Embed Size (px)
Citation preview
Original Investigation
Erythropoiesis-Stimulating Agent Responsiveness and Mortalityin Hemodialysis Patients: Results from a Cohort Study From the
Dialysis Registry in Japan
Shingo Fukuma, MD,1 Takuhiro Yamaguchi, PhD,2 Seiji Hashimoto, MD, PhD,3
Shigeru Nakai, MD, PhD,3 Kunitoshi Iseki, MD, PhD,3
Yoshiharu Tsubakihara, MD, PhD,3 and Shunichi Fukuhara, MD1
Background: Patient responsiveness to erythropoiesis-stimulating agents (ESAs), notoriously difficult tomeasure, has attracted attention for its association with mortality. We defined categories of ESA responsive-ness and attempted to clarify their association with mortality.
Study Design: Cohort study.Setting & Participants: Data from Japan’s dialysis registry (2005-2006), including 95,460 adult hemodialy-
sis patients who received ESAs.Predictor: We defined 6 categories of ESA responsiveness based on a combination of ESA dosage (low
[�6,000 U/wk] or high [�6,000 U/wk]) and hemoglobin level (low [�10 g/dL], medium [10-11.9 g/dL], or high[�12 g/dL]), with medium hemoglobin level and low-dose ESA therapy as the reference category.
Outcomes: All-cause and cardiovascular mortality during 1-year follow-up.Measurements: HRs were estimated using a Cox model for the association between responsiveness
categories and mortality, adjusting for potential confounders such as age, sex, postdialysis weight, dialysisduration, comorbid conditions, serum albumin level, and transferrin saturation.
Results: Median ESA dosage (4,500-5,999 U/wk) was used as a cutoff point, and mean hemoglobin levelwas 10.1 g/dL in our cohort. Of 95,460 patients during follow-up, 7,205 (7.5%) died of all causes, including5,586 (5.9%) cardiovascular deaths. Low hemoglobin levels and high-dose ESA therapy were both associatedwith all-cause mortality (adjusted HRs, 1.18 [95% CI, 1.09-1.27] for low hemoglobin level with low-dose ESAand 1.44 [95% CI, 1.34-1.55] for medium hemoglobin level with high-dose ESA). Adjusted HRs for high-doseESA with low hemoglobin level (hyporesponsiveness) were 1.94 (95% CI, 1.82-2.07) for all-cause and 2.02(95% CI, 1.88-2.17) for cardiovascular mortality. We also noted the interaction between ESA dosage andhemoglobin level on all-cause mortality (likelihood ratio test, P � 0.002).
Limitations: Potential residual confounding from unmeasured factors and single measurement of predic-tors.
Conclusions: Mortality can be affected by ESA responsiveness, which may include independent andinteractive effects of ESA dose and hemoglobin level. Responsiveness category has prognostic importanceand clinical relevance in anemia management.Am J Kidney Dis. 59(1):108-116. © 2011 by the National Kidney Foundation, Inc.
INDEX WORDS: Erythropoiesis-stimulating agent responsiveness; mortality; hemodialysis.
Erythropoiesis-stimulating agent (ESA) respon-siveness has attracted attention for its associa-
tion with mortality in hemodialysis (HD) patients.1-3
Responsiveness can vary widely, and hyporesponsive-
From the 1Department of Epidemiology and Healthcare Research,Graduate School of Medicine and Public Health, Kyoto University,Kyoto, Japan; 2Biostatistics, Tohoku University Graduate School ofMedicine, Sendai; and 3Committee of Renal Data Registry of the Japa-nese Society for Dialysis Therapy, Tokyo, Japan.
Received March 10, 2011. Accepted in revised form July 13,2011. Originally published online September 5, 2011.
Address correspondence to Shingo Fukuma, MD, Department ofEpidemiology and Healthcare Research, Graduate School of Medi-cine and Public Health, Kyoto University, Yoshida-Konoe-cho,Sakyo-ku, Kyoto 606-8501, Japan. E-mail: [email protected]
© 2011 by the National Kidney Foundation, Inc.0272-6386/$36.00
doi:10.1053/j.ajkd.2011.07.014108
ness in particular is believed to predict poorer out-comes than other conditions. Both high-dose ESAtherapy and low hemoglobin levels may contribute topoorer outcomes in patients with ESA hyporesponsive-ness.1,4-6 However, a recent meta-analysis and system-atic review7,8 were unable to determine a recom-mended hemoglobin level suitable for use in clinicalguidelines9-11 or estimate the independent effects ofESA dose on mortality. In addition, the impact of ESAdose–hemoglobin level interaction on mortality is notprecisely understood.
ESA responsiveness is relative, not absolute; the 2determinants of responsiveness are ESA dose andhemoglobin level. Although previous studies havedefined ESA responsiveness based on ESA dose,1
hematocrit change per increase in ESA dose,2 orerythropoietin resistance index,3,12 these indicatorsare rarely used in practice because physicians com-
monly adjust ESA dosage according to hemoglobinAm J Kidney Dis. 2012;59(1):108-116
ESA Responsiveness and Mortality
level (or hematocrit), and the association betweenESA dose and hemoglobin level is not linear. TheNational Kidney Foundation– KDOQI (Kidney Dis-ease Outcomes Quality Initiative) and European guide-lines define ESA hyporesponsiveness as failing toachieve target hemoglobin levels while receiving anESA dosage �500 U/kg/wk.9,10 However, given thatthe mean ESA dosage in Japan is less than one thirdthat in the United States,13 whether the associationbetween ESA responsiveness and mortality is similarfor lower ESA dosages is uncertain.
