atenolol‐based regimen. Debate occurred shortly after the ASCOT publication over whether the benefits of the amlodipine‐based regimen was attributed to superior BP reduction, as measured by conventional brachial cuff measurement, or to effects beyond the noted differences in BP measurement. The CAFE study, by using radial applanation tonometry and pulse‐wave analysis to generate computer‐derived central aortic pressures as well as other hemodynamic parameters, is the first large clinical outcome trial to evaluate the impact of two different BP‐lowering regimens utilizing these noninvasive measurements. CAFE found that the β blocker atenolol as initial antihypertensive therapy does not lower central systolic BP as much as an amlodipine‐based regimen. This probably explains why, in the Losartan Intervention for Endpoint Reduction in Hypertension (LIFE) trial, despite equal BP reduction, a losartan‐based regimen reduced left ventricular hypertrophy more than an atenolol‐based regimen. Differences seen between measurements in central aortic BP and brachial BP probably have to do with clinically relevant changes that occur in reflected waves in the central arterial tree that are not detected with conventional BP measurement. Furthermore, it is the forward pressure wave amplification (augmentation index), not detected on brachial cuff measurement, that was substantially higher with atenolol‐based therapy and may be responsible for the clinical outcomes seen.

The CAFE study provides a novel mechanism to explain why certain antihypertensive agents lead to outcome improvement despite similar BP reduction determined by standard measurement technique. Whether noninvasive techniques such as radial tonometry should be used more often in clinical medicine has not been adequately determined. While β‐blocker therapy has been considered appropriate initial therapy in patients with hypertension, and atenolol remains one of the most widely used β blockers within that class, the CAFE substudy provides a unique mechanism supporting the recent meta‐analysis in those with hypertension that found β‐blocker therapy no better than placebo for the prevention of CV disease. Whether this applies to all agents within the β‐blocker class needs to be evaluated in future trials.

RIMONABANT IMPROVES MULTIPLE CARDIOMETABOLIC PARAMETERS IN OVERWEIGHT AND OBESE INDIVIDUALS

Through activation of the cannabinoid‐1 (CB1) receptor, the endocannabinoid system (ECS) represents a novel potential target of intervention in patients with obesity and the metabolic syndrome. Rimonabant represents the first specific CB1 receptor blocker to enter clinical development. Previous reports suggest that rimonabant enhances weight loss and improves metabolic risk factors in subjects with type 2 diabetes mellitus and cardiovascular disease. Two phase 3 prospective randomized clinical trials comparing rimonabant to placebo in overweight patients have recently been published.

In the Rimonabant in Obesity‐Lipids Study (RIO‐LIPIDS), 1036 overweight or obese subjects with untreated dyslipidemia were randomized in double‐blind fashion to placebo or rimonabant 5 or 20 mg daily for 12 months, in addition to a hypocaloric diet. Rates of study completion were approximately 60% in each group, and all analyses were by “intention to treat.” As compared with placebo, 20 mg of rimonabant led to significantly greater weight loss (−6.7 kg), reduction in waist circumference (−5.8 cm), increase in high‐density lipoprotein cholesterol (HDL‐C) (+10.6%), and decrease in triglycerides (−13.0%). There was no significant difference in low‐density lipoprotein cholesterol (LDL‐C), but rimonabant 20 mg was also associated with significant reductions in LDL‐C particle size, total cholesterol/HDL‐C, LDL‐C/HDL‐C, apoprotein B/apoprotein A‐I, fasting insulin, leptin, C‐reactive protein, and systolic and diastolic blood pressure. Adiponectin levels were increased with rimonabant 20 mg as compared with placebo, and this increase was independent of weight loss. There were no deaths in any of the three groups, but treatment‐related adverse events (mostly depression, anxiety, and nausea) were more common with rimonabant therapy than with placebo.

In the Rimonabant‐North America Study (RIO‐North America), 3045 obese subjects (mean age, 45 years; 80% women; body mass index, 38 kg/m²) with treated or untreated dyslipidemia or hypertension were randomized in double‐blind fashion to one of three therapies: placebo (607 subjects), 5 mg rimonabant (1216 subjects), or 20 mg rimonabant (1222 subjects) daily for 1 year in addition to a hypocaloric diet (deficit, 600 kcal/d). After Year 1, rimonabant‐treated patients were rerandomized to rimonabant or to placebo for an additional year. At the end of 1 year, completion rates were approximately 50%–55% in all three groups, with about 40% of subjects nonadherent and 7% dropouts. There was no significant difference between groups in adherence, and all results were analyzed by intention to treat.

