Abstract
Background
Targeting a higher hemoglobin in patients with chronic kidney disease leads to adverse cardiovascular outcomes, yet the reasons remain unclear. Herein, we sought to determine whether changes in erythropoiesis-stimulating agent (ESA) dose and in hemoglobin were predictive of changes in blood pressure (BP) and whether these changes were associated with cardiovascular outcomes.
Methods
In this secondary analysis of 1421 Correction of Hemoglobin and Outcomes in Renal Disease (CHOIR) participants, mixed model analyses were used to describe monthly changes in ESA dose and hemoglobin with changes in diastolic BP (DBP) and systolic BP (SBP). Poisson modeling was performed to determine whether changes in hemoglobin and BP were associated with the composite end point of death or cardiovascular outcomes.
Results
Monthly average DBP, but not SBP, was higher in participants in the higher hemoglobin arm. Increases in ESA doses and in hemoglobin were significantly associated with linear increases in DBP, but not consistently with increases in SBP. In models adjusted for demographics and comorbid conditions, increases in ESA dose (>0 U) and larger increases in hemoglobin (>1.0 g/dL/month) were associated with poorer outcomes [event rate ratio per 1000 U weekly dose per month increase 1.05, (1.02–1.08), P = 0.002 and event rate ratio 1.70 (1.02–2.85), P = 0.05, respectively]. However, increasing DBP was not associated with adverse outcomes [event rate ratio 1.01 (0.98–1.03), P = 0.7].
Conclusion
Among CHOIR participants, higher hemoglobin targets, increases in ESA dose and in hemoglobin were associated both with increases in DBP and with higher event rates; however, increasing DBP was not associated with adverse outcomes.
Keywords: anemia, blood pressure, cardiovascular events, chronic kidney disease, erythropoietin dose
Introduction
Anemia correction trials in chronic kidney disease (CKD) patients have demonstrated that targeting a higher hemoglobin with erythropoiesis-stimulating agents (ESA) leads to adverse outcomes [1–3]; however, the underlying mechanisms remain unknown. Potential mechanisms for adverse outcomes with higher hemoglobin targets include higher blood viscosity, fluctuations in blood volume, a direct toxic effect of higher ESA dose and cardiovascular complications from increased blood pressure (BP) [4, 5]. However, the hypothesis that increased BP among patients randomized to higher hemoglobin targets may lead to adverse outcomes has not been confirmed [6–9].
The administration of ESA (versus placebo) has been demonstrated in human studies to increase BP, yet not consistently [7, 9–12]. In some small studies, increasing hemoglobin by increasing red blood cell (RBC) mass without the use of ESA has not always been shown to influence BP [13], suggesting the hypertensive effect of anemia correction with ESA may be independent of its effects on RBC mass. In vitro, ESA exerts direct vasoconstrictor effects [14, 15] and increases endothelin-1 levels, particularly when high intravenous ESA doses are employed [16, 17]. Furthermore, endothelial cells treated with ESA exhibit decreased nitric oxide (NO) release and a dose-dependent increase in asymmetric dimethyl arginine, an inhibitor of NO [18, 19]. Taken together, these factors suggest that increased ESA doses required to target a higher hemoglobin may contribute to higher BP and excess cardiovascular risk. But, it remains unknown whether the subcutaneous ESA doses used in the Correction of Hemoglobin and Outcomes in Renal Disease (CHOIR) study would exert the same effects.
Considering both large increases in hemoglobin and high-dose ESA may be associated with decreased survival in patients with end-stage renal disease (ESRD) and CKD and may raise BP [20–23], higher BP may be in the mechanistic pathway for harm with higher hemoglobin targets. Thus, we sought to determine whether changes in ESA doses or in hemoglobin were associated with higher BP in participants enrolled in a randomized controlled trial comparing two hemoglobin targets, in the CHOIR study. In addition, we sought to determine whether these changes in ESA dose, hemoglobin or BP were associated with adverse cardiovascular outcomes.
Materials and methods
Description of cohort
CHOIR was a randomized controlled open-label trial comparing the effects of treatment with epoetin alfa to two different hemoglobin targets on the primary composite end point of death, congestive heart failure hospitalization, stroke and myocardial infarction in anemic CKD patients. The methods, baseline characteristics and results have been reported previously [3]. Inclusion criteria included age >18 years, hemoglobin <11.0 g/dL and estimated glomerular filtration rate of 15–50 mL/min/1.73m2 determined by the Modification of Diet in Renal Disease equation. All participants who received at least one dose of ESA were included in this analysis (n = 1421/1432).
Outcome
Analyses were performed using the primary composite end point of death, congestive heart failure hospitalization, stroke and myocardial infarction.
Study measurements
Participants enrolled in CHOIR were randomized to target hemoglobin of 11.3 or 13.5 g/dL based on two different dosing algorithms and were administered weekly or bi-weekly epoetin alfa subcutaneously. Information on ESA dose, hemoglobin and BP was collected on at least a bi-weekly basis. Standardized BP (an average of three readings made in the brachial artery with the patient sitting for at least 5 min and arm at heart level) was obtained by trained personnel at each study visit.
