Pharmacogenomics of Hypertension in CKD: The CKD-PGX Study

Key Points

• The CKD-PGX study assessed the feasibility of pharmacogenomic testing for a panel of antihypertensive agent efficacy predictors.

• Most patients with uncontrolled hypertension had one or more drug-gene interactions predicting reduced efficacy of their medications.

• In 36% of cases, practitioners used genetic data to change BP management in their patients with CKD.

Visual Abstract

Abstract

Background

Patients with CKD often have uncontrolled hypertension despite polypharmacy. Pharmacogenomic drug-gene interactions (DGIs) may affect the metabolism or efficacy of antihypertensive agents. We report changes in hypertension control after providing a panel of 11 pharmacogenomic predictors of antihypertensive response.

Methods

A prospective cohort with CKD and hypertension was followed to assess feasibility of pharmacogenomic testing implementation, self-reported provider utilization, and BP control. The analysis population included 382 subjects with hypertension who were genotyped for cross-sectional assessment of DGIs, and 335 subjects followed for 1 year to assess systolic BP (SBP) and diastolic BP (DBP).

Results

Most participants (58%) with uncontrolled hypertension had a DGI reducing the efficacy of one or more antihypertensive agents. Subjects with a DGI had 1.85-fold (95% CI, 1.2- to 2.8-fold) higher odds of uncontrolled hypertension, as compared with those without a DGI, adjusted for race, health system (safety-net hospital versus other locations), and advanced CKD (eGFR <30 ml/min). CYP2C9-reduced metabolism genotypes were associated with losartan response and uncontrolled hypertension (odds ratio [OR], 5.2; 95% CI, 1.9 to 14.7). CYP2D6-intermediate or -poor metabolizers had less frequent uncontrolled hypertension compared with normal metabolizers taking metoprolol or carvedilol (OR, 0.55; 95% CI, 0.3 to 0.95). In 335 subjects completing 1-year follow-up, SBP (−4.0 mm Hg; 95% CI, 1.6 to 6.5 mm Hg) and DBP (−3.3 mm Hg; 95% CI, 2.0 to 4.6 mm Hg) were improved. No significant difference in SBP or DBP change were found between individuals with and without a DGI.

Conclusions

There is a potential role for the addition of pharmacogenomic testing to optimize antihypertensive regimens in patients with CKD.

Introduction

Hypertension and CKD are common intersecting diseases with enormous economic burden, morbidity, and mortality. The Centers for Disease Control and Prevention reports that 45% of the US population has hypertension, with approximately half of those individuals inadequately controlled (1). The prevalence of hypertension increases with the severity of CKD because 36% of patients with grade 1, 48% with grade 2, 60% with grade 3, and 84% with grade 4/5 CKD have concomitant hypertension (2). Impaired sodium excretion, extracellular volume expansion, activation of the renin-angiotensin system, and numerous vasoconstrictive effects all conspire to impair BP control in patients with CKD (3). International guidelines emphasize control of hypertension to reduce cardiovascular events in patients with CKD (4).

Many antihypertensive agents are subject to different forms of pharmacokinetic or pharmacodynamic drug-gene interactions (DGIs), each of which effect efficacy. For example, metoprolol is metabolized by the enzyme cytochrome P450 2D6 (CYP2D6), wherein poor metabolizers possess higher circulating concentrations of the drug at a given dose (5). The β-blocker class may be pharmacodynamically affected by β-1 adrenergic receptor (ADRB1) polymorphisms (6). The angiotensin receptor blocker losartan potassium is a prodrug metabolized by cytochrome P450 2C9 (CYP2C9) (7). Poor metabolizers of CYP2C9 have lower concentrations of its active metabolite (8). Hydralazine hydrochloride undergoes phase-2 metabolism by N-acetyltransferase 2 (9). Fast and intermediate acetylators will have lower concentrations and reduced efficacy of hydralazine at a given dose. The evidence supporting these (and other) pharmacogenomic drug-gene pairs has been generated through prior clinical studies, and this evidence has been previously summarized (10,11).

Over 80% of individuals with CKD and hypertension take two or more antihypertensives and 32% take four or more agents (12). Despite polypharmacy, 10.3 million US individuals with apparent treatment-resistant hypertension remain uncontrolled (13).

