Predictors of systolic BP <140 mmHg and systolic BP level by randomly assigned treatment group (benazepril plus amlodipine or hydrochlorothiazide) in the ACCOMPLISH Study

Abstract

Background. The ACCOMPLISH Trial investigated intensive antihypertensive combination treatment with benazepril + amlodipine (B+A) or benazepril + hydrochlorothiazide (B+H) on cardiovascular outcomes in patients with systolic hypertension. We analyzed the baseline predictors of achieving a systolic blood pressure (SBP) <140 mmHg and achieved SBP level by the end of 12 months in both treatment groups. Methods. Baseline and 12-month SBP was available in 10,506 patients, of whom 6250 had diabetes. Univariate and multivariate logistic regression models were used for SBP control at 12 months and multivariable regression models were used for the prediction of SBP at 12 months. A stepwise procedure was used to select significant (p < 0.001) predictors in multivariate analyses. Results. Mean (± SD) BP fell from 145.4/80.1 (± 18.3/10.7) mmHg at randomization to 132.8/74.7 (± 16.0/9.6) mmHg at 12 months. The main baseline predictors of SBP control <140 mmHg were region (USA >Nordic region) and Caucasian ethnicity in both randomization arms. A higher diastolic BP and the use of lipid lowering agents indicated favorable effects in the B+H arm only. The predictors of uncontrolled SBP were: (i) higher baseline SBP values, (ii) higher number of previous antihypertensive medications in both arms, (iii) the previous use of insulin in the B+A arm, and (iv) pre-trial calcium channel blocker (CCB) use in the B+H arm. Additionally, pre-trial use of thiazides and electrocardiogram (ECG)-left ventricular hypertrophy (LVH) at baseline predicted higher, and smoking lower absolute SBP in the B+A arm and the use of thiazides and proteinuria a higher SBP in the B+H arm. Conclusion. Irrespective of treatment, patients in the USA and Caucasians achieved better SBP control, whereas higher baseline SBP and more previous antihypertensive medications indicated less control. Concomitant use of lipid lowering treatment indicated a better SBP control in the benazepril + hydrochlorothiazide arm. Lastly, insulin use and ECG-LVH in the benazepril + amlodipine arm and proteinuria in the benazepril + hydrochlorothiazide arm indicated poor control.

Key Words::

• clinical trial
• combination therapy
• hypertension
• predictor variables

Introduction

The ACCOMPLISH (Avoiding Cardiovascular Events Through Combination Therapy in Patients Living with Systolic Hypertension) study was designed to compare the efficacy of two types of commonly used antihypertensive drug combinations in preventing major clinical outcomes in hypertensive patients at high risk for cardiovascular events. ACCOMPLISH showed that the ACE inhibitor (benazepril) plus calcium channel blocker (amlodipine; CCB) (B+A) combination was more effective than the benazepril plus hydrochlorothiazide (B+H) combination in reducing major cardiovascular outcomes (1).

Of the patients enrolled, 97% were previously treated for hypertension, though only 37% of them had blood pressure (BP) adequately controlled (<140/90 mmHg) when entered the study. Within 12 months after the study therapy, 73% of patients achieved the BP goal of <140/90 mmHg (2). Since this cohort achieved remarkable BP control early in the trial, it has been possible to relate the degree of BP control and achieved absolute BP to clinical and demographic characteristics of the enrolled patients, giving for the first time a unique opportunity to assess predictors of the BP control by initial combination therapy in hypertension. While ACCOMPLISH was still ongoing and drug treatment blinded, we extensively investigated predictors of the 12 months of BP responses (3). Now after study end we are able to do final analyses of the predictors of BP responses in ACCOMPLISH both at BP <140/90 and <130/80 mmHg, by treatment arms, and in the large (60% of all patents) subpopulation with diabetes. The aim of this paper is to report the final data on the multivariate, stepwise selected predictors of systolic BP (SBP) control <140 mmHg and achieved SBP level by treatment group at 12 months.