Definitions
ESA dose was calculated as a weekly average of the dose received during 4-week (i.e., 1-month) intervals starting from enrollment until patients experienced an event. All data following an event were excluded. Weekly averages were analyzed to standardize the variables given the different dosing intervals (from 2 × /week up to bi-weekly). To analyze the change in ESA dose, the ‘current’ weekly dose (averaged over 4 weeks) minus the ‘prior’ weekly dose (averaged over 4 weeks) was calculated.
 |
A negative change indicates a reduction in dose, while a positive change indicates an increase in dose. BP was averaged from BP recordings obtained from enrollment and during all visits within 4-week intervals until the occurrence of a primary end point. Change in BP was computed by the average ‘next’ 4-week average BP minus the ‘current’ 4-week average BP. A negative change indicates a decrease in BP and a positive change indicates an increase in BP. Change in hemoglobin was defined the same way as change in BP.
For data used in Poisson modeling of primary event rates, the days in the study were divided into monthly intervals to minimize loss of events for analysis due to missing hemoglobin or dose values prior to the event. The number of events defined the outcome during each monthly interval for each patient. If there was more than one event within the same monthly interval, the first event was used for analysis. If there was an event within the monthly interval, the averages of BP and hemoglobin before the event were used as covariates. If there was no event in the interval, averages of BP and hemoglobin within the whole month were used. The change in the monthly ‘average’ Hb (or diastolic BP, DBP) was derived as the current month Hb (or DBP) minus the previous month Hb (or DBP).
Statistical analysis
Descriptive statistics were performed to describe the average change in ESA dose, hemoglobin and in BP between patients randomized to the high- and low-dose ESA arm.
Repeated mixed effects regression models (using PROC MIXED in SAS) were used to evaluate the effect of change in ESA dose on systolic BP (SBP) and DBP. ESA doses were changes in the weekly averages during the 4-week period immediately prior to the changes in the ‘next’ 4-week averages in SBP and DBP. Random effect of subject was used to account for correlation between measurements within the same patient. Furthermore, these effects were examined separately based on subsets of records in which the ESA dose was either reduced or increased. Smoothed plots (loess in SAS) of the change in average BP versus the change in ESA dose were used to describe the data. Plots of the predicted change in BP based on the mixed models are presented. In additional analyses, we examined the relationships between changes in ESA dose during 2-week intervals with subsequent 2-week changes in SBP and DBP.
To examine the association between the change in hemoglobin and the change in SBP and DBP, we assessed the relationship between the ‘current’ (4-week) minus the ‘prior’ (4-week) hemoglobin with the ‘current’ minus the ‘prior’ SBP and DBP. Otherwise, the methods are similar to those described above assessing dose and BP. In additional analyses, we examined the relationships during 2-week intervals. Finally, we examined the relationship between ‘current’ (4-week) changes in hemoglobin with subsequent ‘next’ (4-week) changes in SBP and DBP.
To assess the associations between BP and each primary event, the monthly primary event rates were examined using descriptive statistics and Poisson regression models with generalized estimating equations (GEE) approach was used to account for clustering effect of multiple measurements within the same patient. The adjusted event rate ratio estimated from the Poisson regression models was used for interpretation of the models. To account for differences across patient-level covariates, Poisson regression was used to model the adjusted relationships between monthly event rates and ESA dose using both average monthly values and change of ESA dose values. Poisson regression was used to model the adjusted relationships between monthly event rates and BP using both average monthly values and change of BP values. The baseline covariates used for these adjustments are the ones previously identified to be significant in survival analyses and included baseline albumin (g/dL), cholesterol (in splines of <240 versus >240 mg/dL), age in linear splines, history of congestive heart failure (composite including history of congestive heart failure, cardiomyopathy, left or right ventricular dilation), history of cerebrovascular accident or transient ischemic attack, prior deep vein thrombosis, history of solid organ malignancy, prior atrial fibrillation or flutter and the use of either an angiotensin receptor blocker or an angiotensin-converting enzyme inhibitor [21]. For linear variables with a non-linear relationship with the primary end point, piecewise linear splines were used and cut-points were determined based on the lowest quasi-likelihood information criterion (QIC). Unger et al. [22] reported an association between increases in hemoglobin >1 g/dL and increased risk of cardiovascular and thromboembolic events, so the cut-point of 1.0 in the rate of change in hemoglobin was fixed in the linear splines for this variable during the search for other cut-points. QIC and QICu statistics proposed by Pan [24] for GEE models were used for goodness of fit and model selection.
All statistical analyses were performed using SAS (version 9.2; SAS Institute, Cary, NC). The institutional review board (IRB) at each center approved the original study and Duke University IRB approved this analysis.
Results
Characteristics of cohort
Baseline characteristics of the participants enrolled in CHOIR have been published previously [3]. On average, the mean age was 66 years and ∼62% had diabetes mellitus (Table 1). Although there were no significant differences in SBP between treatment arms throughout the trial (Figure 1a), monthly average DBP was slightly higher during follow-up and study end among participants randomized to higher hemoglobin targets (Figure 1b). At baseline and throughout the study, there was similar utilization of all classes of anti-hypertensive medications between participants in the higher and lower hemoglobin arms (Table 1).
Table 1.
