Acute Declines in Estimated Glomerular Filtration Rate in Patients Treated With Benazepril and Hydrochlorothiazide Versus Amlodipine and Risk of Cardiovascular Outcomes

Abstract

Background

Acute declines in estimated glomerular filtration rate (eGFR) occur commonly after starting angiotensin‐converting enzyme inhibitors. Whether declines in eGFR that occur after simultaneously starting angiotensin‐converting enzyme inhibitors with other antihypertensive agents modifies the benefits of these agents on cardiovascular outcomes is unclear.

Methods and Results

We identified predictors of acute declines in eGFR (>15% over 3 months) during randomization to benazepril plus amlodipine versus benazepril plus hydrochlorothiazide in the ACCOMPLISH (Avoiding Cardiovascular Events through Combination Therapy in Patients Living with Systolic Hypertension) trial. We then determined the relation between declines in eGFR (treated as a binary variable, ≤15% versus >15% and separately, as a restricted spline variable) and the composite risk of fatal and nonfatal cardiovascular events using Cox proportional hazards models. We included 10 714 participants (median age 68 years [Q1 63, Q3 73]), of whom 1024 reached the trial end point over median follow‐up of 2.8 years. Predictors of acute declines in eGFR>15% over 3 months included assignment to hydrochlorothiazide (versus amlodipine) and higher baseline albuminuria. Overall, declines in eGFR ≥15% (versus <15%) were associated with a 26% higher hazard of cardiovascular outcomes (95% CI, 1.07–1.48). In spline‐based analysis, risk for cardiovascular outcomes was higher in the hydrochlorothiazide arm at every level of decline in eGFR compared with the same magnitude of eGFR decline in the amlodipine arm.

Conclusion

Combined use of benazepril and amlodipine remains superior to benazepril and hydrochlorothiazide for cardiovascular outcomes, regardless of the magnitude of the decline in eGFR that occurred with initiation of therapy.

Nonstandard Abbreviations and Acronyms

ACCOMPLISH

Avoiding Cardiovascular Events through Combination Therapy in Patients Living with Systolic Hypertension

RAS

renin angiotensin system

SBP

systolic blood pressure

Research Perspective.

What Is New?

• This study examines predictors of acute declines in estimated glomerular filtration that may occur with randomized assignment to the initiation of combined antihypertensive therapy (benazepril + amlodipine or hydrochlorothiazide) and its association with cardiovascular outcomes.

What Question Should Be Addressed Next?

• Additional studies are needed to understand the implications of acute declines in estimated glomerular filtration with the use of different combinations of agents on kidney outcomes.

Acute declines in estimated glomerular filtration rate (eGFR) are common during the treatment of hypertension, especially in patients who are starting renin‐angiotensin system (RAS) inhibitors. However, acute declines in eGFR can also occur in the setting of diuretic use. Theoretically, the concomitant initiation of RAS inhibitors with diuretics may increase the risk for the occurrence of acute declines in eGFR during the initiation of combination antihypertensive therapy. ¹

Historically, tolerance of up to a 30% increase in serum creatinine after the initiation of RAS inhibitors has been recommended, because such changes in eGFR are considered to be a hemodynamic response to efferent arteriolar vasodilation. ² , ³ However, in some observational studies, increases in serum creatinine after RAS inhibitor use were associated with an increased risk of myocardial infarction, heart failure, and need for kidney replacement therapy. ⁴ , ⁵ , ⁶ , ⁷ Few studies have examined whether specific combinations of antihypertensive agents are more likely to induce acute declines in eGFR (eg, with concurrent use of a RAS inhibitor and diuretics). Additionally, combined use of amlodipine with RAS inhibition may be preferred by providers, because the eGFR typically improves (rather than declines) when amlodipine is initiated and could theoretically counter the eGFR declines that may occur with RAS inhibition. ⁸

In this study, we had 2 objectives. First, we aimed to identify predictors of acute declines in eGFR in the first 3 months following randomized assignment of patients at high cardiovascular risk to treatment with either benazepril and amlodipine versus benazepril and hydrochlorothiazide in the ACCOMPLISH (Avoiding Cardiovascular Events through Combination Therapy in Patients Living with Systolic Hypertension) trial. We then examined the association between acute declines in eGFR and risk of the cardiovascular outcomes within and across randomized treatment arms.