We present the results of a prospective cohort study, entitled CKD-PGX, that enrolled and genotyped 382 adults with hypertension, most with CKD, from Indiana University Health (IU Health) physicians nephrology clinics in three settings: a university health system, a county safety-net health system, and outlying suburban clinics near Indianapolis. A clinical genotyping assay was developed and implemented in these health systems, with results and recommendations recorded in the electronic health records (EHRs) for 40 variants and 11 drug-gene pairs relevant to hypertension control (14). The panel of pharmacogenomic predictors of antihypertensive response was embedded in routine clinical practice to aid patients and practitioners in arriving at an efficacious BP regimen, either by identifying less efficacious medications in an individual’s current regimen or selecting an efficacious drug as the “next” antihypertensive agent. The goal of this implementation study was to assess provider utilization, patient attitude, prevalence of actionable DGIs, BP control after pharmacogenomic panel testing, and, ultimately, encourage the successful uptake of evidence-based pharmacogenomic interventions.

Materials and Methods

Study Design

This was a prospective, observational cohort study. Subjects were recruited and provided informed consent during a nephrology clinic visit in the IU Health or Eskenazi Health systems between 2017 and 2019. BP was assessed upon enrollment and at 1-year follow-up. This study was approved by the Institutional Review Board of Indiana University (number 1705413046).

Study Population

Subjects were eligible for inclusion if they were aged ≥18 years old with the ability to provide consent and a genotyping sample. Subjects were required to have at least one of the following: systolic BP (SBP) ≥140 mm Hg on any two readings in the 12 preceding months, eGFR <60 ml/min per 1.73 m², or daily proteinuria >0.2 g by 24-hour urine collection or >0.2 g/g urine protein-creatinine ratio (15). Our analysis population comprised 425 individuals for the baseline subject survey, 382 individuals for the cross-sectional analysis between genotype and hypertension control, and 335 individuals for the longitudinal 1-year follow-up BP outcomes. The prevalence of hypertension in those presenting to our clinics during the study period determined the sample size.

Study Procedures

BP was obtained at baseline immediately after a nephrology clinic appointment, using a standard sphygmomanometer while seated at rest, and again from the clinic visit closest to 1 year after enrollment (approximately 6 months). Three BP measurements were acquired, each separated by 5 minutes. SBP and diastolic BP (DBP) were each averaged separately for the three measurements. Participants provided a whole blood or saliva sample and were genotyped for 40 variants in 11 genes related to antihypertensive response (Table 1, Supplemental Table 1). Genotyping was performed on a custom Taqman OpenArray (Fisher Scientific, Waltham, MA) as previously described (14). Providers received encrypted email alerts when genetic data were deposited in the EHR (approximately 2 weeks after testing) along with interpretations on drug efficacy. Efficacy was determined on the basis of the Food and Drug Administration (FDA) or Dutch Pharmacogenomic Working Group (DPWG) guidance. When guidance was unavailable, the most prevalent genotype was set as the normal efficacy group. Less prevalent genotypes were set as either reduced or increased efficacy on the basis of their relative effect size from prior clinical trials.

Table 1.

Variants of antihypertensive response

Genea	Biologic/Functional Significance	Genotype or Metabolizer Status	Predicted Phenotype	Reference(s)
CYP2C9	Encodes cytochrome P450 2C9, which metabolizes losartan into active metabolitesb	*1/*1, *1/*2	Standard exposure	(7,8,26–34)
		*2/*2, *1/*3,*2/*3, *3/*3, *1/*8, *1/*11, *3/*8	Reduced active metabolite exposure
CYP2D6	Encodes cytochrome P450 2D6, which metabolizes/inactivates metoprolol and carvedilol	Ultrarapid metabolizer	Reduced drug exposure	(5,22,35)
		Normal metabolizer	Standard efficacy
		Intermediate or poor metabolizer	Increased drug exposure
NAT2	Encodes N-acetyletransferase 2, which acetylates hydralazine to its inactive metabolite	Fast or intermediate acetylator	Reduced hydralazine exposure	(9)
		Slow acetylator	Increased hydralazine exposure
F7 (rs6046)c	Encodes clotting factor VII, BP effect may be related to its role in endothelial homeostasis	G/G	Standard amlodipine efficacy	(36)
		G/A or A/A	Reduced amlodipine efficacy
ADRB1	Encodes β1 adrenoceptor	Zero copies of 49S-389Rd	Reduced β-blocker response	(6,37–40)
		One copy of 49S-389Rd	Standard β-blocker response
		Two copies of 49S-389Rd	Greater β-blocker response
GRK4	Encodes G-protein–coupled receptor kinase 4, which maintains ADRB1 cell surface localization	Zero copies of 65L-142V	Greater β-blocker response	(41,42)
		One copy of 65L-142V	Standard β-blocker response
		Two copies of 65L-142V	Reduced β-blocker response
NEDD4L (rs4149601)e	Encodes an E3 ubiquitin ligase, which regulates ENaC expression	G/G	Increased diuretic efficacy	(43,44)
		G/A	Standard diuretic efficacy
		A/A	Reduced diuretic efficacy
NPHS1 (rs3814995)	Encodes the principal structural protein of the glomerular podocytes, nephrin	G/G	Standard ARB efficacy	(45,46)
		G/A or A/A	Increased ARB efficacy
VASP (rs10995)e	Encodes vasodilator-stimulated phosphoprotein, regulates smooth muscle contraction	A/A	Standard thiazide efficacy	(47)
		A/G or G/G	Increased thiazide efficacy
YEATS4 (rs7297610)c	Expression quantitative trait locus associated with thiazide efficacy in Black individuals	C/C	Standard thiazide efficacy	(23,48)
		C/T or T/T	Reduced thiazide efficacy
EBF1/FGF5/SH2B3 e	Three-gene model identified by GWAS but no evidence for direct functional/biologic significance	Zero efficacy alleles	Reduced thiazide efficacy	(24)
		One or two efficacy alleles	Standard thiazide efficacy
		Three or more efficacy alleles	Increased thiazide efficacy