Methods

A full description of the ACCOMPLISH study design has been published previously (4). The following is a brief summary: ACCOMPLISH was a randomized, double blind trial that compared the efficacy of two fixed dose combination therapies, B+A, and B+H, in preventing fatal and non-fatal cardiovascular outcomes. Patients eligible for this study were >55 years with either SBP ≥160 mmHg or currently receiving antihypertensive therapy. Eligible patients had evidence for cardiovascular, renal disease or other target organ damage. Upon study entry, patients had their study medication force-titrated during the first 2 months of the trial to receive either B+A 40/5 mg or B+H 40/12.5 mg. In order to reach BP targets (SBP <140/90 mmHg in all and SBP <130/80 mmHg in patients with diabetes), the study medication would be increased to B+A 40/10 mg or B+H 40/25 mg. Additionally, other antihypertensive agents, excluding the drug classes involved in the trial, could be added in order to achieve target BP goals. Following the initial 3-month period, patients were seen at 6-month intervals until the end of the trial (4).

The data in this report are based on predictor-variable observations at baseline (the beginning of the study treatment period immediately prior to randomization). The total number of patients randomized in the trial (11,506 patients randomized to two treatment groups) reflects power calculations based on testing the principal study hypothesis for the primary cardiovascular mortality/morbidity endpoint; the details of this calculation have been published (4). A total of 10,400 patients of whom 6250 had diabetes at baseline have been analyzed. We chose the 12 months of BP data for the present analysis, since the majority of patients were appropriately titrated and the study medications were mostly stabilized by this time point.

Patient selection

The inclusion and exclusion criteria for hypertensive patients at high cardiovascular risk have been described previously (4). In general, ACCOMPLISH participants are hypertensive, aged 55 years or more, with either cardiovascular and/or renal disease, and were living in the USA or Nordic area (Sweden, Norway, Denmark or Finland). Patients whose age was 60 years or under and had evidence of two target organs damaged (heart, renal, vascular, diabetes) were recruited (4). Nordic-area review boards only allowed enrollment of patients with BP not currently controlled (not a requirement for USA enrollment), and thus the baseline BP control rate was lower for Nordic countries than for the USA (21.1% vs 44.3%, respectively, with SBP/DBP <140/90 mmHg). Accordingly, mean SBP and DBP were also higher at baseline for Nordic patients (152.6/84.4 mmHg) than for US patients (142.4/78.3 mmHg) (5). US black patients also had higher BPs than US Caucasian patients (145.1/81.3 vs 142.4/79.9 mmHg) at baseline.

Statistical analysis

All analyses are based on all patients with 12 months mean sitting BP data. Results are presented as mean ± SD for continuous or as n (%) for discrete variables; 1106 patients randomized did not attend the 12-month visit. Thirty-three potential baseline predictors of BP (demographic, risk and disease factors, baseline laboratory variables and prior antihypertensive medication) were identified and are listed in Table I. Univariate and multivariable logistic regression models were used for BP control at 12 months and simple and multiple linear regression models were used for BP value at 12 months as outcome. Stepwise forward and backward procedures was used to select significant (p < 0.001) predictors in multivariable analyses. Missing values of the baseline predictors were imputed and replaced by median values of the variables. The Statistical Analysis System (SAS Inc. Cary, North Carolina, USA) was used for all statistical analyses.

Table I. Potential baseline predictors of mean sitting systolic blood pressure (SBP) < 140 mmHg and achieved SBP at 12 months.