Baseline characteristics and anti-hypertensive medication utilization among CHOIR participantsa
| Parameter | High hemoglobin 13.5 g/dL (N = 711) | Low hemoglobin 11.3 g/dL (N = 710) | P-value |
|---|---|---|---|
| Age (years), mean (SD) | 65.9 (14.4) | 66.3 (13.5) | 0.97 |
| Female gender, % | 56.4 | 53.8 | 0.33 |
| White race, % | 62.5 | 61.1 | 0.87 |
| Prior DM or DM as cause of CKD, % | 61.5 | 63.8 | 0.39 |
| Baseline BP, mmHg | |||
| SBP | 136.8 (19.7) | 135.6 (19.9) | 0.16 |
| DBP | 71.6 (11.7) | 70.9 (11.1) | 0.23 |
| End of study BPb, mmHg | |||
| SBP | 134.5 (21.1) | 133.0 (19.6) | 0.34 |
| DBP | 71.9 (12.0) | 70.1 (11.4) | 0.004 |
| Baseline medications, % | |||
| ACE-inhibitor and/or ARB | 73.9 | 74.1 | 0.93 |
| Calcium channel blocker | 45.9 | 48.5 | 0.33 |
| Diuretic | 71.8 | 69.8 | 0.43 |
| Beta-blocker (including Labetolol) | 46.8 | 48.1 | 0.63 |
| Other anti-hypertensive | 24.4 | 22.8 | 0.47 |
| Alpha-adrenoreceptor antagonist | 9.3 | 9.8 | 0.73 |
| Minoxidil | 1.3 | 1.3 | 0.99 |
| Clonidine/guanfacine | 12.0 | 11.2 | 0.66 |
| Hydralazine | 4.7 | 4.0 | 0.51 |
| Other anti-hypertensive | 0 | 0 | N/A |
| Post-baseline medications, % | |||
| ACE-inhibitor and/or ARB | 64.6 | 64.9 | 0.91 |
| Calcium channel blocker | 51.5 | 51.1 | 0.87 |
| Diuretic | 70.1 | 69.0 | 0.64 |
| Beta-blocker (including Labetolol) | 47.4 | 49.0 | 0.55 |
| Anti-hypertensive | 29.4 | 28.1 | 0.59 |
| Alpha-adrenoreceptor antagonist | 10.1 | 10.4 | 0.86 |
| Minoxidil | 2.9 | 2.5 | 0.62 |
| Clonidine/guanfacine | 15.2 | 16.5 | 0.51 |
| Hydralazine | 8.2 | 6.9 | 0.36 |
| Other anti-hypertensive | 0.6 | 0.7 | 0.99 |
aACE-I, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; DM, diabetes mellitus.
bThe end of study BP is the last BP measurement before termination for each patient.
Fig. 1.
(a) Monthly average SBP during CHOIR among participants randomized to higher and lower hemoglobin targets. (b) Monthly average DBP during CHOIR among participants randomized to higher and lower hemoglobin targets.
Changes in ESA dose and BP
Of the 1421 participants, the vast majority (n = 1263, 89%) had a reduction in ESA dose at least once during the trial (Table 2). Following a decrease in ESA dose (average of −4486 U), SBP and DBP decreased by −0.79 and −0.51 mmHg, respectively. A similar number of participants (n = 1271, 89%) had an increase in ESA dose (Table 2). Following an increase in ESA dose (4197 U), SBP and DBP increased 0.34 and 0.42 mmHg, respectively. Results of the changes for each randomization arm are included in Table 2.
Table 2.
Average change in monthly SBP and DBP among participants in CHOIR following an ESA dose reduction or an ESA dose increasea
| ESA dose reductions |
ESA dose increases |
|||
|---|---|---|---|---|
| Parameter | Group | Mean (SD) | Group | Mean (SD) |
| Average ESA change (units) | Entire cohort (n = 1263 with 8285 dose reductions) | −4486 (3759) | Entire cohort (n = 1271 with 8923 dose increases) | 4197 (3519) |
| Hb 13.5 (n = 619 with 3931 dose reductions) | −5487 (4143) | Hb 13.5 (n = 644 with 4527 dose increases) | 5166 (3939) | |
| Hb 11.3 (n = 644 with 4354 dose reductions) | −3582 (3110) | Hb 11.3 (n = 627 with 4396 dose increases) | 3198 (2682) | |
| Average SBP change (mmHg) | Entire cohort | −0.79 (12.2) | Entire cohort | 0.34 (11.3) |
| Hb 13.5 | −0.70 (12.1) | Hb 13.5 | 0.23 (11.1) | |
| Hb 11.3 | −0.88 (12.2) | Hb 11.3 | 0.46 (11.6) | |
| Average DBP change (mmHg) | Entire cohort | −0.51 (6.9) | Entire cohort | 0.42 (6.4) |
| Hb 13.5 | −0.51 (6.9) | Hb 13.5 | 0.42 (6.2) | |
| Hb 11.3 | −0.50 (7.0) | Hb 11.3 | 0.41 (6.5) | |
aHb, hemoglobin.
In the mixed model analyses, there was a significant relationship between decreasing ESA dose and subsequent decreases in both SBP (Figure 2a) and DBP (Figure 2b) (P < 0.01 overall and within both the high and low hemoglobin arm). While increasing ESA dose was not consistently associated with an increase in SBP (Figure 2a), increasing ESA dose was significantly associated with an increase in DBP (Figure 2b) (P < 0.05 overall and within both the high and low hemoglobin arm). While our primary analysis looked at the relationship between changes in ESA dose and subsequent changes in BP over 4-week intervals, we also assessed the relationship with 2-week lag intervals. The findings were similar to those presented (data not shown).