ENaC, epithelial sodium channel of the principal cell; ARB, angiotensin receptor blocker; GWAS, genome-wide association study.

^aOnly antihypertensive pharmacogenomic genes are included here. For multiple single nucleotide variant models, please see Supplemental Table 1 for the variants tested. Additional pharmacogenomic variants unrelated to hypertension, but for genetic risk prediction genotype data, was given to providers.

^bNot all ARBs are metabolized by CYP2C9, e.g., olmesartan does not undergo significant metabolism.

^cThis variant only has supporting data in individuals of African ancestry. Clinicians were advised not to extrapolate findings to other populations.

^d49S refers to the A allele of rs1801252, and 389R refers to the C allele of rs1801253.

^eThis variant only has supporting evidence in individuals of European ancestry. Clinicians were advised not to extrapolate findings to other populations.

Nephrology providers (n=39) gave assent to enroll their patients and were trained on the interpretation of pharmacogenomic DGIs. The principal investigators did not alter or suggest changes to subjects’ prescriptions; all clinical care was at the behest of the primary nephrology provider.

Three surveys were administered: (1) each subject’s attitude toward genetic testing was evaluated in a 15-question survey at baseline (Supplemental Appendixes 1), (2) each provider (n=76) completed a baseline survey regarding their attitude toward the testing (16), and (3) each provider completed a return-of-results survey for every one of their enrolled patients to query whether they believed testing affected their clinical management (Supplemental Appendix 2).

Variables

Variables, including demographic characteristics, biochemical parameters, CKD status, comorbidities, socioeconomic factors (care location [safety-net hospital versus others], highest education level), and medication lists, were obtained from the EHR at baseline and at 1-year follow-up.

Outcomes

Study outcomes included (1) prevalence of uncontrolled hypertension (uHTN) associated with actionable DGIs at baseline, and (2) change in SBP and DBP at 1-year follow-up. A DGI or “actionable” genotype was defined as the presence of at least one variant predicting reduced efficacy for an antihypertensive agent a subject was taking at the time of enrollment. Secondary outcomes included patient attitudes toward genetic testing and provider utilization, as defined by the return-of-results survey.

Statistical Analysis

Baseline data and survey responses were analyzed descriptively and provided as percentages for categoric variables, mean±SD for normally distributed variables, and median (25th and 75th percentile) for non-normally distributed variables. Comparisons of categoric variables were expressed as an odds ratio (OR) with 95% CI. For our primary outcome analyses, subjects were considered to have a relevant DGI if one or more of their genetic variants predicted reduced efficacy of their prescribed antihypertensives. DGIs were coded as a binary variable (present or absent). Evaluation of the relationship between DGIs and hypertension control was performed by chi-squared test and adjusted for significant covariables (race, health system, and presence of CKD stage ≥3) using logistic regression. Subgroup analyses were performed for each individual drug-gene(s) pair. Variants predicting increased efficacy of antihypertensives were assessed in subgroup analyses. Change in BP within each individual at 1 year was assessed by paired t test.

