- Email: [email protected]
Comparative Short-term Safety of Sodium Ferric Gluconate Versus Iron Sucrose in Hemodialysis Patients
Original Investigation Comparative Short-term Safety of Sodium Ferric Gluconate Versus Iron Sucrose in Hemodialysis Patients M. Alan Brookhart, PhD,1,2 Janet K. Freburger, PhD,2 Alan R. Ellis, PhD, MSW,2 Wolfgang C. Winkelmayer, MD, ScD,3 Lily Wang, PhD,2 and Abhijit V. Kshirsagar, MD, MPH 4 Background: Despite different pharmacologic properties, little is known about the comparative safety of sodium ferric gluconate versus iron sucrose in hemodialysis patients. Study Design: Retrospective cohort study using the clinical database of a large dialysis provider (2004-2005) merged with administrative data from the US Renal Data System. Setting & Participants: 66,207 patients with Medicare coverage who received center-based hemodialysis. Predictors: Iron formulation use assessed during repeated 1-month exposure periods (n 5 278,357). Outcomes: All-cause mortality, infection-related hospitalizations and mortality, and cardiovascular-related hospitalizations and mortality occurring during a 3-month follow-up period. Measurements: For all outcomes, we estimated 90-day risk differences between the formulations using propensity score weighting of Kaplan-Meier functions, which controlled for a wide range of demographic, clinical, and laboratory variables. Risk differences were also estimated within various clinically important subgroups. Results: Ferric gluconate was administered in 11.4%; iron sucrose, in 48.9%; and no iron in 39.7% of the periods. Risks for most study outcomes did not differ between ferric gluconate and iron sucrose; however, among patients with a hemodialysis catheter, use of ferric gluconate was associated with a slightly decreased risk for both infection-related death (risk difference, 20.3%; 95% CI, 20.5% to 0.0%) and infection-related hospitalization (risk difference, 21.5%; 95% CI, 22.3% to 20.6%). Bolus dosing was associated with an increase in infection-related events among both ferric gluconate and iron sucrose users. Limitations: Residual confounding and outcome measurement error. Conclusions: Overall, the 2 iron formulations studied exhibited similar safety profiles; however, ferric gluconate was associated with a slightly decreased risk for infection-related outcomes compared to iron sucrose among patients with a hemodialysis catheter. These associations should be explored further using other data or study designs. Am J Kidney Dis. 67(1):119-127. ª 2016 by the National Kidney Foundation, Inc. Published by Elsevier Inc. All rights reserved. INDEX WORDS: Anemia; chronic kidney disease (CKD); end-stage renal disease (ESRD); hemodialysis (HD); intravenous iron formulations; sodium ferric gluconate; iron sucrose; mortality; infection; hospitalization; cardiovascular events; safety.
A
nemia is common among patients with end-stage renal disease (ESRD)1 and is associated with increased morbidity, mortality, and risk for hospitalization.2 The anemia seen in these patients is primarily caused by impaired production of endogenous renal erythropoietin. It is worsened by depletion of iron reserves caused by hemodialysis-related blood loss and poor intestinal absorption of iron.3 Erythropoiesisstimulating agents are the primary treatment for the anemia of ESRD. The iron deficiency is addressed
through administration of intravenous (IV) iron formulations and, much less effectively, oral iron supplementation.4 Currently, there are 6 formulations of IV iron available on the US market: low-molecular-weight iron dextran, high-molecular-weight iron dextran, iron sucrose, sodium ferric gluconate, ferumoxytol, and ferric carboxymaltose. High-molecular-weight iron dextran has been linked to hypersensitivity reactions, including anaphylaxis,5,6 and carries a black box advisory on its
From the 1Department of Epidemiology, UNC Gillings School of Global Public Health UNC Chapel Hill; 2Cecil G. Sheps Center for Health Services Research, University of North Carolina, Chapel Hill, NC; 3Selzman Institute for Kidney Health, Section of Nephrology, Department of Medicine, Baylor College of Medicine, Houston, TX; and 4University of North Carolina Kidney Center, UNC School of Medicine, Chapel Hill, NC. Received February 2, 2014. Accepted in revised form July 15, 2015. Originally published online September 15, 2015. Because an author of this article is an editor for AJKD, the peerreview and decision-making processes were handled entirely by an
Associate Editor (Michel Jadoul, MD) who served as Acting Editor-in-Chief. Details of the journal’s procedures for potential editor conflicts are given in the Information for Authors & Journal Policies. Address correspondence to M. Alan Brookhart, PhD, Department of Epidemiology, 2105F McGavran-Greenberg, CB#7435, Chapel Hill, NC 27599. E-mail: [email protected] 2016 by the National Kidney Foundation, Inc. Published by Elsevier Inc. All rights reserved. 0272-6386 http://dx.doi.org/10.1053/j.ajkd.2015.07.026
Am J Kidney Dis. 2016;67(1):119-127
119
Brookhart et al
label. Consequently, there is currently little use of iron dextran in the US dialysis population. Ferumoxytol was approved by the US Food and Drug Administration in 2009 and is not widely used; ferric carboxymaltose was approved in 2013, but remains labeled for use in non2dialysis-dependent chronic kidney disease. Thus, the iron formulations most commonly used in the US ESRD population are sodium ferric gluconate and iron sucrose. Sodium ferric gluconate and iron sucrose preparations are iron-carbohydrate complexes or colloids of varying molecular size with a spheroid ironcarbohydrate core. Because of differences in core size, carbohydrate chemistry, and molecular weight, the 2 formulations have different pharmacokinetic and pharmacodynamic properties.7 Despite concerns about oxidative stress and infection risk associated with the use of IV iron,8,9 pharmacologic differences between the different iron formulations,7 and apparent differences in effectiveness,10 there is currently little information about the comparative safety of ferric gluconate versus iron sucrose.11 To address this evidence gap, we conducted a large-scale comparative safety study of ferric gluconate versus iron sucrose in a cohort of patients undergoing maintenance hemodialysis.