We attempted to clarify the impact of ESA respon-siveness, including the effects of ESA dose and hemo-globin level, on mortality. We defined ESA responsive-ness categories by combining ESA dose andhemoglobin level and then examined the associationbetween category and mortality at 1 year in a largecohort using Japan’s nationwide dialysis registry.
METHODS
StudyPopulation andData Source
This cohort study included HD patients who were older than 20years, had spent at least 3 months on HD therapy, received ESAs,and had available data for hemoglobin levels. We excluded pa-tients with multiple myeloma and polycystic kidney disease be-cause previous studies have reported that response to ESAs differsbetween these patients and others.14,15 We also excluded patientswith acute cardiovascular disease (CVD) at baseline because theyconfound the association between ESA responsiveness and mortal-ity. Ultimately, 95,460 enrolled patients were analyzed.
Data were obtained from the nationwide dialysis registry of theJapanese Society for Dialysis Therapy, a database that containsdata for large numbers of HD patients in Japan, including demo-graphic information (eg, age, sex, time on dialysis therapy, primarycause of end-stage renal disease, diabetes, history of CVD, postdi-alysis body weight, and weekly ESA dosage) and clinical data (eg,hemoglobin, serum albumin, C-reactive protein [CRP], transferrinsaturation, and ferritin). The precise design of the dialysis registryhas been described previously.16 We used standard analysis filecoded as JRDR-08005. Baseline demographic data for all patientswere collected from the database in December 2005. All ESAsused in the present study were epoetin � or epoetin �, and ESAdosage was reported as categorical data (units/week; �1,500,1,500-2,999, 3,000-4,499, 4,500-5,999, 6,000-8,999, or �9,000).Data regarding route and frequency of ESA administration werenot documented in the registry. Mortality data (time and cause)were collected from the database between December 2005 andDecember 2006 (follow-up, 1-12 months).
Definitionof Exposure: ESAResponsiveness
We defined ESA dosage as 2 categories (�6,000 or �6,000U/wk) according to the median values in our cohort (4,500-5,999U/wk) and hemoglobin level as 3 categories (�10, 10-11.9 or �12g/dL) according to clinical guidelines.11 ESA responsiveness thuswas allotted to 6 categories combining the categorical determi-nants of ESA dosage (2 groups) and hemoglobin level (3 groups).We combined the 2 categorical variables and used them as expo-sure categories to evaluate the biological interaction between ESAdosage and hemoglobin level. We also defined 12 categories ofESA responsiveness based on narrower categories of ESA dosage
(�3,000, 3,000-5,999, 6,000-8,999, or �9,000 U/wk) and catego-Am J Kidney Dis. 2012;59(1):108-116
ries of hemoglobin level (�10, 10-11.9, or �12 g/dL) to examine adose-dependent association.
Outcomes
The primary outcome was all-cause mortality rate, and thesecondary outcome was cardiovascular mortality rate. Interna-tional Classification of Diseases, 10th Revision codes were used todefine cardiovascular mortality (codes I10-I79). Cardiovasculardeath included sudden death; death from heart failure, myocardialinfarction, or stroke; and death from other vascular disease.
Statistical Analysis
Exploratory multivariable logistic regression analysis was per-formed to determine which factors may be used to predict ESA-hyporesponsive patients. We defined ESA hyporesponsiveness ashaving hemoglobin level �10 g/dL while receiving an ESA dosage�6,000 U/wk. The model included all potential risk factors, suchas age, sex, time on dialysis therapy, history of CVD, diabetes,postdialysis body weight, transferrin saturation, and CRP andserum albumin levels. Odds ratios and their 95% confidenceintervals (CIs) were estimated.
For each category of ESA responsiveness, we calculated mortal-ity rate and 95% CIs for all-cause and CVD-associated mortality at1 year. We conducted Cox proportional regression analysis toestimate hazard ratios (HRs) and their 95% CIs for the associationbetween categories of ESA responsiveness and mortality for all-cause and CVD-associated mortality. Patients who underwenttransplant or changed modalities were treated as censored.
Categories of ESA responsiveness were included in models bycreating multiple dichotomous variables. Patients with hemoglo-bin levels of 10-11.9 g/dL receiving an ESA dosage �6,000 U/wkwere our reference category because evidence suggests that thiscategory has the lowest mortality risk.2 For Cox regression, ourmodel included adjustment for age, sex, time on dialysis therapy,postdialysis body weight, history of CVD, diabetes, serum albu-min level, and transferrin saturation. Given that previous studiesfound these covariates to be potential confounding factors,1-3,5 wealso examined a model including these factors with ferritin level,CRP level, single-pool Kt/V, and normalized protein nitrogenappearance. Each variable in the model was checked and logtransformation or construction of categorical variables was usedwhen appropriate. We assessed the statistical interaction betweenESA dosage and hemoglobin level using the likelihood ratio testcomparing the Cox model with and without the product terms. Wealso assessed the biological interaction between high-dose ESAtherapy and low hemoglobin level calculating the relative excessrisk due to interaction (RERI). We calculated RERI using adjustedHRs for all-cause mortality.17,18 If there is no biological interac-tion, RERI is equal to 0.
We also defined 12 categories of ESA responsiveness that havenarrower categories of ESA dosage, and used Cox regressionanalysis to examine whether the association between ESA dosageand mortality was dose dependent across hemoglobin levels.Patients with hemoglobin levels �10 g/dL and ESA dosage �3,000U/wk were used as the reference category.