Fig. 2.
(a) Average change in SBP based on changes in ESA dose. Following an ESA dose decrease (per 1000 U), average SBP decreased −0.10 ± 0.03 mmHg overall, P = 0.002; −0.16 ± 0.04 mmHg in the higher hemoglobin arm, P = 0.0003, and 0.06 ± 0.05 mmHg in the lower hemoglobin arm, P = 0.3. Following an ESA dose increase (per 1000 U), SBP increased 0.057 ± 0.03 mmHg overall, P = 0.08; 0.044 ± 0.04 mmHg in the higher hemoglobin arm, P = 0.3, and 0.13 ± 0.06 mmHg in the lower hemoglobin arm, P = 0.04. (b) Average change in DBP based on changes in ESA dose. Following an ESA dose decrease (per 1000 U), average DBP decreased −0.059 ± 0.02 mmHg overall, P = 0.003; −0.010 ± 0.03 mmHg in the higher hemoglobin arm, P = 0.007, and −0.05 ± 0.03 mmHg in the lower hemoglobin arm, P = 0.1. Following an ESA dose increase (per 1000 U), DBP increased 0.068 ± 0.02 mmHg overall, P = 0.003; 0.069 ± 0.02 mmHg in the higher hemoglobin arm, P = 0.0025, and 0.082 ± 0.04 mmHg in the lower hemoglobin arm, P = 0.02.
Changes in hemoglobin and BP
During the trial, 1320 participants (93%) had a decrease in hemoglobin (Table 3). Associated with a decrease in hemoglobin (average of −0.70 g/dL), SBP decreased by −0.85 mmHg and DBP decreased by −0.80 mmHg. Throughout the trial, 1378 participants (97%) had an increase in hemoglobin (Table 3). Associated with an increase in hemoglobin (0.75 g/dL), SBP increased by 0.40 mmHg and DBP increased by 0.60 mmHg. Results of the changes for each randomization arm are included in Table 3.
Table 3.
Average change in monthly SBP and DBP among participants in CHOIR associated with a decrease or increase in hemoglobina
| Hemoglobin decreases |
Hemoglobin increases |
|||
|---|---|---|---|---|
| Parameter | Group | Mean (SD) | Group | Mean (SD) |
| Average hemoglobin change (g/dL) | Entire cohort (n = 1320 with 9462 Hb reductions) | −0.70 (0.59) | Entire cohort (n = 1378 with 12 531 Hb increases) | 0.75 (0.57) |
| Hb 13.5 (n = 654 with 4323 Hb reductions) | −0.78 (0.66) | Hb 13.5 (n = 688 with 6522 Hb increases) | 0.80 (0.59) | |
| Hb 11.3 (n = 666 with 5139 Hb reductions) | −0.63 (0.51) | Hb 11.3 (n = 690 with 6009 Hb increases) | 0.69 (0.55) | |
| Average SBP change (mmHg per 1 g/dL change in hemoglobin) | Entire cohort | −0.85 (12.1) | Entire cohort | 0.40 (11.9) |
| Hb 13.5 | −0.69 (12.0) | Hb 13.5 | 0.25 (11.6) | |
| Hb 11.3 | −1.00 (12.2) | Hb 11.3 | 0.56 (12.3) | |
| Average DBP change (mmHg per 1 g/dL change in hemoglobin) | Entire cohort | −0.80 (6.9) | Entire cohort | 0.60 (6.8) |
| Hb 13.5 | −0.74 (6.9) | Hb 13.5 | 0.53 (6.6) | |
| Hb 11.3 | −0.85 (7.0) | Hb 11.3 | 0.68 (7.0) | |
aHb, hemoglobin.
In the mixed model analyses, there was a significant relationship between decreasing hemoglobin and decreases in both SBP (Figure 3a) and DBP (Figure 3b). There was also a significant association between an increase in hemoglobin and an increase in both SBP (Figure 3a) and DBP (Figure 3b). While these relationships were consistently found across both treatment groups for DBP, the associations between changes in Hb and changes in SBP were not significant among participants in the higher hemoglobin arm (Figure 3a and b). The findings were similar when 2-week intervals were used. We also analyzed whether changes in hemoglobin affected subsequent BP over longer time intervals. While our primary findings showed that changes in hemoglobin were directly associated with changes in BP, changes in hemoglobin during 1 month were not associated with changes in BP during subsequent months (data not shown).
Fig. 3.
(a) Average change in SBP graphed per change in hemoglobin. Associated with a decrease in hemoglobin (per 1 g/dL), average decrease in SBP was −0.75 ± 0.20 mmHg overall, P = 0.0003; −0.24 ± 0.27 mmHg in the higher hemoglobin arm, P = 0.37, and −1.66 ± 0.32 mmHg in the lower hemoglobin arm, P < 0.0001. Associated with an increase in hemoglobin (per 1 g/dL), average increase in SBP was 0.47 ± 0.18 mmHg overall, P = 0.01; 0.13 ± 0.24 mmHg in higher hemoglobin arm, P = 0.6, and 1.01 ± 0.28 mmHg in the lower hemoglobin arm, P = 0.0004. (b) Average change in DBP graphed per change in hemoglobin. Associated with a hemoglobin decrease (per 1 g/dL), average DBP decreased −1.13 ± 0.12 mmHg overall, P < 0.0001; −0.94 ± 0.15 mmHg in the higher hemoglobin arm, P < 0.0001, and −1.51 ± 0.19 mmHg in the lower hemoglobin arm, P < 0.0001. Associated with a hemoglobin increase (per 1 g/dL), DBP increased 0.77 ± 0.10 mmHg overall, P < 0.0001; 0.70 ± 0.13 mmHg in the higher hemoglobin arm, P < 0.0001, and 0.092 ± 0.16 mmHg in the lower hemoglobin arm, P < 0.0001.