Results

Participants

A total of 472 adult subjects were recruited and consented from outpatient nephrology clinics within the IU Health system, the Eskenazi Health safety-net system, and associated outlying suburban clinics (Figure 1). A total of 37 subjects withdrew from the study, most frequently due to personal reasons, an inability to follow-up during the coronavirus disease 2019 pandemic, or a genotyping failure for which they would not provide a repeat sample. The remaining 435 subjects were genotyped and completed baseline surveys. The characteristics of the overall population are given in Table 2. There were approximately equal numbers of women and men, and the average age was 58.1±14.9 years old. The average body mass index was 33.7±8.4 kg/m². The majority of participants had baseline CKD, with 95% having any CKD stage, 78% had stage ≥3 CKD, and 27% had CKD stage ≥4. An International Classification of Diseases, Tenth Revision diagnosis of hypertension was recorded in 92% (n=401) of participants; however, only 382 subjects (88%) were treated with antihypertensive agents. The average number of antihypertensive agents prescribed was 3.0±1.4 in the 382 subjects taking antihypertensives. Angiotensin-converting enzyme inhibitors and angiotensin receptor blockers (58%) or β-blockers (56%) were commonly prescribed. Common comorbidities included diabetes, heart disease, and sleep apnea. The overall mean±SD SBP was 139.9±22.1 mm Hg, and the mean±SD DBP was 80.7±12.0 mm Hg. Of the 382 subjects on antihypertensives, 335 subjects completed 1-year follow-up.

Figure 1.

Enrollment and inclusion in the CKD-PGX study. Survey data were available in 435 adult participants who received the pharmacogenomic genotyping panel. There were 382 subjects who had a hypertension diagnosis and were prescribed one or more antihypertensives. These 382 participants were included in the cross-sectional analysis. In the longitudinal analysis, there were 335 subjects included who completed a 1-year follow-up with subsequent BP assessment.

Table 2.

Baseline characteristics of subjects included in overall genotype-analysis

Characteristic	Entire Cohort (N=435)	Uncontrolled Hypertensionb (n=189)	Controlled Hypertensionc (n=193)	P Value (Uncontrolled Hypertension versus Controlled Hypertension)
Female sex, n (%)	220 (51)	90 (48)	97 (50)	0.71
Age, yr, mean±SD	58.1±14.2	59.5±3.9	57.9±13.4	0.65
BMI, kg/m2, mean±SDa	33.7±8.4	33.8±8.9	34.1±7.9	0.62
Race, n (%)				0.004
White	253 (58)	92 (49)	125 (65)
Black	175 (40)	93 (49)	65 (34)
Asian	6 (1)	2 (1)	4 (2)
Native American	1 (0.2)	1 (0.5)	0
Hispanic ethnicity, n (%)	6 (1)	4 (2)	2 (1)	0.66
Baseline BP, mm Hg, mean±SD
Average systolic	139.9±22.1	158.0±17.6	123.3±9.6	<0.001
Average diastolic	80.7±12.0	86.9±12.6	75.4±8.7	<0.001
CKD stage, n (%)				0.1
1 (eGFR >90 ml/min per 1.73 m2)	18 (4)	10 (5)	6 (3)
2 (eGFR 60–89 ml/min per 1.73 m2)	56 (13)	26 (14)	27 (14)
3a (eGFR 45–59 ml/min per 1.73 m2)	109 (25)	47 (25)	51 (26)
3b (eGFR 30–44 ml/min per 1.73 m2)	114 (26)	41 (22)	62 (32)
4 (eGFR 29–15 ml/min per 1.73 m2)	79 (18)	40 (21)	32 (17)
5 (eGFR <15 ml/min per 1.73 m2)	38 (9)	18 (10)	9 (5)
No kidney disease	21 (5)	7 (4)	6 (3)
Any GFR ≥30 ml/min per 1.73 m2	318 (73)	131 (69)	152 (79)	0.04
Any GFR <30 ml/min per 1.73 m2	117 (27)	58 (31)	41 (21)
Health system, n (%)				0.001
Academic health center	137 (32)	49 (26)	53 (28)
Suburban clinic	127 (29)	50 (26)	80 (41)
Safety-net health system	171 (39)	90 (48)	60 (31)
Education level, n (%)				0.18
Some high school	48 (11)	26 (14)	15 (8)
Completed high school or GED	107 (25)	47 (25)	49 (25)
Some college	129 (30)	64 (34)	58 (30)
Bachelor’s degree	89 (21)	32 (17)	40 (21)
Graduate degree	43 (10)	15 (8)	24 (12)
Did not respond	19 (4)	5 (1)	7 (4)
Visit type at enrollment, n (%)
Initial consult	29 (7)	13 (7)	13 (7)	0.96
Follow-up appointment	406 (93)	176 (93)	180 (93)
Coexisting medical conditions, n (%)
Hypertension	401 (92)	189 (100)	193 (100)	>0.99
Diabetes	198 (46)	93 (49)	83 (43)	0.22
Coronary artery disease/prior MI	72 (17)	33 (18)	35 (18)	0.86
Congestive heart failure	71 (16)	35 (19)	29 (15)	0.36
Peripheral vascular disease	22 (5)	12 (6)	7 (4)	0.22
Obstructive sleep apnea	119 (27)	51 (27)	57 (30)	0.58
Cancer	80 (18)	31 (16)	39 (20)	0.34
Nonsteroidal anti-inflammatory use	65 (15)	31 (16)	24 (12)	0.27
Current smoking status	68 (16)	33 (18)	27 (14)	0.35

BMI, body mass index; GED, general equivalency diploma; MI, myocardial infarction.