Methods

Study Population

The ACCOMPLISH trial was a randomized double‐blinded trial in which patients at high risk for cardiovascular events were assigned to receive treatment with benazepril plus amlodipine or benazepril plus hydrochlorothiazide. ⁹ The primary trial end point was a composite cardiovascular outcome. Details of the ACCOMPLISH trial have been previously described and the data that support the findings of this study are available from Novartis. ⁹ , ¹⁰ The trial was prematurely terminated after 36 months when prespecified end points were met and benazepril‐amlodipine was found to be superior to benazepril‐hydrochlorothiazide for the primary composite cardiovascular outcome.

We included ACCOMPLISH trial participants who had a serum creatinine available at month 3 of study (Figure S1) in this analysis. We excluded patients with missing covariates (outside of albuminuria) and missing outcome data, or those who developed the primary outcome within the first 3 months of randomization.

Definition of an Acute Decline in eGFR and the Primary Outcome

During the ACCOMPLISH trial, serum creatinine was ascertained every 3 months in the first year of study. Therefore, we defined the window for determination of the change in eGFR between randomization and month 3 of study. We defined the predictor using 2 different approaches: first, as a binary variable (an acute decline in eGFR ≤15% or >15 in the first 3 months of the trial based on prior published studies ⁶ , ⁷ ) and subsequently, taking a flexible continuous spline‐based approach with 4 knots as done in prior studies. ⁷ , ¹¹

The primary outcome was the composite cardiovascular outcome used in the parent trial, which included death from cardiovascular causes and nonfatal cardiovascular events, including hospitalization for angina, stroke, myocardial infarction, coronary revascularization, and resuscitated sudden cardiac death. ¹⁰

Statistical Analysis

We first compared the characteristics of individuals who had a ≤15% decline in eGFR over the first 3 months of the trial compared with those who had a >15% decline in eGFR. We used t tests, χ² tests, and Kruskal–Wallis tests as appropriate to test for these differences.

Predictors of Acute Declines in eGFR Between Randomization and Month 3 of Follow‐Up

Next, we sought to identify predictors of an acute decline in eGFR that was >15% including covariates such as demographic characteristics (age, sex, race, and ethnicity), baseline systolic blood pressure (SBP), history of coronary artery disease, dyslipidemia, body mass index, diabetes, eGFR, logarithm of albumin/creatinine ratio, and randomized treatment assignment using multivariable logistic regression. Urine albumin/creatinine ratio was ascertained at baseline enrollment in only a subset of patients in the parent trial. Thus, we imputed albumin/creatinine ratio in N=2900 (27%) of participants given the large proportion of missingness in this variable using single‐chained imputation with 20 burn‐in iterations. We identified candidate variables that were predictive of the outcome using a 2‐step, adaptive, least absolute shrinkage and selection operator (LASSO) approach in which cross‐validation was used to choose the penalty parameter that minimized the prediction error. This LASSO approach was applied to the overall cohort, as well as separately by randomized arm. ¹² The LASSO model considered all covariates described above as candidates for inclusion.

Association Between Acute Declines in eGFR and the Primary Cardiovascular Outcome

In our time‐to‐event analysis with Cox proportional hazards models, we examined the association between acute declines in eGFR (≤15% versus >15% decline) and cardiovascular outcomes in the overall trial, and then separately within each treatment arm. Time at risk in these models started at 3 months after randomization and was censored at losses to follow‐up, death, or end of trial. Our models were adjusted for the randomized assignment to amlodipine versus hydrochlorothiazide, demographic, clinical, and laboratory characteristics derived at baseline, in addition to change in systolic blood pressure between baseline and month 3 of study. We then tested for interaction between acute declines in eGFR and randomized assignment.

In addition, we performed adjusted analyses treating acute declines in eGFR as a continuous spline predictor with 4 knots that included an interaction term between randomized assignment to intervention and acute declines in eGFR, and examined its association with risk of cardiovascular outcomes. The splines were generated using a 0% decline in the amlodipine arm as the reference group, but we then qualitatively examined differences in the hazard ratio for cardiovascular events across the same level of decline in the amlodipine and hydrochlorothiazide arms.