We also performed subgroup analysis according to transferrinsaturation (�20% or �20%), ferritin level (�200 or �200 ng/mL), and age (�65 or �65 years). The cutoff points for thesefactors were defined according to clinical guidelines.11 We usedCox regression analysis to examine whether the association be-tween ESA responsiveness and mortality was consistent withinthese subgroups.
A multiple imputation approach using chained equations wasused to account for missing covariates. Results were similar to
those obtained on complete set analysis.109
Fukuma et al
All analyses were performed using STATA (version 11.0; STATA,www.stata.com) and JMP (version 8.0.2; SAS Institute, www.sas.com) software, with 2-sided significance set at 0.05.
RESULTS
StudyParticipants andBaselineCharacteristics
A total of 95,460 HD patients were included in thisstudy, and our process of selecting participants fromthe database is shown in Fig 1. Baseline characteris-tics (overall and based on ESA responsiveness cat-egory) are listed in Table 1.
With regard to patient characteristics, mean agewas 64.3 years, 59.0% were men, and 32.5% werefound to have diabetes. Median ESA dosage was4,500-5,999 U/wk, and mean hemoglobin level was10.1 � 1.2 (standard deviation) g/dL. In general, the20,939 patients (22%) with low hemoglobin levelswho received high-dose ESA therapy were older,more frequently had a history of CVD, and had lowerserum albumin levels, higher CRP levels, and lowertransferrin saturations than other patients.
FactorsAssociatedWith ESAHyporesponsiveness
Results of exploratory multivariable logistic regres-sion analysis to determine predictors of ESA hypore-sponsiveness are listed in Table 2. After including95,460 patients in exploratory logistic regression,factors found to be associated with ESA hyporespon-siveness included older age, being female, lengthytime on dialysis therapy, history of CVD, transferrinsaturation �20%, elevated CRP level, and decreasedserum albumin level. However, neither diabetes statusnor postdialysis body weight was associated withresponsiveness.
OutcomeData
During the 1,103,393 person-months of follow-up(range, 1-12 months), 7,205 patients (7.5%) died of allcauses, including 5,586 (5.9%) who died of CVD. Of95,460 total patients, data for 202 (�0.1%) were
censored because of transplant (n � 178) and modal-110
ity change (n � 24). Rates of all-cause and CVD-associated mortalities at 1 year postexamination basedon ESA responsiveness category are listed in Table 3.The all-cause mortality rate at 1 year postexaminationwas highest in patients with hemoglobin levels �10g/dL receiving ESA dosages �9,000 U/wk (165.6deaths/1,000 person-years) and lowest in those withhemoglobin levels of 10-11.9 g/dL receiving ESAdosages �3,000 U/wk (43.0 deaths/1,000 person-years).
ESAResponsiveness andMortality
Table 4 lists unadjusted and multivariable-adjustedassociations between category of ESA responsivenessand mortality for all-cause and CVD-associated deaths.After adjusting for potential confounding factors, suchas age, sex, time on dialysis therapy, postdialysis bodyweight, diabetes, history of CVD, serum albuminlevel, and transferrin saturation, low hemoglobin leveland high-dose ESA therapy were found to be associ-ated separately with all-cause mortality. In particular,we observed the highest all-cause mortality rate in thepresent study in patients with low hemoglobin levelsreceiving high-dose ESA therapy (adjusted HR, 1.94;95% CI, 1.82-2.07), a finding that was unchangedeven after adjusting for ferritin level (HR, 1.92; 95%CI, 1.80-2.05), CRP level (HR, 1.76; 95% CI, 1.65-1.88), single-pool Kt/V (HR, 1.84; 95% CI, 1.73-1.96), and normalized protein nitrogen appearance(HR, 1.89; 95% CI, 1.77-2.02). Assessment of theinteraction between ESA dose and hemoglobin levelclearly showed statistical significance (likelihood ra-tio test, P � 0.002). The RERI between low hemoglo-bin level and high-dose ESA therapy was 0.32(RERI � 1.94�1.18�1.44 � 1). RERI is calculatedusing HRs of 2 risk factors. In this study, we defined 2risk factors as high ESA dosage and low hemoglobinlevel. This method allows us to consider biologicalinteraction, which is the departure from the additivityof 2 risk factors. RERI � 0 indicates positive biologi-
Figure 1. Selection process for studypopulation. Abbreviations: CVD, cardio-vascular disease; ESA, erythropoiesis-stimulating agent; HD, hemodialysis.
cal interaction.