Changes in ESA dose, hemoglobin, DBP and outcomes
In adjusted models, increases in ESA dose (>0 U) were associated with higher events rates [estimated event rate ratio per 1000 U weekly dose per month increase of 1.05, 95% confidence interval (CI): 1.02–1.08, P = 0.002] (Table 4, Model 1). In adjusted analyses, there was a non-linear relationship between change in hemoglobin and outcomes. Using linear splines, the relationship between the rate of change in hemoglobin and the outcome was piecewise linear where moderate increases in hemoglobin were not associated with the primary end point, while large increases in hemoglobin (with rate of change >1.0 g/dL/month) were linearly associated with poor outcomes (estimated event rate ratio per 1 g/dL increase in monthly hemoglobin rate of 1.70, 95% CI: 1.02–2.85, P = 0.04) (Table 4, Model 2). In adjusted analyses, increases in DBP were not associated with the primary end point (estimated event rate ratio per 1 mmHg of 1.01, 95% CI: 0.98–1.03, P = 0.7) (Table 4, Model 3).
Table 4.
Adjusted event rate ratioa for the primary end point based on changes in ESA dose, DBP and hemoglobinb
| Parameter | Adjusted event rate ratio estimate | Confidence interval | P-value | |
|---|---|---|---|---|
| Model 1c | Change in weekly average ESA dose per month, per 1000 U (>0 increase) | 1.05 | 1.02–1.08 | 0.002 |
| Model 2d | Change in monthly rate of change of hemoglobin with three piecewise linear splines | |||
| <0.7 g/dL | 0.87 | 0.73–1.4 | 0.1 | |
| 0.7–1.0 g/dL | 0.30 | 0.05–1.83 | 0.2 | |
| ≥1.0 g/dL | 1.70 | 1.02–2.85 | 0.04 | |
| Model 3e | Change in monthly average DBP, per 1 mmHg | 1.01 | 0.98–1.03 | 0.7 |
| Model 4af | Change in monthly average DBP, per 1 mmHg | 1.00 | 0.98–1.03 | 0.9 |
| Change in weekly average ESA dose per month, per 1000 U (>0 increase) | 1.03 | 1.00–1.07 | 0.03 | |
| Model 4bg | Change in monthly-average DBP, per 1 mmHg | 1.01 | 0.98–1.03 | 0.7 |
| Change in monthly rate of change of hemoglobin with three piecewise linear spline | ||||
| <0.7 g/dL | 0.86 | 0.72–1.03 | 0.1 | |
| 0.7–1.0 g/dL | 0.26 | 0.04–1.67 | 0.2 | |
| ≥1.0 g/dL | 1.73 | 1.00–2.97 | 0.05 |
aAll models were adjusted for the following: treatment group, baseline albumin, cholesterol, age, prior congestive heart failure, prior cerebrovascular disease or transient ischemic attack, prior deep venous thrombosis, history of solid organ malignancy, prior atrial fibrillation/atrial flutter and the use of an angiotensin-converting enzyme inhibitor or angiotensin receptor blocker.
bHb, hemoglobin. Due to collinearity between changes in ESA dose and changes in hemoglobin, they were not included in models together.
cModel 1 also adjusted for monthly average ESA dose.
dModel 2 also adjusted for monthly average DBP.
eModel 3 also adjusted for monthly average hemoglobin.
fModel 4a also adjusted for monthly average DBP and monthly average ESA dose.
gModel 4b also adjusted for monthly average DBP and monthly average hemoglobin.
There were significant correlations between changes in ESA dose and changes in hemoglobin, thus they were included separately in models with DBP. In adjusted models that included changes in ESA dose and changes in DBP, increases in ESA dose remained associated with higher event rates (estimated event rate ratio 1.04, 95% CI: 1.00–1.07, P = 0.03), while increasing DBP was still not associated with the primary end point (estimated event rate ratio 1.00, 95% CI: 0.98–1.03, P = 0.9) (Table 4, Model 4a). In fully adjusted models that included changes in hemoglobin and changes in DBP, large increases in hemoglobin continued to be associated with higher event rates (estimated event rate ratio 1.73, 95% CI: 1.00–2.97, P = 0.05), while increasing DBP was not predictive of adverse outcomes (estimated event rate ratio 1.01, 95% CI: 0.98–1.03, P = 0.7) (Table 4, Model 4b).
Discussion
In this secondary analysis of the CHOIR trial, higher hemoglobin targets, greater increases in ESA dose and larger increases in hemoglobin were associated with higher DBP and with higher event rates. However, higher DBP was not in turn associated with adverse outcomes. Thus, these hypothesis-generating findings do not support the notion that increases in DBP are in the causal pathway for the detrimental outcomes observed among CHOIR participants targeted to a higher hemoglobin goal.