All analyses were performed in Stata 17 except for spline analyses, which were conducted in R. Informed consent was obtained from all trial participants by ACCOMPLISH investigators at time of enrollment. The University of California San Francisco Institutional Review Board reviewed and approved this study.

Results

We included a total of 10 714 ACCOMPLISH participants for analysis, which represents 92% of the original trial cohort after excluding those with missing data (Figure S1). A total of 15.8% of individuals experienced an acute decline in eGFR over 3 months (N=1690). The distribution of decline in eGFR is shown in Figure S2 by treatment arm.

We compared characteristics of individuals by the magnitude of the acute decline in eGFR that occurred over the first 3 months (≤15% versus >15%, Table 1). We found that those with a >15% acute decline in eGFR were older, more commonly randomized to receive hydrochlorothiazide as opposed to amlodipine, and had lower SBP at baseline and larger declines in their BP after 3 months. There was no statistically significant difference in the prevalence of baseline diabetes by the level of decline in eGFR that occurred. Those who exhibited a >15% decline in eGFR experienced a median decrease of 16 mL/min per 1.73 m² (Table 1).

Table 1.

Patient Characteristics by Magnitude of the Acute Declines in eGFR That Occurred During the First 3 Months of the Trial

n (Row %), Median (Q1, Q3), or Mean±SD	≤15% Decline	>15% Decline	P value
	N=9024 (84.2%)	N=1690 (15.8%)
Blood pressure treatment
Amlodipine	4784 (89.3%)	575 (10.7%)	<0.001
HCTZ	4240 (79.2%)	1115 (20.8%)
Age at baseline (y)	67.0 (63.0, 73.0)	68.0 (63.0, 74.0)	0.003
Sex
Male	5608 (85.8%)	929 (14.2%)	<0.001
Female	3416 (81.8%)	761 (18.2%)
Race or ethnicity
Hispanic	457 (81.0%)	107 (19.0%)	<0.001
Non‐Hispanic Black	964 (79.3%)	252 (20.7%)
Non‐Hispanic White	7505 (85.1%)	1311 (14.9%)
Other	98 (83.1%)	20 (16.9%)
Baseline BMI (kg/m2)	30.8 ± 6.2	31.4 ± 6.5	0.001
Baseline SBP (mm Hg)	144.3 (133.3 to 158.0)	142.0 (131.3 to 156.0)	<0.001
SBP at 3 mo (mm Hg)	134.0 (124.7 to 145.0)	128.7 (119.0 to 139.0)	<0.001
SBP change (mm Hg) from baseline to 3 mo	−10.0 (−21.3, 1.3)	−13.3 (−26.0 to −2.7)	<0.001
Baseline albumin/creatinine ratio (mg/g)	14.0 (5.0 to,50.9)	17.6 (5.5 to 69.0)	<0.001
Baseline eGFR (mL/min per 1.73 m2)	76.0 (62.7 to 89.3)	78.1 (61.6 to 89.9)	0.039
eGFR at 3 mo (mL/min per 1.73 m2)	78.0 (64.4 to 90.4)	58.8 (46.4 to 70.4)	<0.001
eGFR change (mL/min per 1.73 m2) from baseline to 3 mo	−0.1 (−3.8 to 5.8)	−15.8 (−20.9 to −13.0)	<0.001
Diabetes
Absent	3578 (85.0%)	630 (15.0%)	0.067
Present	5446 (83.7%)	1060 (16.3%)
Dyslipidemia
Absent	2383 (85.5%)	405 (14.5%)	0.036
Present	6641 (83.8%)	1285 (16.2%)
History of cardiovascular disease*
Absent	4077 (84.8%)	731 (15.2%)	0.14
Present	4947 (83.8%)	959 (16.2%)

BMI indicates body mass index; eGFR, estimated glomerular filtration rate; HCTZ, hydrochlorothiazide; and SBP, systolic blood pressure.

^*History of cardiovascular disease includes myocardial infarction, stroke, coronary bypass, or percutaneous coronary intervention.

Predictors of the Occurrence of an Acute Decline in Kidney Function >15%

The predictors of a larger acute decline in eGFR after the initiation of the intervention for the overall trial are shown in Table 2. The predictor that was most strongly associated with the odds of an acute decline in eGFR was randomized assignment to hydrochlorothiazide (versus amlodipine; odds ratio 2.22 [95% CI, 1.99–2.48]). In addition, age, female sex, Black race, lower baseline SBP, higher baseline body mass index, baseline albumin/creatinine ratio, and history of cardiovascular disease predicted higher odds of an acute decline in kidney function, whereas presence of diabetes and dyslipidemia did not (Table 2).