As compared with placebo, 1 year of 20 mg of rimonabant led to significantly greater weight loss (−1.6 kg vs. −6.1 kg), reduction in waist circumference (−2.5 cm vs. −6.1 cm), decrease in triglycerides (+7.9% vs. −5.3%), and increase in HDL‐C (+5.4% vs. +12.6%), placebo vs. rimonabant, respectively. The percentage of subjects achieving a ≥5% decrease in weight at Year 1 was 20.0% for placebo and 48.5% for rimonabant 20 mg. In subjects randomized to 20 mg of rimonabant, the prevalence of the metabolic syndrome (as defined by National Cholesterol Education Program Adult Treatment Panel III [ATP III] guidelines) declined from 34.8% at baseline to 21.2% at Year 1. Insulin resistance (measured by homeostasis model assessment of insulin resistance) increased in subjects taking placebo, but not in those taking rimonabant 20 mg. Systolic and diastolic blood pressure decreased slightly, but not significantly, in subjects randomized to 20 mg of rimonabant. Subjects who were switched from 20 mg of rimonabant to placebo for Year 2 experienced a significant weight regain, while those who remained on rimonabant 20 mg for Year 2 maintained their weight loss and favorable changes in metabolic parameters. Compared with subjects on placebo, the incidence of adverse events leading to withdrawal was higher with rimonabant than placebo: mainly psychiatric (depression, anxiety, and irritability), nervous system, and gastrointestinal (nausea)‐related events.

In summary, these two well designed and conducted clinical trials found that rimonabant, at a dose of 20 mg daily, positively affected all significant components of the metabolic syndrome in overweight and obese individuals. The most common side effects were depression, anxiety, and nausea, each of which was seen in >5% of subjects randomized to rimonabant 20 mg. Given the high nonadherence rate (almost half of the participants studied), long‐term tolerability of rimonabant may be an important clinical issue.—Després JP, Golay A, Sjöström L, for the Rimonabant in Obesity‐Lipids Study Group. Effects of rimonabant on metabolic risk factors in overweight patients with dyslipidemia. N Engl J Med. 2005;353:2121–2134.—Pi‐Sunyer FX, Aronne LJ, Heshmati HM, et al., for the RIO‐North America Study Group. Effect of rimonabant, a cannabinoid‐1 receptor blocker, on weight and cardiometabolic risk factors in overweight or obese patients: RIO‐North America: a randomized controlled trial. JAMA. 2006;295:761–775.

COMMENT

Activation of the ECS appears to be intimately involved in worsening multiple cardiometabolic risk factors in overweight and obese individuals. CB1 receptors are found in several areas of the nervous system and in the gut, liver, muscle, and adipocytes. In the brain, activation of CB1 (mostly in the hypothalamus and limbic system) appears to provoke food intake even in the setting of satiety (thus providing a potential explanation for the “munchies” in persons who consume exogenous cannabinoids like marijuana). In fat cells, activation of CB1 promotes lipogenesis and inhibits production of the beneficial hormone adiponectin. In animal models, it appears that ECS activation acts as a positive feedback loop, whereby chronic overfeeding leads to greater activation of the ECS and subsequent further overeating, weight gain, and worsening of cardiometabolic risk factors. Interestingly, many of the effects on cardiometabolic risk factors seen with ECS activation, including effects on adiponectin levels, appear to be somewhat independent of weight, highlighting the important peripheral effects of ECS activation.

Most previous attempts to treat cardiometabolic risk have approached each risk factor individually. While none of the individual beneficial metabolic effects reported here is unique or of greater magnitude than can be seen with other classes of medications, rimonabant at a dose of 20 mg daily appears to offer a unique ability to target multiple cardiovascular risk factors at once. While there was little reduction observed in systolic and diastolic blood pressure, it will be of interest to see whether the overall effects seen with rimonabant will prove to be additive when combined with other interventions used to treat these metabolic parameters, including lipid lowering, oral hypoglycemic, and antihypertensive medications in our patients with hypertension. Although the overall incidence of the metabolic syndrome decreased with rimonabant therapy in RIO‐North America, these trials did demonstrate some significant potential limitations to this therapy. The relatively low completion rates in both trials are similar to what has been seen in other weight‐loss trials and were not different between treatment groups; however, a significant number of subjects randomized to rimonabant 20 mg did experience important psychiatric and gastrointestinal side effects, which may be a significant limitation to its use in clinical practice. The fact that in RIO‐North America subjects who discontinued rimonabant after 1 year of therapy regained significant weight suggests that rimonabant therapy may need to be continued long‐term to see maintenance of weight loss and improvements in cardiometabolic risk factors. For now, regardless of how rimonabant is used in clinical practice, long‐term, comprehensive lifestyle changes—including a reduced‐calorie, low‐fat diet as well as increased physical activity—remain critical components for combating obesity in patients with hypertension.