Region – USA Or Nordic countries Age (years) Gender Race (Caucasian or non-Caucasian) Hispanic ethnicity Body mass index (kg/m^2) Smoker Peripheral arterial occlusive disease Stroke Atrial fibrillation Diabetes mellitus Myocardial infarction or unstable angina Revascularization (CABG or PCI) Proteinuria Dyslipidemia (as determined by investigator) Glomerular filtration rate (GFR, ml/min/1.73 m^2) Chronic kidney disease as defined by GFR

Left ventricular hypertrophy by ECG Heart rate (beats/min) Mean sitting systolic blood pressure (mmHg) Mean sitting diastolic blood pressure (mmHg) Serum creatinine (μmol/l) Fasting glucose (mmol/l) Hemoglobin (g/l) Number of antihypertension medication classes (unit) Angiotensin converting enzyme inhibitors Angiotensin receptor blockers Thiazide Beta-blocker Lipid lowering agents Oral hypoglycemic agents and/or insulin sensitizer Insulin treatment Calcium channel blocker

CABG, coronary artery bypass graft; PCI, percutaneous coronary intervention; ECG, electrocardiogram.

Results

Mean BP fell from 145.4/80.1 (± 18.3/10.7) mmHg at randomization to 132.8/74.7 (16.0/9.6) mmHg at 12 months. Detailed comparisons of groups, assessments of univariate predictors and multiple regression analyses have been analyzed (data on file).

The main baseline predictors of SBP control <140 mmHg at 12 months in the multivariate analyses were the region (USA) and Caucasian ethnicity in both treatment arms, whereas a higher diastolic BP (DBP) and the use of lipid lowering agents indicated favorable effects in the B+H arm (Tables II and III). The predictors of uncontrolled SBP at 12 months were higher baseline SBP values and the pre-trial use of more antihypertensive drugs prior to the trial in both arms and also insulin use in the B+A arm (Table II) and pre-trial CCB use in the B+H arm (Table III).

Table II. Multivariate predictors (stepwise selection from full model) of 12 months mean sitting systolic blood pressure (SBP) <140 mmHg in the benazepril/amlodipine (B+A) treatment group.

Variable	Odds ratio	95% confidence intervals	Chi-squared value	p-value
Mean sitting systolic BP at baseline	0.963	(0.959–0.967)	344.01	<0.0001
Number of different antihypertensive medication classes at baseline	0.739	(0.695–0.785)	93.59	<0.0001
Region (USA)	1.843	(1.584–2.145)	62.62	<0.0001
Race (Caucasian)	1.457	(1.193–1.780)	13.61	0.0002
Insulin treatment	0.721	(0.597–0.870)	11.62	0.0007

Units indicated in Table I.

Table III. Multivariate predictors (stepwise selection from full model) of 12 months mean sitting systolic blood pressure (SBP) <140 mmHg in the benazepril/hydrochlorothiazide (B+H) treatment group.

Variable	Odds ratio	95% confidence interval	Chi-squared value	p-value
Mean sitting SBP at baseline	0.958	(0.954–0.962)	347.29	<0.0001
Region (USA)	2.115	(1.822–2.455)	97.12	<0.0001
Number of different antihypertensive medication classes at baseline	0.803	(0.752–0.858)	42.33	<0.0001
Calcium channel blocker	0.639	(0.552–0.741)	35.44	<0.0001
Mean sitting diastolic BP at baseline	1.018	(1.011–1.025)	23.15	<0.0001
Race (Caucasian)	1.474	(1.218–1.783)	15.89	<0.0001
Lipid lowering agents	1.331	(1.154–1.534)	15.50	<0.0001

Units indicated in Table I.

Additionally, pre-trial use of thiazides and ECG (electrocardiogram)-left ventricular hypertrophy (LVH) predicted higher achieved SBP and baseline DBP and smoking lower achieved SBP in the B+A arm (Table IV). Pre-trial use of thiazides and proteinuria (macroalbuminuria) predicted higher SBP in the B+H arm (Table V).

Table IV. Multivariate predictors (stepwise selection from full model) of mean sitting systolic blood pressure (SBP) at 12 months in the benazepril/amlodipine (B+A) treatment group.