antibiotic use. The USRDS is a national data system funded by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) that collects, analyzes, and distributes information about the treatment of ESRD. USRDS includes data from the Medical Evidence Report Form, the Medicare Enrollment database, the ESRD Death Notification Form, and the standard analytic files, which contain final action claims data.12 These data were used to obtain information about demographic characteristics, health care use (eg, hospitalizations and outpatient care), and additional clinical characteristics (eg, comorbid conditions). We examined clinical data from 2 years (2004-2005) during which the dialysis provider used both ferric gluconate and iron sucrose. After 2005, the dialysis provider used mostly iron sucrose.
METHODS
We first identified center-based outpatient hemodialysis patients defined as follows: individuals who had undergone at least 9 months of maintenance dialysis (which accounted for the 6-month baseline period and an additional 3 months from the initiation of dialysis therapy to allow for stability in the Centers for Medicare & Medicaid Services claims processing),12 received hemodialysis in a dialysis facility, were covered by Medicare Parts A and B, and had at least one TSAT measurement during January 30, 2004, to November 30, 2005 (the November 30 date was chosen to allow for the 1-month exposure period and at least one day of followup). Patients were excluded if they had polycystic kidney disease because anemia management may be considerably different in these patients. TSAT records were excluded if: (1) iron dextran or both ferric gluconate and iron sucrose were delivered in the exposure period; (2) the duration of Part A claims prior to index was insufficient (ie, ,120 days of Part A claims), suggesting
Data Sources The data used for this study came from the clinical research database of a large dialysis provider and the US Renal Data System (USRDS), which were merged at the patient level. The dialysis provider owns and manages more than 1,500 outpatient dialysis facilities located throughout the United States in urban, rural, and suburban areas. Their clinical database captures detailed clinical, laboratory, and treatment data for patients receiving care at all their dialysis units. All data are collected using standardized electronic health record systems. For this study, we used the clinical data to obtain detailed information about iron formulation use and dosing, erythropoiesis-stimulating agent use and dosing, clinical laboratory values (eg, hemoglobin, transferrin saturation [TSAT], and serum ferritin), current vascular access, and recent IV
Study Design We used a retrospective cohort design with repeated measures on iron treatments and outcomes. The index date of the 1-month exposure assessment period was anchored on a TSAT measurement because this information is used to guide iron administration. The 6-month baseline period prior to the TSAT measurement was used to identify potential confounders and effect modifiers. Patients were followed up for 3 months after the exposure period. Eligible patients could contribute multiple exposure/follow-up periods (Fig 1). Our design was implicitly attempting to mimic a randomized trial in which patients prescribed a 1-month course of iron were randomly assigned to receive the iron as either ferric gluconate or iron sucrose and then followed up for 90 days, with further iron treatment decisions left to the discretion of the physician.
Cohort Identification
second observation first observation
baseline
exposure
TSAT lab
follow-up
baseline
exposure
follow-up
TSAT lab
Figure 1. Study design schematic. Abbreviations: lab, laboratory; TSAT, transferrin saturation. 120
Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients incomplete data; (3) the TSAT measurement occurred during the follow-up period for a prior eligible TSAT; or (4) there were fewer than 9 dialysis sessions in the month preceding the index date or during the exposure period (suggesting that the individual was not receiving regular in-center hemodialysis). We also excluded records with missing covariate information.