Prior studies have not consistently identified differences in BP in participants randomized to higher versus lower hemoglobin targets. In a small substudy of 60 ESRD participants in the Normalization of Hematocrit study [25], no difference in ambulatory BP or routine BP parameters between participants randomized to higher versus lower hematocrit arms was seen [26]. The CREATE (Cardiovascular risk Reduction by Early Anemia Treatment with Epoetin β) study which randomized participants with CKD to early versus late initiation of ESA also did not demonstrate a difference in mean BP between groups at the end of the study. However, during the course of the study, there was a 10% increased incidence of hypertension (defined as SBP >160 mmHg) among participants randomized to the higher hemoglobin arm [27]. The TREAT study (Trial to Reduce Cardiovascular Events with Aranesp Therapy), which randomized participants to a hemoglobin of 13 g/dL with ESA versus placebo, did not find a difference in SBP during the trial but identified higher DBP in those assigned to ESA versus placebo (73 versus 71 mmHg, respectively) [1]. Finally, a recent meta-analysis of anemia correction trials in CKD patients identified a higher cumulative relative risk for hypertension among patients targeted to a higher hemoglobin [11]. In the study reported here, no difference in SBP was seen between treatment arms, but DBP was higher over time in participants randomized to the higher hemoglobin target. Thus, it appears that anemia correction with ESA therapy tends to increase BP in CKD patients.
In this study, one reason for higher DBP observed in participants randomized to higher hemoglobin appears to be due to the higher ESA dose requirements in these participants. This is consistent with prior studies which have demonstrated ESA to exert a vasoconstrictor effect in vitro and in vivo. In vitro, ESA stimulates vascular endothelial cells to increase endothelin-1 (ET-1) release, a potent vasoconstrictor, and ET-1 release is increased with higher ESA doses [17]. In vitro, endothelial cells treated with ESA exhibit decreased nitric oxide (NO), a vasodilator, and a dose-dependent increase in asymmetric dimethyl arginine (an inhibitor of NO) [18, 19]. In in vivo studies among predialysis and ESRD patients, intravenous (IV) administration of ESA in increasing doses thrice weekly has been associated with increasing incidences of hypertension [28, 29]. Thus, while we only identified a modest relationship between increasing ESA doses and increases in BP, our analysis extends these prior studies and demonstrates that varying ESA doses are associated with changes in BP.
In this study, the other reason for higher DBP in participants targeted to higher hemoglobin appears to be partly explained by greater increases in hemoglobin. However, previous studies have not consistently shown a relationship between raising RBC mass and raising DBP. In a small prospective study which raised hemoglobin from 0.25 to 0.32 g/dL in 15 hemodialysis patients with the use of IV iron, there was no significant effect of increasing RBC mass on either SBP or DBP; however this study was likely underpowered to demonstrate a significant difference [13]. In a secondary analysis of 118 hemodialysis patients randomized to placebo versus low-dose ESA versus high-dose ESA, the use of ESA was associated with an increase in DBP which correlated with the rise in hemoglobin [30]. In our study, we identified a stepwise increase in DBP with ESA-induced increases in hemoglobin. These changes appear to occur concurrently as we failed to find a persistent effect of increases in hemoglobin on subsequent BP during longer time intervals. While increasing hemoglobin appears associated with increased DBP, the mechanisms are uncertain. Some have hypothesized that increased total blood volume may contribute to higher BP, however, compensatory decreases in plasma volume should occur and minimize increases in BP. This may partially explain why we found direct associations between hemoglobin and BP which did not persist with longer time intervals. Alternatively, an increase in peripheral vascular resistance with higher blood viscosity and increased inhibition of hypoxia-mediated small vessel vasodilation may underlie increased DBP with higher hemoglobin target [31].
Despite identifying increased DBP in participants with higher hemoglobin targets, increasing DBP did not appear to be a mediator of risk in those randomized to the higher target. While higher DBP may be a marker of increased small artery vascular resistance, high DBP has not been strongly associated with outcomes in many large epidemiological studies of patients with high cardiovascular risk [32]. In fact, increasing evidence suggests very low DBP and an increase in pulse pressure (with loss of distensibility of the larger arteries) appears to be a stronger mortality predictor [33, 34]. Thus, while DBP may be increased with higher hemoglobin targets, this does not appear to explain the greater adverse outcomes among patients randomized to higher hemoglobin targets.
This study is not without limitations. As a secondary analysis of a randomized trial, no causal relationships can be inferred from the observed associations. In addition, anti-hypertensive medications were increased at the discretion of local investigators as deemed necessary for maintaining BP control. Furthermore, although medication data (i.e. name of agent) was collected during the course of the trial, dose adjustments were not collected precluding an assessment of whether anti-hypertensive medications were intensified. It remains possible that larger differences in BP would have been observed if changes to anti-hypertensive medications had required protocol assessments. In addition, the BP used in these analyses was collected as part of routine clinical care and we observed small changes in BP which were statistically significant but may not be clinically significant. Given ambulatory BP and central BP measurements, which are increasingly recognized as better predictors of BP burden, were not measured as part of this study, our results may underrepresent BP burden within individual participants. However, such misclassification should have been similar in both groups of participants.