Am J Kidney Dis. 2012;59(1):108-116
Table 1. Baseline Demographics
VariableOverall
(n � 95,460)
ESA <6,000 U/wk (n � 56,521) ESA >6,000 U/wk (n � 38,939)
Hb <10 g/dL(n � 18,028)
Hb 10-11.9 g/dL(n � 34,660)
Hb >12 g/dL(n � 3,833)
Hb <10 g/dL(n � 20,939)
Hb 10-11.9 g/dL(n � 16,705)
Hb >12 g/dL(n � 1,295)
Age (y) 64.3 � 12.6 64.6 � 12.6 63.2 � 12.8 62.9 � 13.2 66.0 � 12.2 64.6 � 12.4 63.4 � 12.9
Men 59.0 52.3 60.7 65.6 56.9 63.3 63.8
Time on dialysis (y) 5.1 (2.4; 9.9) 5.2 (2.5; 10.1) 5.1 (2.4; 10.0) 4.9 (2.0; 9.8) 5.2 (2.4; 10.0) 5.1 (2.4; 9.7) 4.7 (2.0; 9.3)
Diabetes 32.5 32.5 31.4 32.7 33.4 32.3 33.6
History of CVD 19.5 19.4 17.8 19.1 22.6 19.6 19.5
Serum albumin (g/dL) 3.78 � 0.43 3.76 � 0.41 3.86 � 0.38 3.88 � 0.39 3.61 � 0.48 3.79 � 0.42 3.82 � 0.46
CRP (mg/dL) 0.12 (0.06; 0.40) 0.11 (0.05; 0.36) 0.10 (0.05; 0.30) 0.10 (0.05; 0.30) 0.22 (0.10; 0.80) 0.13 (0.06; 0.40) 0.10 (0.05; 0.36)
TSAT 26.8 � 13.6 26.6 � 13.7 28.4 � 13.1 31.1 � 15.1 23.9 � 14.0 25.6 � 12.4 30.1 � 15.8
Ferritin (ng/mL) 102 (39; 227) 100 (37; 223) 106 (43; 224) 123 (56; 277) 102 (34; 243) 93 (36; 201) 121 (50; 290)
Postdialysis body weight (kg) 52.5 � 10.8 51.7 � 10.6 53.1 � 10.8 52.9 � 10.9 51.6 � 10.7 53.2 � 10.8 52.5 � 11.0
Single-pool Kt/V 1.37 � 0.29 1.39 � 0.30 1.39 � 0.29 1.36 � 0.29 1.34 � 0.30 1.35 � 0.28 1.32 � 0.29
nPNA 0.91 � 0.19 0.91 � 0.18 0.93 � 0.18 0.92 � 0.19 0.89 � 0.20 0.92 � 0.19 0.91 � 0.20
Note: Count data are expressed as a percentage, and continuous variables, as mean � standard deviation) if normally distributed or median (25th; 75th percentile) if skewed. Categories ofESA responsiveness were defined by combination of ESA dosage (�6,000 and �6,000 U/wk) and Hb level (�10, 10-11.9, and �12 g/dL). Conversion factors for units: Hb in g/dL to g/L, �10;serum albumin in g/dL to g/L, �10; no conversion necessary for ferritin in ng/mL and �g/L.
Abbreviation: CRP, C-reactive protein; CVD, cardiovascular disease; ESA, erythropoiesis-stimulating agent; Hb, hemoglobin; nPNA, normalized protein nitrogen appearance; TSAT,transferrin saturation.
Am
JK
idneyD
is.2012;59(1):108-116111
ES
AR
esponsivenessand
Mortality
Fukuma et al
We also noted that the relationship between CVD-associated mortality and ESA responsiveness cat-egory was similar to that between all-cause mortalityand responsiveness: the highest rate of CVD-associ-ated mortality was noted in patients with low hemoglo-bin levels receiving high-dose ESA therapy (adjustedHR, 2.02; 95% CI, 1.88-2.17).
Table 2. Exploratory Logistic Regression Analysis toDetermine Predictors of ESA Hyporesponsiveness
Variable OR (95% CI)
Age (/10-y increase) 1.02 (1.00-1.03)
Man (vs woman) 0.93 (0.90-0.97)
Time on dialysis (/1-y increase) 1.00 (1.00-1.01)
History of CVD 1.04 (1.00-1.09)
Diabetes 1.02 (0.98-1.06)
Transferrin saturation�20% 1.90 (1.82-1.98)20%-49.9% 1.00 (reference)�50% 1.03 (0.94-1.14)
C-Reactive proteinFirst quartile 1.00 (reference)Second quartile 1.03 (0.98-1.09)Third quartile 1.20 (1.14-1.27)Fourth quartile 1.74 (1.65-1.83)
Serum albumin (/1-g/dL increase) 0.38 (0.37-0.40)
Postdialysis body weight (/10-kg increase) 1.00 (0.98-1.02)
Note: N � 95,460. ESA hyporesponsiveness was defined ashaving a hemoglobin level �10 g/dL while receiving an ESAdosage �6,000 U/wk.
Abbreviations: CI, confidence interval; CVD, cardiovasculardisease; ESA, erythropoiesis-stimulating agent; OR, odds ratio.
Table 3. Mortality Rates for All-Cause and Cardiovas
Hb (g/dL) ESA (U/wk) Person-years No.
�10 �3,000 4,922.8 277
3,000-5,999 12,537.7 894
6,000-8,999 6,380.5 678
�9,000 13,124.8 2174
10-11.9 �3,000 12,045.9 518
3,000-5,999 21,874.9 1063
6,000-8,999 7,792.0 463
�9,000 8,276.6 843
�12 �3,000 1,677.9 84
3,000-5,999 2,063.3 120
6,000-8,999 630.7 44
�9,000 622.3 47
Note: N � 95,460. Categories of ESA responsiveness we6,000-8,999, and �9,000 U/wk) and Hb level (�10, 10-11.9, ancodes were used to define cardiovascular mortality (codes I00 to
Abbreviations: ESA, erythropoiesis-stimulating agent; Hb, hemoglob
112
Dose-DependentAssociationBetweenESAandAll-CauseMortality
The association between 12 categories of ESAresponsiveness and all-cause mortality is shown inFig 2. We defined 4 categories of ESA dosage (�3,000,3,000-5,999, 6,000-8,999, or �9,000 U/wk) and 3categories of hemoglobin level (�10, 10-11.9, or �12g/dL). Patients with hemoglobin levels �10 g/dL andESA dosage �3,000 U/wk were our reference cat-egory. Results of analysis showed that the all-causemortality rate increased monotonically by ESA dos-age in patients with hemoglobin levels �10 g/dL.
