Clinical and molecular approaches to individualize antihypertensive drug therapy

Abstract

Interindividual variation of blood pressure (BP) responses to antihypertensive drugs is extensive. Several clinical, laboratory, and genetic predictors of BP responses to blood pressure-lowering agents have been suggested. We describe here the principal findings from the GENRES Study which is primarily a pharmacogenetic study of antihypertensive drug responses but also includes analysis of certain clinical and laboratory predictors. In this placebo-controlled, double-blinded, and randomized study, more than 200 male subjects with essential hypertension were treated with four antihypertensive drug monotherapies (amlodipine, bisoprolol, hydrochlorothiazide, and losartan) in a cross-over fashion, resulting in more than 800 treatment periods. Generally, placebo BP level was the best predictor of BP responses. In addition, higher baseline plasma renin activity predicted better BP response to losartan and bisoprolol, and weaker response to hydrochlorothiazide. A number of candidate gene polymorphisms analysed so far have given negative results in relation to BP responses, with the exception of an STK39 variant associating with losartan responsiveness. In future, genome-wide association studies on antihypertensive pharmacogenetics may identify novel pathways of BP regulation and provide new tools for both basic research and clinical use.

Key words::

• Adrenergic beta-1 receptor antagonists
• angiotensin II type 1 receptor blockers
• calcium channel blockers
• drug therapy
• hypertension
• pharmacogenetics
• thiazides

Abbreviations
ABP	=	ambulatory blood pressure
BP	=	blood pressure
DBP	=	diastolic blood pressure
GWAS	=	genome-wide association studies
OBP	=	office blood pressure
PRA	=	plasma renin activity
SBP	=	systolic blood pressure

Key messages

• Several predictors of antihypertensive responses have been identified.

• No genetic polymorphisms can currently be used in clinical practice to predict blood pressure responses in essential hypertension.

• Large-scale genomic approaches may identify novel pathways of blood pressure regulation.

Introduction

Although hypertensive patients as a group respond to most antihypertensive drug treatments with a reasonably predictable blood pressure (BP) decrease, the individual responses are highly variable. In occasional cases, BP responses in essential hypertension may compare to those seen in cases with monogenic forms of hypertension where mutation-directed tailored treatments are initiated, while in other cases no response or even increased BP levels are observed, whether encountered in clinical practice or in controlled studies (1,2).

The search for predictors of antihypertensive drug responses has two motives. First, clinical treatment of hypertension requires intensification, since few treated patients reach target BP levels (3,4). Although most patients in clinical trials appear to need combination therapy to reach desirable BP levels when treatment protocols are fixed, rotational studies have shown that monotherapy may be effective in the majority of patients if the proper drug can be found (5). Availability of robust predictors of BP response could facilitate success in treatment. Second, BP is a tricky phenotype to assess. In addition to the difficulties in obtaining reliable BP measurements, multiple known and unknown genetic and lifetime environmental factors and physiological pathways affect BP levels and underlying regulatory mechanisms. Genome-wide association studies (GWAS) on population BP levels have given a number of positive results, but only with small effect sizes after expansion of subject numbers to several tens of thousands, possibly because these studies were mostly based on single time point BP measurements with high intraindividual variability. On the other hand, using BP response to a specific pharmacological therapy as the phenotype may provide an alternative way to identify gene loci influencing BP levels, as use of BP levels prior to and during drug treatment may partially control for interindividual contribution of environmental factors.

In this mini-review, we present data on the variability and the predictors of BP responses to antihypertensive treatment from the GENRES Study (2).