^aBMI was missing for seven individuals.

^bBP ≥140/90 mm Hg.

^cBP ≥140/90 mm Hg.

Association of DGIs and BP Control

Of the 382 individuals prescribed antihypertensive agents at baseline, 189 had uHTN, defined as a SBP of ≥140 mm Hg or DBP of ≥90 mm Hg (Table 2). The mean±SD SBP in subjects with uHTN was 158.0±17.6 mm Hg. The mean±SD SBP in individuals with controlled hypertension (cHTN; n=193) was significantly lower at 123.3±9.7 (P<0.001). Similarly, the average DBP in subjects with uHTN was higher (P<0.001). Those with cHTN and uHTN were similar in age, sex distribution, body mass index, and frequency of comorbidities. The prevalence of CKD stages 4 and 5 was higher in the uHTN group (P=0.04). The distribution of race was significantly different between groups: 65% of subjects with cHTN were White, whereas only 49% of subjects with uHTN were White (P=0.004). Subjects who received care in the safety-net health system (a potential surrogate for lower socioeconomic status) were more likely to have uHTN (P=0.001).

In the 382 subjects at baseline, relevant DGIs predicted to reduce efficacy of a currently prescribed antihypertensive agent were assessed. The genotype distribution is summarized in Supplemental Table 2. The majority (58%) of participants with uHTN had an actionable genotype. Subjects with an actionable DGI had two-fold (95% CI, 1.3- to 3.0-fold) higher odds of uHTN as compared with those without an actionable genotype (P=0.0008; Table 3). Race was also associated with DGI prevalence in a univariate analysis: 61% of Black individuals had a DGI, whereas 43% of White subjects had a DGI (P<0.001). When adjusted for significant variables in the univariable analysis (race, eGFR of <30 ml/min per 1.73 m², and health system), subjects with a relevant DGI had 1.85-fold (95% CI, 1.2- to 2.8-fold) increased odds of uHTN. In summary, individuals who had uHTN were more likely to have a genetic variant that predicted reduced efficacy of an antihypertensive agent that they were prescribed at baseline.

Table 3.

Association of antihypertensive drug-gene interactions with BP

Group	n (%)		Odds Ratio (95% Confidence Interval)	P Value
	uHTN(N=189)	cHTN(N=193)
Unadjusted chi-squared analysis
Relevant DGIa	110 (58)	79 (41)	2.01 (1.3 to 3.0)	0.0008
No relevant DGI	79 (42)	114 (59)
Logistic regression analysis
Relevant DGI			1.85 (1.2 to 2.8)	0.0001
Race (non-White)			1.37 (0.8 to 2.2)
CKD (eGFR <30 ml/min)			1.37 (0.9 to 2.2)
Health system (safety net)			1.63 (1.01 to 2.6)

Unadjusted analysis performed with a chi-squared analysis. Logistic regression analysis was adjusted for race, stage of CKD, and health system. uHTN, uncontrolled hypertension; cHTN, controlled hypertension; DGI, drug-gene interaction.

^aRelevant DGI refers to the presence of a reduced-efficacy variant for an antihypertensive agent a subject was taking on enrollment.

Individual Drug-Gene Analyses

As exploratory analyses, we examined the association between relevant DGIs and baseline uHTN for each individual DGI (Supplemental Table 3). Significant associations between uHTN and DGIs were found for participants prescribed losartan, metoprolol, and carvedilol. Variants in CYP2C9 that predicted reduced efficacy of losartan were associated with uHTN in participants taking the drug (OR, 5.2; 95% CI, 1.9 to 14.7). Intermediate or poor CYP2D6 metabolizers have higher circulating concentrations of metoprolol or carvedilol. These individuals were less likely to have uHTN than normal metabolizers taking either agent (OR, 0.55; 95% CI, 0.3 to 0.95). No other significant DGIs were identified in this relatively small sample size.