SubgroupAnalysis
We analyzed patients in subgroups established ac-cording to transferrin saturation (�20% or �20%),ferritin level (�200 or �200 ng/mL), and age (�65 or�65 years). The association between all-cause mortal-ity and ESA responsiveness category is summarizedin Fig 3. The association between ESA hyporespon-siveness (hemoglobin �10 g/dL and ESA dosage�6,000 U/wk) and mortality was consistent in allsubgroups.
DISCUSSION
In this study, we found that ESA responsiveness,defined based on ESA dosage and hemoglobin level,was associated with 1-year mortality for all causesand CVD after adjustment for a range of potentialconfounding factors. In particular, the worst outcomeswere noted in patients with low hemoglobin levelsreceiving high-dose ESA therapy (hyporesponsive
Death at 1 Year by Category of ESA Responsiveness
ause Mortality Cardiovascular Mortality
Rate/1,000 person-years No.
Rate/1000 person-years
56.3 (50.0-63.1) 208 42.3 (36.8-48.2)
71.3 (66.9-75.9) 689 55.0 (51.0-59.1)
106.3 (98.8-114.1) 540 84.6 (77.9-91.7)
165.6 (159.3-172.1) 1,751 133.4 (127.6-139.3)
43.0 (39.4-46.8) 389 32.3 (29.2-35.6)
48.6 (45.8-51.5) 793 36.3 (33.8-38.8)
59.4 (54.3-64.9) 357 45.8 (41.3-50.7)
101.8 (95.4-108.6) 634 76.6 (71.0-82.5)
50.1 (40.1-61.6) 62 36.9 (28.4-47.1)
58.2 (48.5-69.2) 90 43.6 (35.2-53.4)
69.7 (51.1-92.5) 33 52.3 (36.3-72.7)
75.6 (56.0-99.2) 40 64.3 (46.3-86.5)
fined by combination of ESA dosage (�3,000, 3,000-5,999,2 g/dL). International Classification of Diseases, 10th RevisionThe 95% confidence intervals shown are for rates.
cular
All-C
re ded �1I99).
in.
Am J Kidney Dis. 2012;59(1):108-116
ESA Responsiveness and Mortality
patients). We also noted that the impact of hemoglo-bin level on mortality differed by ESA dose, indicat-ing an interaction between ESA dose and hemoglobinlevel. These findings suggest that ESA dose andhemoglobin level should not be considered separatelywhen evaluating these items’ effects on mortality, andthe category of ESA responsiveness has prognosticimportance in anemia management.
ESA responsiveness varies widely depending on anumber of clinical factors,19-21 and we identifiedpatients fitting a range of responsiveness categories inthe present study, even those receiving relatively lowESA dosages, because the difference in ESA dosagewas �9,000 U/wk in the same hemoglobin category.In an exploratory logistic regression model, older age,being female, lengthy time on dialysis therapy, historyof CVD, low serum albumin level, high CRP level,and low transferrin saturation in turn were found to beassociated with ESA hyporesponsiveness. Transferrinsaturation and serum albumin and CRP levels aresurrogate markers of iron deficiency, malnutrition,and inflammation, respectively. These factors may bemodified by iron supplementation or improving nutri-tion, and we also may be able to identify modifiableunderlying inflammatory processes22,23 by tracking
Table 4. Hazard Ratios for All-Cause and Cardiov
ESA (U/wk) Hb (g/dL)
All-Cause Mor
Unadjusted Model
�6,000 �10 1.44 (1.33-1.55)
10-11.9 1.00 (reference)
�12 1.17 (1.01-1.35)
�6,000 �10 3.13 (2.95-3.33)
10-11.9 1.74 (1.62-1.88)
�12 1.56 (1.26-1.92)
Note: N � 95,460. Values shown are hazard ratios (95% conficombination of ESA dosage (�6,000 and �6,000 U/wk) and Hb l
Abbreviations: ESA, erythropoiesis-stimulating agent; Hb, hema Adjusted for age, sex, time on dialysis therapy, postdialys
albumin level, and transferrin saturation.
Figure 2. Association between 12 cat-egories of erythropoiesis-stimulating agent(ESA) responsiveness and mortality for all-cause and cardiovascular disease. Catego-ries of ESA responsiveness were definedby combination of ESA dosage (�3,000,3,000-5,999, 6,000-8,999, and �9,000U/wk) and hemoglobin level (�10, 10-11.9, and �12 g/dL). Adjusted model in-cluded age, sex, time on dialysis therapy,postdialysis body weight, diabetes, historyof cardiovascular disease, serum albuminlevel, and transferrin saturation. Abbrevia-
tion: CI, confidence interval.Am J Kidney Dis. 2012;59(1):108-116
occult infections,24 clotted arteriovenous grafts,25 andundiagnosed malignancies.26,27
Given that the mentioned factors are associatedwith both ESA responsiveness and mortality,28-30 wefeared that they might confound the association be-tween ESA responsiveness and mortality. However,the association appeared to be present within allsubgroups for transferrin saturation ratio, ferritin level,and age, suggesting that the association is indepen-dent of these factors. In addition, given that thesefindings suggested that ESA responsiveness may bean intermediate factor between the mentioned factorsand mortality, we compared Cox regression models todetermine whether to include ESA responsiveness inmodels. We found that the association between thesefactors and mortality was not attenuated on adjust-ment for ESA responsiveness, suggesting that ESAresponsiveness was not only an intermediate factor,but also an independent predictive factor.