Table 2.

Predictors of an Acute Decline in eGFR >15% Over the First 3 Months After Randomization in ACCOMPLISH in Multivariable Analysis

	OR	95% CI	P value
HCTZ (vs smlodipine)	2.22	[1.99–2.48]	<0.001
Age at baseline (per 5‐y increase)	1.13	[1.08–1.18]	<0.001
Female (vs male)	1.33	[1.19–1.48]	<0.001
Race or ethnicity
Non‐Hispanic Black	1.37	[1.17–1.61]	<0.001
Hispanic	1.24	[0.99–1.55]	0.06
Other	1.21	[0.74–1.97]	0.45
Non‐Hispanic White	Reference
Baseline BMI (per 10 kg/m2 increase)	1.13	[1.03–1.23]	0.007
eGFR at baseline (per 10 mL/min per 1.73 m2 increase)	1.06	[1.03–1.10]	<0.001
SBP at baseline (per 10 mm Hg increase)	0.94	[0.92–0.97]	<0.001
Urine albumin/creatinine ratio (per doubling)	1.04	[1.02–1.06]	<0.001
Diabetes	1.12	[0.98–1.28]	0.09
Dyslipidemia	1.13	[1.00–1.28]	0.06
History of cardiovascular disease*	1.24	[1.10–1.41]	0.001

BMI indicates body mass index; eGFR, estimated glomerular filtration rate; HCTZ, hydrochlorothiazide; OR, odds ratio; and SBP, systolic blood pressure.

^*History of cardiovascular disease includes myocardial infarction, stroke, coronary bypass, or percutaneous coronary intervention.

When we examined predictors of an acute decline in eGFR separately by randomized assignment using a LASSO approach, some differences were noted based on whether patients were assigned to amlodipine versus hydrochlorothiazide (Table S1). For example, baseline dyslipidemia was a statistically significant predictor of the outcome in the hydrochlorothiazide arm, but not the amlodipine arm. Female sex, baseline eGFR, and body mass index were statistically significant predictors of an acute decline in eGFR in the amlodipine arm but not the hydrochlorothiazide arm. However, there was no interaction between these predictors and randomized treatment assignment for predicting a decline in eGFR (all P >0.05).

Association Between Declines in eGFR and Risk of the Primary Outcome

During 2.8 years of median follow‐up (Q1 2.4, Q3 3.3 years), 1024 individuals reached the primary cardiovascular outcome; 11.7% occurred in those with an acute decline in eGFR >15% and 9.2% among those with a ≤15% decline in eGFR.

In the overall trial, the risk of the primary cardiovascular outcome was higher for those with a >15% decline in eGFR (hazard ratio 1.26 [95% CI, 1.07–1.48]) compared with those with a ≤15% decline in eGFR.

Within the amlodipine arm, the hazard ratio for the risk of the primary outcome was higher for those with a >15% decline in eGFR (versus ≤15% decline; hazard ratio 1.47 [95% CI, 1.12–1.92]). This effect size was larger than that in the hydrochlorothiazide arm (hazard ratio 1.17 for a >15% versus ≤15% decline in eGFR [95% CI, 0.96–1.43]; Table 3). However, we did not find an interaction between randomized assignment to hydrochlorothiazide versus amlodipine and the presence of a 15% decline in eGFR in unadjusted (P for interaction=0.23) or adjusted analysis (P for interaction=0.17).

Table 3.

Risk of the Primary Cardiovascular Outcome in the Overall Cohort, and Stratified by Treatment in Adjusted Models

	Overall			Amlodipine			HCTZ
	HR	95% CI	P value	HR	95% CI	P value	HR	95% CI	P value
HCTZ (vs amlodipine)	1.22	[1.08–1.38]	0.002
GFR decline >15%	1.26	[1.07–1.48]	0.005	1.47	[1.12–1.92]	0.006	1.17	[0.96–1.43]	0.12

All models adjusted for age, sex, race, baseline body mass index, eGFR at baseline, baseline SBP and change in SBP between baseline and month 3, baseline albuminuria, diabetes, dyslipidemia, and history of cardiovascular disease. GFR indicates glomerular filtration rate; HCTZ, hydrochlorothiazide; HR, hazard ratio; and SBP, systolic blood pressure.