Methodological aspects of the GENRES Study

GENRES is a double-blind, placebo-controlled, randomized, and cross-over study with four 4-week monotherapies from the major antihypertensive drug groups (losartan, bisoprolol, amlodipine, and hydrochlorothiazide). Both 24-h ambulatory (ABP) and office (OBP) blood pressure measurements were carried out. More than 200 middle-aged (35–59 years) Finnish men completed the study, resulting in more than 800 drug treatment periods. There were 4-week placebo periods before and after each treatment period. Consequently, the placebo BP level estimates should be very accurate, since they are derived from four different placebo periods. Estimation of the number of subjects required was based on a priori power calculations (2). For specific genetic polymorphisms, post-hoc power calculations may be performed. For example, assuming genotype group numbers of 180 (wild-type homozygotes) and 25 (variant allele heterozygotes), alpha level of 0.05, the observed SD of diastolic blood pressure (DBP) responses, and a DBP response difference of 3 mmHg, a power of 93% is obtained.

Variability of blood pressure responses

The ABP measurements were more repeatable during the placebo periods than the OBP measurements: the coefficients of variation were 3.6%/3.5% (systolic/diastolic) in ABP and 5.4%/5.2% in OBP measurements. Therefore, we have chosen ABP responses as the primary target variable in our studies. The BP responses were highly variable, confirming earlier research and clinical data. For example, the mean 24-h systolic ABP response to 50 mg of losartan was –9.1 mmHg (n = 204). The SD of responses was 6.7 mmHg, a value lower than reported for other pharmacogenetic studies with ABP measurements, and the responses ranged from + 9 (increase) to –32 (decrease) mmHg.

The within-subject correlations of BP responses between the study drugs were variable, with best correlations between bisoprolol and losartan (r = 0.32/0.39; SBP/DBP responses), and amlodipine and hydrochlorothiazide (r = 0.35/0.35), largely confirming the findings of Brown and co-workers (5). The data are presented in detail in our previous paper (2).

Multiple regression modelling of blood pressure responses using clinical chemical and endocrine variables

We have analysed BP responses to the four antihypertensive treatments using parameters available in clinical practice (2,6). Here we describe the main results of stepwise multiple regression modelling using a combined set of parameters. In this exploratory stepwise modelling, carried out with SPSS software (version 17.0, SPSS Inc., Chicago, IL, USA), a P value threshold of <0.10 was used. The 24-h ABP responses were normally distributed, or had only minor deviations not affecting the models. The baseline predictors in the analyses were transformed to normal distribution when needed and included the following: placebo 24-h ABP level (systolic or diastolic), age, earlier use of antihypertensive medication (no/yes) or a diuretic (no/yes), body mass index, waist-to-hip ratio, current smoking (no/yes), plasma renin activity (PRA), serum aldosterone, serum aldosterone-to-PRA ratio, daily urinary excretions of sodium and albumin, and serum sodium, potassium, insulin, fasting glucose, urate, cholesterol, triglycerides, calcium, and creatinine. Of highly correlated parameters, only the best predictor was included in the final model. The combined results are shown in Table I.

Table I. Stepwise multiple regression modelling of ambulatory blood pressure responses to four antihypertensive monotherapies. P values, direction of the effect, and semipartial R² values are shown.