Longitudinal BP Control

Overall, the 335 subjects who completed a 1-year follow-up in the nephrology clinic had a significant decrease in BP, both SBP (−4.0 [95% CI, 1.6 to 6.5] mm Hg) and DBP (−3.3 [95% CI, 2.0 to 4.6] mm Hg). Among the 163 individuals with uHTN, 71 were “controlled” by 1 year follow-up, with a SBP of <140 mm Hg and a DBP of <90 mm Hg. Table 4 shows the within-group comparisons of baseline and 1-year follow-up BPs in the overall cohort, with and without a DGI, and in those with uHTN at baseline, with and without a baseline actionable genotype. Most comparisons were significant except reduction in SBP among subjects without a DGI. Between individuals with and without a baseline DGI, no significant difference was identified in change in SBP (P=0.54) or DBP (P=0.10), although the genotyping may affect BP management in both groups by avoidance of a future DGI in the 1-year follow-up.

Table 4.

Longitudinal BP assessment

Group	SBP				DBP
	Baseline, Mean±SD	1 Year, Mean ±SD	Change (95% Confidence Interval)	P Value	Baseline, Mean ±SD	1 Year, Mean ±SD	Change (95% Confidence Interval)	P Value
All subjects (N=335)	140.2±22.5	136.2±19.8	4.0 (1.6 to 6.5)	0.002	80.8±12.2	77.6±11.3	3.3 (2.0 to 4.6)	8.6×10−7
Subjects with DGI (n=160)a	143.2±22.6	138.3±18.9	4.8 (1.3 to 8.3)	0.008	81.5±12.8	77.1±11.1	4.4 (2.5 to 6.3)	1.0×10−5
Subjects with no DGI (n=175)a	137.4±22.0	134.1±20.3	3.4 (-0.5 to 6.9)	0.14	80.2±11.6	77.9±11.4	2.3 (0.5 to 4.0)	0.01
Subjects with uHTN (n=163)	157.9±17.9	143.0±20.9	14.9 (9.5 to 20.3)	3.0×10−13	86.7±12.5	79.4±12.3	7.4 (4.5 to 10.3)	9.6×10−9
Subjects with uHTN and DGI (n=90)a	158.8±16.9	144.0±19.5	14.8 (10.2 to 19.4)	1.3×10−9	86.1±13.7	77.7±11.6	8.4 (5.9 to 10.9)	4.1×10−9
Subjects with uHTN but no DGI (n=73)a	156.9±19.2	141.9±22.3	15.0 (9.7 to 20.3)	2.1×10−7	87.6±10.8	81.4±12.9	6.1 (3.4 to 8.8)	1.7×10−5

SBP, systolic BP; DBP, diastolic BP; DGI, drug-gene interaction; uHTN, uncontrolled hypertension.

^aRelevant DGI refers to the presence of a reduced-efficacy variant for an antihypertensive agent a subject was taking on enrollment.

Provider Utilization

Action-driven item surveys for each recruited subject were completed by their primary nephrology provider. The surveys queried the utility of genotype results in each subjects’ antihypertensive drug management. The response rate was 54% (180 of 335). Physicians reported that the genetic testing altered their diagnosis or management in 36% of all participants and 38% of participants with baseline uHTN. In 85% of survey responses, physicians stated they had or would discuss results with their patients.

Patient-Reported Attitudes to Genotyping

Among the full cohort of 435 subjects with hypertension, proteinuria, or an eGFR of <60 ml/min per 1.73 m², 425 subjects completed a baseline survey. Supplemental Figure 1 illustrates the response from selected questions regarding subjects’ attitude toward genetic testing. The surveys revealed that few subjects (5%) were familiar with the terms pharmacogenomics, genetic testing, or personalized medicine. More than 96% of participants reported that knowledge of their genetic code would prompt them to invest more to control their BP and enable their providers to deliver enhanced antihypertensive care.

Discussion

Antihypertensive medication response is frequently unpredictable and varies among individuals. In CKD, BP control at recommended targets is clearly important to prevent cardiovascular disease and end organ dysfunction. However, 25%–33% of individuals with grade ≥3 CKD have uncontrolled BP, despite therapy with three or more medications (17). Society guidelines often provide initial and secondary agent recommendations extrapolated from the general population (18,19), but they may fail to provide specific direction for those with CKD and apparent treatment-resistant (uncontrolled) hypertension.

In this study, relevant DGIs were found in 58% of subjects. We identified a significant OR of 1.8 (95% CI 1.2-2.8) between reduced-efficacy DGIs and uHTN, even after adjusting for presence of CKD, race, and health system. When providers were armed with, and trained on, the genotype information, they reported adjusting the antihypertensive regimen in 36% of subjects. Study subjects were receiving ongoing care in nephrology clinics, including frequent follow-up, ambulatory BP monitoring when required, secondary hypertension workup, and dietary counseling. The only addition to standard care was the provision of the genotype report in the EHR and notification to the subject’s physician. Medication changes, if any, were made by the primary nephrologists who indicated they discussed the genetic results with 85% of subjects. Patient engagement is a necessary component of hypertension management. Indeed, surveyed subjects agreed that understanding their pharmacogenomic report would facilitate greater action, on their part, to control BP. The adjustments made by the patients and providers resulted in a decrease of 4 mm Hg in SBP within individuals across the entire cohort and 14.9 mm Hg in those with baseline uHTN.