When we examined the relation between acute declines in eGFR as a continuous spline predictor and risk of the primary outcome, we noted that generally, the risk of the primary cardiovascular outcome was higher in the hydrochlorothiazide arm in adjusted analysis across all levels of eGFR decline compared with the amlodipine arm (Figure 1 and Table S2).

Figure 1. Hazard ratio for the primary composite cardiovascular outcome by both magnitude of the decline in eGFR and randomized assignment in adjusted analyses.

All models adjusted for age, sex, race, baseline body mass index, eGFR at baseline, baseline SBP and change in SBP between baseline and month 3, baseline albuminuria, diabetes, dyslipidemia, and history of cardiovascular disease. *Circles with filled centers are significantly different from the reference point of 0% eGFR change at 3 months (diamond in figure) among patients treated with amlodipine. eGFR indicates estimated glomerular filtration rate; HR, hazard ratio; and SBP, systolic blood pressure.

Discussion

In patients at elevated cardiovascular risk, it is not uncommon for individuals to require more than 1 antihypertensive agent to achieve BP targets. ¹³ , ¹⁴ Depending on the classes of antihypertensive agents that are used, the magnitude of the acute decline in eGFR that occurs may vary substantially. ¹⁵ In this study, we examined the association between acute declines in eGFR that occurred when using a combination of benazepril and a diuretic versus benazepril and a calcium channel blocker. Overall, declines in eGFR were more common in those receiving diuretic compared with amlodipine. Any decline in eGFR >15% was associated with a 26% higher risk of the primary cardiovascular end point compared with those without such eGFR declines in the first 3 months of starting therapy. When we performed stratified analysis in each treatment arm, we observed a 47% higher risk of the primary cardiovascular outcome in those with acute declines in eGFR >15% (versus ≤15%) within the first 3 months of study when using amlodipine, but this risk was not statistically significant in the hydrochlorothiazide arm. However, in spline‐based analysis, the benefit of the amlodipine arm was preserved regardless of the magnitude of the acute decline in eGFR in our spline‐based analysis compared with the same decline in the hydrochlorothiazide arm.

It might have been expected that eGFR declines would be larger when patients were starting a combination of both RAS inhibitors and diuretics simultaneously, which may increase the likelihood of hemodynamic eGFR declines occurring in the setting of potential concurrent volume depletion. ¹¹ The fact that such declines were more common in individuals with larger decreases in their SBP during the same 3‐month timeframe aligns with the potential hemodynamic nature of such changes. Of note, the absolute magnitude of loss of eGFR was clinically significant (average loss of >20 mL/min per 1.73 m²) when the acute decline in eGFR was >15%. However, those with higher SBP were less likely to experience an acute decline in eGFR, the reasons of which were unclear, but possibly related to better ability of patients with hypertension to avoid organ hypoperfusion when BP was lowered.

From an epidemiologic standpoint, it was interesting to observe that the acute declines in eGFR were more strongly associated with the primary outcome in the amlodipine rather than the hydrochlorothiazide arm when conducting within‐arm analysis, though statistical significance for tests of effect modification was not achieved. This observation may stem from the fact that amlodipine initiation is often associated with an increase in eGFR that can counter the efferent vasodilation mediated by RAS inhibition, so patients who experienced acute declines in eGFR despite the use of amlodipine likely have greater severity of illness and have a higher risk of cardiovascular events. For the same reason, when we conduct comparisons of the risk of cardiovascular outcomes across randomized treatment groups in our spline‐based analysis, we believe the lower risk of cardiovascular outcomes observed with amlodipine versus hydrochlorothiazide initiation across any level of decline in eGFR is conservatively estimated, given the greater severity of illness that may be present in those treated with amlodipine who experienced a decline in eGFR. However, in both arms, any acute decline in eGFR that occurred following initiation of the intervention was associated with a higher risk of cardiovascular outcomes.