	Amlodipine responses						Bisoprolol responses						Hydrochlorothiazide responses						Losartan responses
	Systolic			Diastolic			Systolic			Diastolic			Systolic			Diastolic			Systolic			Diastolic
	P	Eff.	PtR^2	P	Eff.	PtR^2	P	Eff.	PtR^2	P	Eff.	PtR^2	P	Eff.	PtR^2	P	Eff.	PtR^2	P	Eff.	PtR^2	P	Eff.	PtR2
Placebo BP level	6E-13	+	0.18	1E-11	+	0.17	0.002	+	0.05	1E-4	+	0.07	4E-5	+	0.07	1E-4	+	0.07	0.003	+	0.04	2E-4	+	0.06
Age	4E-4	+	0.04	6E-6	+	0.07	ns			ns			0.05	+	0.02	ns			0.09	+	0.01	ns	+
Waist-hip ratio	ns			ns			ns			ns			ns			ns			ns			0.04		0.02
Smoking	0.09	+	0.01	0.06	+	0.01	0.06	–	0.02	ns			ns			ns			ns			ns
Earlier diuretic use	ns			0.02	–	0.02	ns			ns			ns			ns			ns			ns
Renin	ns			ns			0.01	+	0.03	0.02	+	0.02	0.05	–	0.02	0.008	–	0.03	1E-4	+	0.06	1E-5	+	0.08
Aldo/renin	0.04	+	0.01	ns			ns			ns			ns			ns			ns			ns
S-Sodium	ns			ns			ns			ns			ns			ns			0.06	+	0.02	0.003	+	0.04
S-Potassium	ns			ns			ns			ns			0.04	–	0.02	ns			ns			0.05	+	0.01
S-Calcium	(0.14)	–		0.02	–	0.02	ns			ns			ns			ns			ns			ns
S-Creatinine	ns			ns			ns			ns			ns			ns			ns			(0.10)	+
S-Glucose	ns			ns			ns			ns			ns			ns			0.005	–	0.04	6E-4	–	0.05
S-Cholesterol	5E-5	–	0.05	0.003	–	0.03	0.02	–	0.02	(0.13)	–		ns			ns			ns			ns
S-Triglycerides	ns			ns			ns			ns			0.03	–	0.02	ns			ns			ns
dU-Sodium	2E-5	–	0.06	0.01	–	0.02	ns			ns			ns			0.08	–	0.01	ns			ns
dU-Albumin	ns			ns			ns			ns			0.04	–	0.02	0.06	–	0.02	ns			ns
R^2 of the model		0.39			0.35			0.09			0.09			0.19			0.12			0.13			0.22

P values > 0.10 in parentheses indicate conditional significancies for the next modelling step (not included in the final model). The following predictors were included in the analyses but were not significant predictors of any of the BP responses: earlier antihypertensive medication use, body mass index, S-Aldosterone, S-Insulin, and S-Urate. Abbreviations: + /– = better/poorer BP response with increasing explanatory variable value; Aldo/renin = S-aldosterone to plasma renin activity ratio; BP = blood pressure; dU = daily urinary excretion; Eff. = effect; ns = P > 0.10; PtR² = drop in R² if the variable was removed from the model; Renin = plasma renin activity; S = fasting serum.

Losartan

Previously proposed predictors of BP response to angiotensin-blocking agents include PRA (higher values, better response) and African-American race (lower response) (7,8). In the GENRES Study, significant predictors of both systolic and diastolic 24-h ABP responses to losartan were PRA, placebo BP level, and fasting serum glucose and sodium levels (Table I). The finding of PRA as the strongest predictor (P values 0.0002/0.00001 for SBP/DBP responses in multiple regression analysis) confirms earlier reports and partially validates the design of the GENRES Study. As novel findings, lower fasting serum glucose and higher serum sodium levels predicted better response to losartan (6). Figure 1 shows the BP responses in extreme quartiles of PRA and serum sodium levels.

Figure 1. Ambulatory 24-h systolic (A) and diastolic (B) blood pressure responses to losartan in the lowest and the highest quartiles of plasma renin activity (PRA) and serum sodium. Means and SDs are shown. Statistical significance between the extreme quartiles was calculated with Student's t test. SBP = systolic blood pressure; DBP = diastolic blood pressure; PRA = plasma renin activity.

Bisoprolol

Higher PRA levels are known to predict better BP response to beta-blockers (9), which was confirmed in our study (6). In addition to placebo BP level and PRA, no other parameters predicted both systolic and diastolic ABP responses to bisoprolol. From the four study drugs, lowest R-square (the coefficient of determination) values were obtained for bisoprolol using the described model (Table I).

Amlodipine

Higher age is known to predict better response to calcium channel blockers (8), which was even observed in the GENRES Study with a relatively narrow age range (35–60 years) of subjects (Table I). In addition, better systolic and diastolic ABP responses to amlodipine were predicted by higher placebo BP level, lower serum cholesterol, lower daily urinary excretion of sodium, and lower serum calcium (Table I, Figure 2) (2,6). The associations of cholesterol and calcium have earlier been described, but only in small studies (10–12).