Our pharmacogenomics panel–based approach included a wide array of variants. A summary of the available evidence is beyond the scope of this discussion; however, antihypertensive variants were selected on the basis of the strength of evidence, minor allele frequency, FDA label annotations, and guidelines. The FDA drug labels of hydralazine (20), losartan (8), and metoprolol (21) each reference metabolic enzymes which affect concentrations of these drugs. Where major society guidelines were available, such as for metoprolol and carvedilol from the DPWG (22), they were used to guide recommendations. As additional guidelines from DPWG and the Clinical Pharmacogenomics Implementation Consortium are made available, these will need to be incorporated into future genotype-guided dosing recommendations.

Forty efficacy variants were included across a breadth of agents to maximize actionability and effect for enrolled subjects. In some cases, pharmacokinetic and pharmacodynamic variants predicted opposite effects for the same drug. For example, the CYP2D6 poor-metabolizer status predicts increased circulating concentrations of metoprolol, yet a subject with a reduced function variant of ADRB1 would have the opposite predicted efficacy. There is insufficient evidence to reconcile these two effects; thus, all variants were reported separately, and the prescribing clinician decided how they would use the information. Consideration was given to evidence in multiple populations. The Pharmacogenomics Evaluation of Antihypertensive Response group identified different predictors of thiazide efficacy in Black and White individuals. As such, YEATS4 was used as a predictor in Black individuals (23), whereas a different three-gene model was used in White individuals (24). In this study, the outcomes of utilization and BP control incorporate a level of decision making on the part of the provider. The direction of effect for drug-gene prediction, the identification of race by the provider and subject, and the choice of medications to treat apparent treatment-resistant hypertension are inherent aspects of the decision-making process for all providers.

A significant strength of the CKD-PGX cohort is generalizability. The study population was 57% White, 40% Black, and was recruited from three diverse environments: a university hospital, a safety-net health system, and several suburban clinics. The primary limitation of the study is that it was a prospective cohort, not a randomized controlled trial. Thus, observation bias in BP control (the Hawthorne effect) may contribute to the outcomes measured. This bias is inherent in all prospective cohort studies. Given the proportion of uHTN in this cohort, regression to the mean is expected in subjects receiving specialty nephrology care. Thus, a randomized controlled trial is necessary to establish a cause-and-effect relationship between genotyping and BP control.

This study was underpowered to detect the significance of specific prescribing behavior. In practice, this may hold less relevance because patients are on multiple antihypertensives, frequently with multiple actionable genotypes. Some providers would elect to change dose, select an alternate medication, or add a medication. Herein, a binary variable of provider self-reported utility was assessed in each subject. A pharmacogenomic panel–based approach was used, and our sample size was insufficient to detect the effect of any single variant. Significant loss to follow-up (9%) did occur, in part due to the coronavirus disease 2019 pandemic which shifted follow-up to telehealth and affected assessment of BP. The action item and subject surveys were not validated as part of this study. Finally, the role of patients’ lifestyles, diets, and medication compliance, and the relationship of medication compliance to BP control was not tracked. These limitations are counterbalanced by our clinically translatable outcomes, such as the association of DGIs with BP, longitudinal BP control, and provider utilization.

This study outlines the implementation of pharmacogenomic panel testing in an outpatient nephrology setting. Currently, individual patient demographics, such as obesity, sex, and race, are important factors in the selection process for an adequate antihypertensive regimen (25). This study suggests a potential role for the addition of pharmacogenomic data to select antihypertensive regimens. In the future, whole genome or exome sequencing will likely be integrated into the clinical setting. At that time, genotype information will be readily available to providers and will usher in an exciting era in the care of patients with hypertension and CKD. A higher proportion of DGIs were found in non-White individuals. Given the widespread prevalence of poor BP control in those with CKD across the world, the ability to select medications on the basis of genotype has far-reaching consequences in potentially addressing complications of hypertension, such as cardiovascular and cerebrovascular mortality.