In summary, although higher hemoglobin targets, increasing ESA doses and larger increases in hemoglobin are associated with higher DBP and higher event rates in CHOIR, this study did not identify higher DBP to be associated with a greater risk of adverse outcomes. Nonetheless, these findings suggest that raising hemoglobin with ESA in CKD patients can increase DBP and judicious attention to BP control in patients initially prescribed ESA or in whom ESA doses are being adjusted is warranted.
Conflict of interest statement
None declared.
Acknowledgements
Funding. The original CHOIR trial was supported by Ortho Biotech Clinical Affairs and Johnson & Johnson Pharmaceutical Research and Development, both subsidiaries of Johnson & Johnson. This analysis was supported by grants from the NIH (R01 DK080094-01A1 and K23 HL092297).
References
- 1.Pfeffer MA, Burdmann EA, Chen CY, et al. A trial of darbepoetin alfa in type 2 diabetes and chronic kidney disease. N Engl J Med. 2009;361:2019–2032. doi: 10.1056/NEJMoa0907845. [DOI] [PubMed] [Google Scholar]
- 2.Phrommintikul A, Haas SJ, Elsik M, et al. Mortality and target haemoglobin concentrations in anaemic patients with chronic kidney disease treated with erythropoietin: a meta-analysis. Lancet. 2007;369:381–388. doi: 10.1016/S0140-6736(07)60194-9. [DOI] [PubMed] [Google Scholar]
- 3.Singh AK, Szczech L, Tang KL, et al. Correction of anemia with epoetin alfa in chronic kidney disease. N Engl J Med. 2006;355:2085–2098. doi: 10.1056/NEJMoa065485. [DOI] [PubMed] [Google Scholar]
- 4.Fishbane S, Besarab A. Mechanism of increased mortality risk with erythropoietin treatment to higher hemoglobin targets. Clin J Am Soc Nephrol. 2007;2:1274–1282. doi: 10.2215/CJN.02380607. [DOI] [PubMed] [Google Scholar]
- 5.Yang W, Israni RK, Brunelli SM, et al. Hemoglobin variability and mortality in ESRD. J Am Soc Nephrol. 2007;18:3164–3170. doi: 10.1681/ASN.2007010058. [DOI] [PubMed] [Google Scholar]
- 6.Abraham PA, Macres MG. Blood pressure in hemodialysis patients during amelioration of anemia with erythropoietin. J Am Soc Nephrol. 1991;2:927–936. doi: 10.1681/ASN.V24927. [DOI] [PubMed] [Google Scholar]
- 7.Krapf R, Hulter HN. Arterial hypertension induced by erythropoietin and erythropoiesis-stimulating agents (ESA) Clin J Am Soc Nephrol. 2009;4:470–480. doi: 10.2215/CJN.05040908. [DOI] [PubMed] [Google Scholar]
- 8.Lebel M, Kingma I, Grose JH, et al. Hemodynamic and hormonal changes during erythropoietin therapy in hemodialysis patients. J Am Soc Nephrol. 1998;9:97–104. doi: 10.1681/ASN.V9197. [DOI] [PubMed] [Google Scholar]
- 9.Raine AE, Roger SD. Effects of erythropoietin on blood pressure. Am J Kidney Dis. 1991;18:76–83. [PubMed] [Google Scholar]
- 10.Cody J, Daly C, Campbell M, et al. Recombinant human erythropoietin for chronic renal failure anaemia in pre-dialysis patients. Cochrane Database Syst Rev. 2005;(3) doi: 10.1002/14651858.CD003266.pub2. CD003266. [DOI] [PubMed] [Google Scholar]
- 11.Palmer SC, Navaneethan SD, Craig JC, et al. Meta-analysis: erythropoiesis-stimulating agents in patients with chronic kidney disease. Ann Intern Med. 2010;153:23–33. doi: 10.7326/0003-4819-153-1-201007060-00252. [DOI] [PubMed] [Google Scholar]
- 12.Strippoli GF, Craig JC, Manno C, et al. Hemoglobin targets for the anemia of chronic kidney disease: a meta-analysis of randomized, controlled trials. J Am Soc Nephrol. 2004;15:3154–3165. doi: 10.1097/01.ASN.0000145436.09176.A7. [DOI] [PubMed] [Google Scholar]
- 13.Kaupke CJ, Kim S, Vaziri ND. Effect of erythrocyte mass on arterial blood pressure in dialysis patients receiving maintenance erythropoietin therapy. J Am Soc Nephrol. 1994;4:1874–1878. doi: 10.1681/ASN.V4111874. [DOI] [PubMed] [Google Scholar]
- 14.Heidenreich S, Rahn KH, Zidek W. Direct vasopressor effect of recombinant human erythropoietin on renal resistance vessels. Kidney Int. 1991;39:259–265. doi: 10.1038/ki.1991.31. [DOI] [PubMed] [Google Scholar]
- 15.Vaziri ND, Zhou XJ, Smith J, et al. In vivo and in vitro pressor effects of erythropoietin in rats. Am J Physiol. 1995;269:F838–F845. doi: 10.1152/ajprenal.1995.269.6.F838. [DOI] [PubMed] [Google Scholar]
- 16.Carlini R, Obialo CI, Rothstein M. Intravenous erythropoietin (rHuEPO) administration increases plasma endothelin and blood pressure in hemodialysis patients. Am J Hypertens. 1993;6:103–107. doi: 10.1093/ajh/6.2.103. [DOI] [PubMed] [Google Scholar]