Figure 2. Ambulatory 24-h systolic (A) and diastolic (B) blood pressure responses to amlodipine in the lowest and the highest quartiles of serum cholesterol and calcium. For abbreviations and statistics, see legend to Figure 1.

Hydrochlorothiazide

Low PRA is known to associate with better BP response to thiazide diuretics (9), which was confirmed in our study (6). In addition, higher placebo BP level and higher age predicted better response to hydrochlorothiazide (Table I) (2).

Pharmacogenetics of blood pressure responses

Using the GENRES Study as a platform, we have analysed potential pharmacogenetic associations of several hypertension candidate genes, as also derived from a number of gene variants showing positive signals in GWAS data (13–17), with BP responses to BP-lowering drugs. No significant positive results on any of the four antihypertensive drugs were obtained for polymorphisms angiotensinogen (AGT) Met235 Thr, angiotensin-converting enzyme (ACE) I/D, angiotensin type 1 receptor (AGTR1) 1166A/C, alpha-adducin (ADD1) Gly460Trp, beta-1-adrenergic receptor (ADRB1) Ser49Gly and Arg389Gly, beta-2-adrenergic receptor (ADRB2) Gly16Arg and Glu27Gln, or CYP2C9 *1/*2/*3 (18–20). The negative results do not, however, exclude the possibility that these polymorphisms are related to clinical outcomes in hypertensive patients. A detailed comparison of these data to the vast field of previous pharmacogenomics studies, mostly carried out with smaller materials, single drugs, and without 24-h BP recordings, is outside the scope of this mini-review. Considering large-scale clinical studies with BP responses as one of the end-points, our results on negative findings with the ACE and ADD1 polymorphisms are in agreement with those reported from the very large GenHAT study (21,22). The candidate gene approach also yielded negative results in the large EUROPA/PERGENE study, where no significant predictors of BP response to an ACE inhibitor among 52 SNPs in 12 genes within the renin-angiotensin-aldosterone system were found (23).

As the next step we proceeded to analysis of variations identified in the large GWAS of population BP levels (13–17). We selected 19 SNPs (24) and analysed their association with BP responses to all four study drugs, which resulted in 76 comparisons (counting the highly correlated SBP and DBP responses as one) and a Bonferroni-corrected P value requirement of <0.0007 for statistical significance (considering the SBP and DBP responses uncorrelated would give 152 comparisons and a P level requirement of <0.0003). In addition to some suggestive findings, one significant association was found. An intronic STK39 SNP (rs6749447) was associated with 24-h ABP response to losartan, with P values of 0.0005/0.0002 (systolic/diastolic) in multivariate analysis (24). Between the extreme genotype groups, the differences of mean systolic/diastolic 24-h ABP responses were 3.3/2.6 mmHg. The association, although not replicated thus far, is physiologically plausible since the STK39-encoded kinase SPAK is involved in the activation of renal ion transporters by angiotensin II (25–27). Although the thiazide-sensitive sodium-chloride cotransporter is also regulated by SPAK, the two analysed STK39 SNPs were not associated with BP response to hydrochlorothiazide (24), in congruence with data from another report (28).

Conclusions

BP responses to antihypertensive drugs display marked interindividual variability. Several baseline demographic and laboratory parameters predict BP responses to some extent, but classification of patients according to these variables would leave significant overlap in BP responses, well demonstrated by the SD bars of the BP responses (Figure 1). Pretreatment profiling of patients may decrease the probability of treatment failure (29), but also the non-preferred drugs should be considered at a later stage if target BP levels are not reached or adverse effects occur when the primary alternative is used as antihypertensive drug. At present, there is insufficient evidence to support the use of genetic polymorphisms as predictors of BP response in essential hypertension. It is still possible that large-scale genomic techniques may help to detect novel pathways of BP regulation and thereby also identify new targets for pharmacogenomics approaches.

Declaration of interest: The authors declare no conflict of interest.