Variable	Regression coefficient	95% confidence interval	F-statistic	p-value
Mean sitting systolic BP at baseline	0.331	(0.307–0.356)	677.46	<0.0001
Region (USA)	−5.737	(−6.636 to − 4.838)	156.49	<0.0001
Number of different antihypertensive medication classes at baseline	1.660	(1.274–2.045)	71.36	<0.0001
Mean sitting diastolic BP at baseline	−0.103	(−0.146 to − 0.060)	22.22	<0.0001
Race (Caucasian)	−2.645	(−3.749 to − 1.540)	22.02	<0.0001
Thiazide use	1.983	(1.126–2.840)	20.58	<0.0001
Left ventricular hypertrophy by ECG	2.564	(1.415–3.713)	19.14	<0.0001
Smoking	−2.017	(−3.202 to − 0.832)	11.13	0.0009

Units indicated in Table I. ECG, electrocardiogram.

Table V. Multivariate predictors (stepwise selection from full model) of mean sitting systolic blood pressure (SBP) at 12 months in the benazepril/hydrochlorothiazide (B+H) treatment group.

Variable	Regression coefficient	95% confidence interval	F-statistic	p-value
Mean sitting systolic BP at baseline	0.382	(0.355–0.408)	811.72	<0.0001
Region (USA)	−5.899	(−6.835 to −4.962)	152.58	<0.0001
Calcium channel blocker use	4.400	(3.448–5.352)	82.08	<0.0001
Mean sitting diastolic BP at baseline	−0.175	(−0.219 to −0.130)	59.55	<0.0001
Race (Caucasian)	−3.106	(−4.257 to −1.954)	27.96	<0.0001
Thiazide use	2.242	(1.299–3.185)	21.74	<0.0001
Proteinuria (macroalbuminuria)	3.918	(1.934–5.902)	14.99	0.0001
Number of different antihypertensive medication classes at baseline	0.903	(0.439–1.366)	14.56	0.0001

Units indicated in Table I.

Discussion

Irrespective of the randomized treatment regimen, patients in the USA and Caucasians vs other ethnic groups achieved better SBP control while those with higher SBP and more previous antihypertensive agents before the study entry, particularly use of thiazides, indicated poorer control at 12 months. Additionally, diabetes and/or target organ damage, insulin use and ECG-LVH in the B+A arm and proteinuria in the B+H arm indicated poor control. The results in identifying important predictors are quite consistent by using either SBP control or achieved SBP measures at month 12 for each arm.

It is well known that BP levels in Northern Europe are higher than in many other populations (6). The higher likelihood of reaching BP targets in American subjects has been previously described both in population studies (7) and in hypertension trials (8–10). In the multivariate analyses, we did simple adjustment for baseline BP differences, which allows for significant residual confounding and these findings should be carefully interpreted. However, since lower baseline SBP values also strongly predicted the subsequent BP control, it is likely that the lower baseline BP values in US patients in ACCOMPLISH can explain their lower achieved BP values. These baseline differences occurred largely because, unlike the US patients, those in the Nordic countries were allowed to enter the study only if they had failed to achieve BP control on previous therapy. Despite access to high doses of effective treatments, the Nordic patients as a group never caught up and achieved the same BP control levels as the Americans. This finding emphasizes that higher BPs are more difficult to control; patients whose BPs fail to respond well to initial antihypertensive therapies often continue to demonstrate resistance to treatment. Similar findings were seen in other clinical trials where early BP differences between treatment arms persisted for years despite the efforts of investigators (11–13). However, inasmuch as the difference in BPs diminished during treatment the Nordic physicians were not less aggressive in up-titrating and adding drugs compared with the US physicians.

It is not clear why being a Caucasian predicted better BP responses to treatments. However, all but three of the 1400 blacks were from the US and had significantly higher baseline SBP than the US non-blacks at entry, and it may be investigated further whether US blacks with similar baseline distributions as matched US Caucasians are more difficult to control. Like for the Nordic vs US issue, we did simple adjustment for baseline BP differences in the multivariate analyses, which allows for significant residual confounding and these findings should be carefully interpreted. BP reductions in African Americans, for instance, did not appear to be markedly different from those in other ethnic groups (2). Moreover, combination therapy of the type used in ACCOMPLISH has been shown generally equally effective in the black and non-black patients (14); however, it has yet not been investigated whether the two drug regimens used in ACCOMPLISH were equally effective in the black subgroup.