- 17.Carlini RG, Dusso AS, Obialo CI, et al. Recombinant human erythropoietin (rHuEPO) increases endothelin-1 release by endothelial cells. Kidney Int. 1993;43:1010–1014. doi: 10.1038/ki.1993.142. [DOI] [PubMed] [Google Scholar]
- 18.Scalera F, Kielstein JT, Martens-Lobenhoffer J, et al. Erythropoietin increases asymmetric dimethylarginine in endothelial cells: role of dimethylarginine dimethylaminohydrolase. J Am Soc Nephrol. 2005;16:892–898. doi: 10.1681/ASN.2004090735. [DOI] [PubMed] [Google Scholar]
- 19.Wang XQ, Vaziri ND. Erythropoietin depresses nitric oxide synthase expression by human endothelial cells. Hypertension. 1999;33:894–899. doi: 10.1161/01.hyp.33.3.894. [DOI] [PubMed] [Google Scholar]
- 20.Singh AK. The FDA's perspective on the risk for rapid rise in hemoglobin in treating CKD anemia: Quo Vadis. Clin J Am Soc Nephrol. 2010;5:553–556. doi: 10.2215/CJN.00490110. [DOI] [PubMed] [Google Scholar]
- 21.Szczech LA, Barnhart HX, Inrig JK, et al. Secondary analysis of the CHOIR trial epoetin-alpha dose and achieved hemoglobin outcomes. Kidney Int. 2008;74:791–798. doi: 10.1038/ki.2008.295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Unger EF, Thompson AM, Blank MJ, et al. Erythropoiesis-stimulating agents—time for a reevaluation. N Engl J Med. 2010;362:189–192. doi: 10.1056/NEJMp0912328. [DOI] [PubMed] [Google Scholar]
- 23.Zhang Y, Thamer M, Stefanik K, et al. Epoetin requirements predict mortality in hemodialysis patients. Am J Kidney Dis. 2004;44:866–876. [PubMed] [Google Scholar]
- 24.Pan W. Akaike's information criterion in generalized estimating equations. Biometrics. 2001;57:120–125. doi: 10.1111/j.0006-341x.2001.00120.x. [DOI] [PubMed] [Google Scholar]
- 25.Besarab A, Bolton WK, Browne JK, et al. The effects of normal as compared with low hematocrit values in patients with cardiac disease who are receiving hemodialysis and epoetin. N Engl J Med. 1998;339:584–590. doi: 10.1056/NEJM199808273390903. [DOI] [PubMed] [Google Scholar]
- 26.Conlon PJ, Kovalik E, Schumm D, et al. Normalization of hematocrit in hemodialysis patients with cardiac disease does not increase blood pressure. Ren Fail. 2000;22:435–444. doi: 10.1081/jdi-100100885. [DOI] [PubMed] [Google Scholar]
- 27.Drueke TB, Locatelli F, Clyne N, et al. Normalization of hemoglobin level in patients with chronic kidney disease and anemia. N Engl J Med. 2006;355:2071–2084. doi: 10.1056/NEJMoa062276. [DOI] [PubMed] [Google Scholar]
- 28.Double-blind, placebo-controlled study of the therapeutic use of recombinant human erythropoietin for anemia associated with chronic renal failure in predialysis patients. The US Recombinant Human Erythropoietin Predialysis Study Group. Am J Kidney Dis. 1991;18:50–59. doi: 10.1016/s0272-6386(12)80290-3. [DOI] [PubMed] [Google Scholar]
- 29.Pollok M, Bommer J, Gurland HJ, et al. Effects of recombinant human erythropoietin treatment in end-stage renal failure patients. Results of a multicenter phase II/III study. Contrib Nephrol. 1989;76:201–211. doi: 10.1159/000417896. discussion 12–18. [DOI] [PubMed] [Google Scholar]
- 30.Effect of recombinant human erythropoietin therapy on blood pressure in hemodialysis patients. Canadian Erythropoietin Study Group. Am J Nephrol. 1991;11:23–26. doi: 10.1159/000168267. [DOI] [PubMed] [Google Scholar]
- 31.Narayan P, Papademetriou V, Wachtell K, et al. Association of hemoglobin delivery with left ventricular structure and function in hypertensive patients: Losartan intervention for End Point Reduction in Hypertension Study. Hypertension. 2006;47:868–873. doi: 10.1161/01.HYP.0000215578.40687.6f. [DOI] [PubMed] [Google Scholar]
- 32.Kengne AP, Czernichow S, Huxley R, et al. Blood pressure variables and cardiovascular risk: new findings from ADVANCE. Hypertension. 2009;54:399–404. doi: 10.1161/HYPERTENSIONAHA.109.133041. [DOI] [PubMed] [Google Scholar]
- 33.Klassen PS, Lowrie EG, Reddan DN, et al. Association between pulse pressure and mortality in patients undergoing maintenance hemodialysis. JAMA. 2002;287:1548–1555. doi: 10.1001/jama.287.12.1548. [DOI] [PubMed] [Google Scholar]
- 34.Panagiotakos DB, Kromhout D, Menotti A, et al. The relation between pulse pressure and cardiovascular mortality in 12,763 middle-aged men from various parts of the world: a 25-year follow-up of the seven countries study. Arch Intern Med. 2005;165:2142–2147. doi: 10.1001/archinte.165.18.2142. [DOI] [PubMed] [Google Scholar]