Few studies investigating the association of ESAdose and hemoglobin level with mortality have ac-counted for the interaction of these factors them-selves, and in the present study, we noted severalfindings indicative of interaction. First, the associa-tion between hemoglobin level and mortality was
lar Mortality by Category of ESA Responsiveness
Cardiovascular Mortality
sted Modela Unadjusted Model Adjusted Modela
(1.09-1.27) 1.47 (1.35-1.61) 1.19 (1.09-1.30)
(reference) 1.00 (reference) 1.00 (reference)
(0.97-1.31) 1.17 (0.98-1.38) 1.12 (0.95-1.33)
(1.82-2.07) 3.37 (3.14-3.61) 2.02 (1.88-2.17)
(1.34-1.55) 1.77 (1.63-1.93) 1.44 (1.32-1.57)
(1.06-1.63) 1.67 (1.32-2.12) 1.39 (1.09-1.76)
e intervals). Categories of ESA responsiveness were defined by�10, 10-11.9, and �12 g/dL).in.
dy weight, diabetes, history of cardiovascular disease, serum
ascu
tality
Adju
1.18
1.00
1.13
1.94
1.44
1.32
dencevel (oglobis bo
113
Fukuma et al
modified statistically by ESA dosage. Second, thecombined effect of having low hemoglobin level andreceiving high-dose ESA therapy was greater than thesum of baseline risk, having a low hemoglobin leveland receiving high-dose ESA therapy, a departurefrom the additive model.18,31,32
Although reports of the effects of ESA dose onmortality risk in observational studies are conflict-ing,33-36 indication-based confounding may contrib-ute to some of these conflicting findings. In addition,in the present study, the association between ESAdose and mortality may have been confounded byunmeasured factors, such as facility-level measures ofESA treatment patterns,37 because we used only dataavailable from the database. Although the mecha-nisms underlying these associations between ESA
dosage and mortality are unclear, our study suggests114
that the need to administer high-dose ESA therapy tothese patients is problematic.
The ESA dosage cutoff point in the present studywas 6,000 U/wk, in accordance with the medianvalue in this cohort. This cutoff point is lower thanthose in clinical guidelines in the United States andEurope.9,10 However, this cutoff point is consideredreasonable based on average practice in Japan. Wealso stratified patients by hemoglobin level (�10,10-11.9, and �12 g/dL) and examined the dose-dependent association between ESA dose and mor-tality. In patients with hemoglobin levels �10 g/dL,the mortality rate increased monotonically by ESAdosage, supporting the clinical importance of thelow cutoff point used in the present study. Inaddition, ESA responsiveness is important for pa-
Figure 3. Association between cat-egories of 12 erythropoiesis-stimulatingagent (ESA) responsiveness and all-cause mortality with subgroups accord-ing to (A) transferrin saturation ratio(�20% and �20%), (B) ferritin level(�200 and �200 ng/mL), and (C) age(�65 and �65 years). Categories of ESAresponsiveness were defined by combina-tion of ESA dosage (�3,000, 3,000-5,999, 6,000-8,999, and �9,000 U/wk)and hemoglobin level (�10, 10-11.9, and�12 g/dL). Adjusted model included age,sex, time on dialysis therapy, postdialysisbody weight, diabetes, history of cardio-vascular disease, serum albumin level,and transferrin saturation. Abbreviation:CI, confidence interval.
tients who receive a relatively low ESA dose.
Am J Kidney Dis. 2012;59(1):108-116
ESA Responsiveness and Mortality
Several studies have reported that ESA dose inJapan is far lower than that in other countries,38,39 andmost prior studies defined ESA hyporesponsivenessas a markedly higher ESA dose than we did here.1,40
Therefore, the high ESA dosage in this study wouldconstitute a relatively low dosage in other countries.This difference may have arisen due to Japaneseclinical guidelines and the reimbursement policy inJapan. The maximum ESA dosage for HD patients inJapan has been limited to 9,000 U/wk. The ironreplacement policy in Japan also is different from thatin other countries because Japanese clinical guide-lines recommend a relatively conservative iron replace-ment policy, indicating criteria for starting iron admin-istration as transferrin saturation �20% or ferritinlevel �100 ng/mL. However, data regarding intrave-nous iron doses were not available from the registry,and we therefore were unable to take these data intoaccount in the present study. We believe it is impor-tant to find the association between ESA responsive-ness and mortality in a Japanese population whoreceive ESAs at a dose far lower than that of othercountries. These findings suggest that ESA responsive-ness is a clinically important predictor even in patientswho receive a relatively low ESA dose, not only inJapan, but also in other countries.
Major strengths of this study were as follows. First,the participating cohort was large (N � 95,460),involving approximately one-third of all HD patientsin Japan. Second, we defined several categories ofESA responsiveness, allowing examination of indepen-dent and interactive effects of ESA dosage and hemo-globin level on mortality. In addition, these categoriesare easy to understand and useful for physicians in aclinical setting. Third, we conducted secondary analy-sis to examine the robustness of the association be-tween ESA responsiveness and mortality, and sub-group analyses support this robustness.