In the ACCOMPLISH trial, patients who have a baseline history of cardiovascular disease were particularly at risk for these acute declines in eGFR, which is consistent with the results of prior studies. ¹¹ Recent post hoc analyses of a few randomized controlled trials (eg, Action in Diabetes and Vascular Disease: Preterrax and Diamicron Modified Release Controlled Evaluation) showed that acute rises in serum creatinine after starting RAS inhibition was associated with higher risk of cardiovascular, kidney, and mortality outcomes, though the changes in serum creatinine were measured over longer intervals after randomization (eg, 4 months). ¹⁶ In Telmisartan Randomized Assessment Study in ACEi Intolerant Subjects with Cardiovascular Disease (TRANSCEND) and Ongoing Telmisartan Alone and in combination with Ramipril Global Endpoint Trial (ONTARGET) trials, there was also a tendency for acute declines in eGFR within 8 weeks after the initiation of angiotensin‐converting enzyme inhibitors/angiotensin II receptor blockers to have a non–statistically significant tendency towards a higher risk of cardiovascular outcomes in post hoc analysis, ¹⁷ which is consistent with the results of our study. We do note that the duration over which changes in eGFR were ascertained in ONTARGET (a 6‐month period) may be more reflective of natural progression of disease, and less likely to be related to hemodynamic eGFR declines after RAS inhibition.

A strength of the ACCOMPLISH trial is the testing of different combinations of antihypertensive therapy, which allows us to examine the relation between acute declines in eGFR in the setting of combination therapy. In addition, ACCOMPLISH included a large sample size of patients at high risk for cardiovascular outcomes, which allowed us to study this end point, whereas prior studies have primarily focused on the relation between acute declines in eGFR and kidney or all‐cause mortality outcomes. ⁵ , ⁶ , ¹⁸ The reasons why acute declines in eGFR may be associated with adverse cardiovascular outcomes is unclear. It is possible that the occurrence of an acute decline in eGFR during antihypertensive therapy is a signal of worse health status and higher risk for cardiovascular outcomes. While a recent analysis of the ACCOMPLISH trial found that lower cumulative BP and BP variability during follow‐up may have contributed to some of the cardiovascular benefit of amlodipine, ¹⁹ there was still a residual benefit of amlodipine that was not explained by BP. In this study, our models did account for changes in SBP between randomization and month 3 and still identified a benefit to randomized assignment to amlodipine. Further studies are needed to understand these observations.

We highlight several limitations to our study. Although we leverage the randomized assignment of participants to either amlodipine or hydrochlorothiazide, because we were interested in the acute declines in eGFR that occurred postrandomization, our findings are observational and do not maintain the benefits of randomization given the inclusion of a postrandomization exposure. In addition, we imputed albuminuria in approximately one‐quarter of participants due to the lack of measured albuminuria, and we also do not have comprehensive data on changes in albuminuria during follow‐up, which was not consistently ascertained. Our study does not inform whether any antihypertensive therapy should be discontinued after an acute decline in eGFR occurs, since our study is observational in nature and patients with greater severity of illness may be more susceptible to such declines. In addition, it is important to note that discontinuation of therapy was not tested in ACCOMPLISH. Instead, the trial evaluated initiation of benazepril + hydrochlorothiazide versus amlodipine. While we examined predictors of an acute eGFR decline with different combinations of antihypertensive medications in this study, these models warrant validation in future studies. Finally, we are unable to examine the risk of adverse kidney outcomes (eg, onset of the need for kidney replacement therapy) due to the rarer occurrence of this event during the ACCOMPLISH trial.

Conclusions

In conclusion, acute declines in eGFR were more common when hydrochlorothiazide is used in combination with benazepril as opposed to amlodipine. The declines in eGFR after initiation of the intervention were associated with higher risk of cardiovascular events. However, across any level of decline in eGFR that occurred following randomization, benazepril plus amlodipine use was associated with lower risk of cardiovascular events than benazepril plus hydrochlorothiazide.

Sources of Funding

This manuscript is funded by NIH R01 DK121904.

Disclosures

Dr Ku serves as a Board of Trustee member of the American Kidney fund and as a member of the American Kidney Fund Health Equity Coalition, and has received research grant support from CareDX and Natera. Dr Sarnak has received consulting fees from Akebia as a Member of the Steering Committee of a Trial. He has also attended an Advisory Board by Boehringer Ingelheim. His spouse is employed by Eli Lilly. The remaining authors have no disclosures to report.

Supporting information

Data S1