Disclosures

A.B. Chapman reports receiving research funding from National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), Reata, and Sanofi; serving on the special emphasis panel and review panel for the National Institutes of Health (NIH)/NIDDK and Small Business Innovation Research, and on the US Department of Defense Review Committee; serving as an external advisor for the O’Brien Center Northwestern University; serving on a speakers bureau for Otsuka; receiving honoraria from Otsuka and Reata; and having consultancy agreements with Otsuka Pharmaceuticals, Reata, and Sanofi Pharmaceuticals. B.W. Miller reports receiving honoraria from DaVita, DaVita Kidney Care, Fresenius Renal Therapies Group, Northpointe Meetings, and UpToDate; having consultancy agreements with Fresenius Medical Services; and serving on the NxStage Medical Scientific Advisory Board. S.M. Moe reports serving on the editorial boards of American Journal of Nephrology and American Journal of Nutrition; having consultancy agreements with, and receiving honoraria from, Amgen, Ardelyx, and Sanifit; receiving research funding from Chugai (research grant), Keryx Research (grant), the NIH (research grant), and Veterans Administration (VA merit); and having ownership interest in Eli Lilly (via stock). V.M. Pratt reports having other interests in/relationships with the American College of Medical Genetics and Genomics and Association for Molecular Pathology; having consultancy agreements with Avalon Healthcare LLC, Everlywell Health, Laboratory Corporation of America, and MindX Sciences LLC; and serving as a scientific advisor for, or member of, the Journal of Molecular Diagnostics, FDA, and Veterans Affairs. A.D. Sinha reports receiving research funding from Bayer, and having consultancy agreements with George Clinical and Johnson & Johnson. T. S. Skaar reports receiving research funding from Indiana University School of Medicine Precision Health Initiative and the NIH. All remaining authors have nothing to disclose.

Funding

Support for this work was provided by the NIH/NIDDK grant K08DK107864 (to M.T. Eadon) and the Indiana University Grand Challenge Precision Health Initiative. S. M. Moe was funded by NIH/NIDDK grant R01DK11087103 and National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) grant P30AR072581-01. R.N. Moorthi was supported by NIH/NIDDK grant K23DK102824 and NIAMS grant R01AR077273.

Author Contributions

A.B. Chapman, M.T. Eadon, S.M. Moe, R.N. Moorthi, V.M. Pratt, S.J. Sher, and T.S. Skaar reviewed and edited the manuscript; A.B. Chapman, M.T. Eadon, S.M. Moe, R.N. Moorthi, V.M. Pratt, and J. Su were responsible for methodology; A.B. Chapman, M.T. Eadon, S.M. Moe, R.N. Moorthi, and T.S. Skaar conceptualized the study; M.T. Eadon, J. Maddatu, R. Melo Ferreira, R.N. Moorthi, and J. Su were responsible for data curation and formal analysis; M.T. Eadon, J. Maddatu, S.M. Moe, R.N. Moorthi, V.M. Pratt, S.J. Sher, and A.D. Sinha were responsible for investigation; M.T. Eadon, J. Maddatu, and R.N. Moorthi wrote the original draft; M.T. Eadon, B.W. Miller, S.M. Moe, R.N. Moorthi, V.M. Pratt, A.D. Sinha, and J. Su were responsible for resources; M.T. Eadon, B.W. Miller, S.M. Moe, R.N. Moorthi, S.J. Sher, and A.D. Sinha provided supervision; M.T. Eadon, B.W. Miller, S.M. Moe, R.N. Moorthi, S.J. Sher, A.D. Sinha, and T.S. Skaar were responsible for project administration; M.T. Eadon, S.M. Moe, R.N. Moorthi, and T.S. Skaar were responsible for funding acquisition and visualization; M.T. Eadon, R.N. Moorthi, V.M. Pratt, and T.S. Skaar were responsible for validation; and M.T. Eadon, R.N. Moorthi, and J. Su were responsible for software.

Supplemental Material

This article contains supplemental material online at http://kidney360.asnjournals.org/lookup/suppl/doi:10.34067/KID.0005362021/-/DCSupplemental.

Supplemental Appendix 1

Subject survey. Download Supplemental Appendix 1, PDF file, 327 KB ^{(326.8KB, pdf)}

Supplemental Appendix 2

Provider survey. Download Supplemental Appendix 2, PDF file, 327 KB ^{(326.8KB, pdf)}

Supplemental Figure 1

Subject reported survey results. Download Supplemental Figure 1, PDF file, 327 KB ^{(326.8KB, pdf)}

Supplemental Table 1

Gene descriptions. Download Supplemental Table 1, PDF file, 327 KB ^{(326.8KB, pdf)}

Supplemental Table 2

Genotype frequency. Download Supplemental Table 2, PDF file, 327 KB ^{(326.8KB, pdf)}

Supplemental Table 3

Cross sectional associations of individual drug-gene interactions with baseline uncontrolled hypertension. Download Supplemental Table 3, PDF file, 327 KB ^{(326.8KB, pdf)}