However, several limitations to the present studywarrant mention. First, baseline data were used todefine exposure categories in this cohort, and we wereunable to examine the effect of changes from thebaseline category during follow-up. Second, we lackedextensive precise information regarding comorbid con-ditions other than CVD and diabetes, and our cohorttherefore may have included patients with other malig-nancies and gastrointestinal bleeding. However, webelieve these cases are not frequent enough to signifi-cantly affect the results obtained and thus are inconse-quential. We also lacked information for route andfrequency of administration of ESAs, secondary hyper-parathyroidism, and certain drugs, such as angiotensin-converting enzyme inhibitors and angiotensin II receptorantagonists,41 which may affect ESA responsiveness.
Previous studies reported that most Japanese patientsAm J Kidney Dis. 2012;59(1):108-116
received ESAs by intravenous administration because ofclinical guidelines and directions on the ESA packageinsert. Finally, patients in this study may not be represen-tative of HD patients in other countries, particularlygiven that ESA dose and hemoglobin level generally arelower in Japan than in other countries. For these reasons,we advise precaution when interpreting and generalizingthese results.
Allowing for these method issues, we found thatESA responsiveness category, defined based on acombination of ESA dosage and hemoglobin level,was associated independently with 1-year mortality.In particular, the category of hyporesponsive, definedas having low hemoglobin levels and receiving high-dose ESA therapy, was associated with the highestmortality rate of all categories. We also noted aninteraction on mortality between hemoglobin leveland ESA dosage. Together, these findings suggest thatESA dosage and hemoglobin level may have indepen-dent and interactive effects on mortality and respon-siveness category has prognostic importance and clini-cal relevance in anemia management.
ACKNOWLEDGEMENTSWe thank participants in the dialysis registry of the Japanese
Society for Dialysis Therapy, members of the committee of renaldata registry, all personnel at participating institutions involved inthis survey, and Shin Yamazaki, PhD, for support in statisticalanalyses.
Support: None.Financial Disclosure: Dr Iseki has grants with Terumo and
Chugai. Dr Tsubakihara has received speaker fees from Chugaiand Kyowa Hakko Kirin. Dr Fukuhara is an advisor on epidemiol-ogy study for Kyowa Hakko Kirin and receives consulting feesfrom Kyowa Hakko Kirin.
REFERENCES1. Zhang Y, Thamer M, Stefanik K, Kaufman J, Cotter D.
Epoetin requirements predict mortality in hemodialysis patients.Am J Kidney Dis. 2004;44:866-876.
2. Kilpatrick R, Critchlow C, Fishbane S, et al. Greater epoetinalfa responsiveness is associated with improved survival in hemo-dialysis patients. Clin J Am Soc Nephrol. 2008;3:1077-1083.
3. López-Gómez J, Portolés J, Aljama P. Factors that conditionthe response to erythropoietin in patients on hemodialysis and theirrelation to mortality. Kidney Int Suppl. 2008;111:S75-S81.
4. Robinson B, Joffe M, Berns J, Pisoni R, Port F, Feldman H.Anemia and mortality in hemodialysis patients: accounting formorbidity and treatment variables updated over time. Kidney Int.2005;68:2323-2330.
5. Szczech L, Barnhart H, Inrig J, et al. Secondary analysis ofthe CHOIR trial epoetin-alpha dose and achieved hemoglobinoutcomes. Kidney Int. 2008;74:791-798.
6. Locatelli F, Pisoni R, Combe C, et al. Anaemia in haemodi-alysis patients of five European countries: association with morbid-ity and mortality in the Dialysis Outcomes and Practice PatternsStudy (DOPPS). Nephrol Dial Transplant. 2004;19:121-132.
7. Strippoli G, Navaneethan S, Craig J. Haemoglobin andhaematocrit targets for the anaemia of chronic kidney disease.
Cochrane Database Syst Rev. 2006;18(4):CD003967.115
Fukuma et al
8. Palmer S, Navaneethan S, Craig J, et al. Meta-analysis:erythropoiesis-stimulating agents in patients with chronic kidneydisease. Ann Intern Med. 2010;153:23-33.
9. National Kidney Foundation. KDOQI, II. Clinical PracticeGuidelines and Clinical Practice Recommendations for Anemia inChronic Kidney Disease in Adults. Am J Kidney Dis. 2006;47(5 suppl 3):S16-S85.
10. Locatelli F, Aljama P, Bárány P, et al. Revised Europeanbest practice guidelines for the management of anaemia in patientswith chronic renal failure. Nephrol Dial Transplant. 2004;19(suppl2):ii1-ii47.
11. Tsubakihara Y, Nishi S, Akiba T, et al. 2008 JapaneseSociety for Dialysis Therapy: guidelines for renal anemia inchronic kidney disease. Ther Apher Dial. 2010;14:240-275.
12. Kaysen G, Müller H, Ding J, Chertow G. Challenging thevalidity of the EPO index. Am J Kidney Dis. 2006;47:157-166.
13. McFarlane P, Pisoni R, Eichleay M, Wald R, Port F, Men-delssohn D. International trends in erythropoietin use and hemoglo-bin levels in hemodialysis patients. Kidney Int. 2010;78:215-223.
14. Verdalles U, Abad S, Vega A, et al. Factors related to theabsence of anemia in hemodialysis patients. Blood Purif. 2011;32:69-74.