Differences in predictors between the two treatment arms were not striking but the previous use of CCB in the B+H group and thiazide use in the B+A group were significant and probably reflect discontinuation of these treatments at the time of randomization and roll-over into blinded study drugs of opposite categories. Pre-trial use of diuretics were categorized as thiazides (37%), loop-diuretics (12%) and aldosterone antagonists (1.5%) (5); though the use of indepamide is low or non-existing in the Nordic area and in the US, the category pre-trial thiazides may include thiazide-like chlortalidone in the US patients. If the predictor ability of pre-trial drug here is caused by the discontinuation at randomization, we assume that thiazide and chlortalidone act similarly in this aspect. Pre-trial treatment with CCB and thiazides may have been effective in these groups of patients; the question of discontinuation effect can be further investigated in the Nordic patients only where lack of control was an entry criterion. That insulin treatment predicted higher SBP in the B+A arm may suggest the need of diuretic treatment in diabetic patients to achieve a good BP control.

It is not clear why the concomitant use of treatment for dyslipidemia should predict better BP-lowering efficacy in the B+H group, though it is possible that patients threatened by multiple risk factors may be motivated to adhere to their therapies, thus achieving better efficacy. The ASCOT trial highlighted the clinical outcomes benefits of jointly administering antihypertensive and lipid-lowering therapies (13,15). However, positive outcome interaction in ASCOT was for CCB and statin treatment, while here we see a positive BP interference between thiazide and lipid lowering treatment. It may seem that statins induce small BP reduction though it does not explain the impact in the B+H arm only.

Strong predictors of non-controlled hypertension in the ACCOMPLISH trial are otherwise the typical markers of more severe hypertension and target organ damage like LVH in the B+A group and proteinuria in the B+H group. It cannot easily be explained why the heart and the kidney organ damage came out differently in this aspect. Because of the design of the study, all patients, regardless of risk status, received well-dosed, effective early treatment. In each group, 32.3% of the patients received approved antihypertensive agents in addition to the highest dose of study medication after 1 year in the study (1), i.e. equally distributed in the two arms and not explaining differential determinants of control of achieved SBP. Further efforts at BP normalization should then have been directed towards those patients with still uncontrolled hypertension or with cardiac, renal or diabetic manifestations in accordance with current hypertension guidelines (16,17).

The positive effect of higher baseline DBP on BP control is also interesting. Possibly this is explained by physicians increasing their efforts for achieving BP control by seeing apparently higher risk in patients. Possible, there may be an underlying mechanism indicating stiffer arteries and a wider pulse pressure in treatment resistant individuals.

In conclusion, irrespective of treatment, patients in the USA and Caucasians achieved better SBP control (in part because of their lower levels of SBP at study entry) while higher baseline SBP and more previous antihypertensive medications indicated less control. Concomitant use of lipid lowering treatment indicated better SBP control in the B+H arm, and diabetes and/or target organ damage like insulin use and ECG-LVH in the B+A arm and proteinuria in the B+H arm indicated poor control.

Declaration of interest: The ACCOMPLISH Study was supported by Novartis. Sverre E. Kjeldsen, Kenneth A. Jamerson, George L. Bakris, Bertram Pitt, Björn Dahlöf, Eric J. Velazquez, Jaakko Tuomilehto, Jan Östergren, Hans Ibsen and Michael Weber have received grant support or honoraria from Novartis. Tsushung A. Hua, Roxzana Y. Kelly, Dion Zappe and Allen Hester are employees of Novartis. The authors alone are responsible for the content and writing of this article.