15. Birgegard G. Managing anemia in lymphoma and multiplemyeloma. Ther Clin Risk Manag. 2008;4:527-539.
16. Iseki K, Nakai S, Yamagata K, Tsubakihara Y. Tachycardiaas a predictor of poor survival in chronic haemodialysis patients.Nephrol Dial Transplant. 2011;26:963-969.
17. Andersson T, Alfredsson L, Kallberg H, Zdravkovic S,Ahlbom A. Calculating measures of biological interaction. Eur JEpidemiol. 2005;20:575-579.
18. Li R, Chambless L. Test for additive interaction in propor-tional hazards models. Ann Epidemiol. 2007;17:227-236.
19. Coladonato J, Frankenfield D, Reddan D, et al. Trends inanemia management among US hemodialysis patients. J Am SocNephrol. 2002;13:1288-1295.
20. Regidor D, Kopple J, Kovesdy C, et al. Associations be-tween changes in hemoglobin and administered erythropoiesis-stimulating agent and survival in hemodialysis patients. J Am SocNephrol. 2006;17:1181-1191.
21. Kalantar-Zadeh K, Lee G, Miller J, et al. Predictors ofhyporesponsiveness to erythropoiesis-stimulating agents in hemo-dialysis patients. Am J Kidney Dis. 2009;53:823-834.
22. Stenvinkel P. Inflammation in end-stage renal failure: couldit be treated? Nephrol Dial Transplant. 2002;17(suppl 8):33-38;discussion 40.
23. Smrzova J, Balla J, Bárány P. Inflammation and resistanceto erythropoiesis-stimulating agents—what do we know and whatneeds to be clarified? Nephrol Dial Transplant. 2005;20(suppl8):viii2-viii7.
24. Kadiroglu A, Kadiroglu E, Sit D, Dag A, Yilmaz M.Periodontitis is an important and occult source of inflammation inhemodialysis patients. Blood Purif. 2006;24:400-404.
25. Nassar G, Fishbane S, Ayus J. Occult infection of old
nonfunctioning arteriovenous grafts: a novel cause of erythropoi-116
etin resistance and chronic inflammation in hemodialysis patients.Kidney Int Suppl. 2002;80:49-54.
26. Maisnar V, Chroust K. Treatment of associated anemia indifferent hematological disorders with epoetin alpha. Neoplasma.2004;51:379-384.
27. Abels R. Use of recombinant human erythropoietin in thetreatment of anemia in patients who have cancer. Semin Oncol.1992;19:29-35.
28. Kalantar-Zadeh K, Ikizler T, Block G, Avram M, Kopple J.Malnutrition-inflammation complex syndrome in dialysis patients:causes and consequences. Am J Kidney Dis. 2003;42:864-881.
29. Kalantar-Zadeh K, McAllister C, Lehn R, Lee G, NissensonA, Kopple J. Effect of malnutrition-inflammation complex syn-drome on EPO hyporesponsiveness in maintenance hemodialysispatients. Am J Kidney Dis. 2003;42:761-773.
30. Kalantar-Zadeh K, McAllister C, Lehn R, Liu E, Kopple J.A low serum iron level is a predictor of poor outcome in hemodi-alysis patients. Am J Kidney Dis. 2004;43:671-684.
31. Rothman KJ. Epidemiology: An Introduction. New York,NY: Oxford University Press; 2002.
32. Hosmer D, Lemeshow S. Confidence interval estimation ofinteraction. Epidemiology. 1992;3:452-456.
33. Bradbury B, Wang O, Critchlow C, et al. Exploring relativemortality and epoetin alfa dose among hemodialysis patients. Am JKidney Dis. 2008;51:62-70.
34. Zhang Y, Thamer M, Cotter D, Kaufman J, Hernán M.Estimated effect of epoetin dosage on survival among elderlyhemodialysis patients in the United States. Clin J Am Soc Nephrol.2009;4:638-644.
35. Bradbury B, Danese M, Gleeson M, Critchlow C. Effect ofepoetin alfa dose changes on hemoglobin and mortality in hemodi-alysis patients with hemoglobin levels persistently below 11 g/dL.Clin J Am Soc Nephrol. 2009;4:630-637.
36. Bradbury B, Do T, Winkelmayer W, Critchlow C, BrookhartM. Greater epoetin alfa (EPO) doses and short-term mortality riskamong hemodialysis patients with hemoglobin levels less than 11g/dL. Pharmacoepidemiol Drug Saf. 2009;18:932-940.
37. Brookhart MA, Schneeweiss S, Avorn J, Bradbury BD, LiuJ, Winkelmayer WC. Comparative mortality risk of anemia man-agement practices in incident hemodialysis patients. JAMA. 2010;303:857-864.
38. McFarlane PA, Pisoni RL, Eichleay MA, Wald R, Port FK,Mendelssohn D. International trends in erythropoietin use andhemoglobin levels in hemodialysis patients. Kidney Int. 2010;78:215-223.
39. Pisoni RL, Bragg-Gresham JL, Young EW, et al. Anemiamanagement and outcomes from 12 countries in the DialysisOutcomes and Practice Patterns Study (DOPPS). Am J Kidney Dis.2004;44:94-111.
40. Zhang Y, Thamer M, Stefanik K, Kaufman J, Cotter DJ.Epoetin requirements predict mortality in hemodialysis patients.Am J Kidney Dis. 2004;44:866-876.
41. Johnson D, Pollock C, Macdougall I. Erythropoiesis-stimulating agent hyporesponsiveness. Nephrology (Carlton). 2007;
12:321-330.Am J Kidney Dis. 2012;59(1):108-116