Study Variables Outcomes We examined a variety of adverse clinical outcomes related to death, infection, and cardiovascular events. These were determined by examining the Medicare death notification and inpatient and outpatient claims data. The specific definition and data source for each outcome are presented in Table S1 (provided as online supplementary material). We examined 6 outcomes: death from any cause, infection-related hospitalization, infection-related death, hospitalization for myocardial infarction, hospitalization for stroke, and cardiovascular disease–related death. In addition, we created 2 composite outcomes: cardiovascular disease–related hospitalization or death and infection-related hospitalization or death.
Exposures Iron formulations were classified as ferric gluconate or iron sucrose based on native codes in the clinical database. The primary exposure contrast of interest was ferric gluconate versus iron sucrose. We also created a no-iron category for individuals who received no iron in the 1-month exposure period. We describe characteristics of the no-iron patients, but they were not otherwise used in the analysis.
Covariates We included a wide range of covariates in our analyses. Descriptions of the covariates and the manner in which they were defined are presented in Table S2. Covariates included demographic characteristics (eg, age, sex, race, Medicaid eligibility, census region, and year), clinical characteristics (eg, cause of ESRD, dialysis vintage, body mass index, type of vascular access, and number of hospital days), laboratory and anemia management variables (baseline values for hemoglobin, ferritin, TSAT, iron dose, epoetin alfa dose, and albumin; receipt of a blood transfusion; and epoetin alfa dose during exposure period), and several comorbid condition measures based on the Elixhauser classification13 and content expertise of the investigative team. Because of the potential relationship between iron use and infections, we created 4 “history of infection” variables: history of pneumonia, sepsis, or vascular access infection during the baseline period and history of any infection or use of IV antibiotics in the last month. Due to the extensive list of comorbid conditions, we selected certain ones to include in a parsimonious model and others to add during sensitivity analyses.
Statistical Analysis To describe the treatment groups, we calculated mean values and frequencies of all covariates within each treatment group. To assess the relationship between formulation choice and outcomes, we estimated unadjusted and adjusted cumulative incidence functions using Kaplan-Meier methods. Adjustment was obtained by applying inverse probability of treatment weighting to the Kaplan-Meier estimator. Weights were estimated using a generalized additive logistic model of the treatment choice, in which all covariates in Table S1 were included as potential confounders. Age, serum albumin level, body mass index, TSAT, and hemoglobin level were entered into the propensity score model using smoothing splines.14 Patients were censored after 90 days, at loss to follow-up, or administratively on December 31, 2005. Hospitalization outcomes were further censored at loss of Medicare Parts A and B eligibility. We graphed the survival function and the Am J Kidney Dis. 2016;67(1):119-127
difference between survival functions (risk difference [RD]) for each day during the 90-day follow-up period. Confidence intervals (CIs) were estimated for the daily RDs between groups using a nonparametric clustered bootstrap based on 250 resamples. Because previous research has reported increased infection risk for bolus relative to maintenance dosing,15 we assessed the contrast between these dosing strategies among patients receiving ferric gluconate and iron sucrose separately. Bolus and maintenance dosing were defined as in the previous report.15 Separate propensity scores were estimated for this contrast. All statistical analyses were conducted using R Statistical Software, version 3.3.16
Subgroup Analysis We examined several demographic and clinical subgroups (Table S3). Individuals were categorized based on race (black and nonblack), sex, hemodialysis catheter use, history of infection in the last month of baseline, TSAT at baseline (,20%, 20%-,50%, and $50%), time since reported incidence of ESRD (,1, 1-,4, and $4 years), hemoglobin level at baseline (,10, 10-12, and .12 g/dL), and ferritin level at baseline (,200, 200-500, and .500 mg/L).
Sensitivity Analyses Because our definition of hospitalization for infection was specific to sepsis, vascular access infection, and pneumonia, in a sensitivity analysis, we considered a broader infection definition, hospitalization for infection of any major organ system, and also receipt of IV antibiotics (Table S1). We also assessed the sensitivity of our results to the specification of the propensity score model by examining how the addition of other potentially relevant covariates to our primary model affected estimates. The additional variables that we considered were cause of ESRD, calendar year, region, Medicaid eligibility, and a number of additional comorbid conditions (Table S2). In this analysis, we increased the degrees of freedom (flexibility) of the regression splines. We also conducted a sensitivity analysis in which we stratified on iron use (vs no use) in the past month. This analysis was conducted to examine whether there may be delayed effects of formulation choice on outcomes. Finally, we conducted 3 sensitivity analyses with shorter iron exposure and/or follow-up periods: (1) a 2-week exposure and 6week follow-up period, (2) a 1-week exposure and 6-week followup period, and (3) a 1-month exposure and 6-week follow-up period. An institutional review board of the University of North Carolina approved this study. The study design, analytic methods, outcomes, and covariate definitions were prespecified in a protocol submitted to the Agency for Healthcare Research and Quality, which funded this study. Some aspects of the study were modified in response to comments from journal reviewers.
RESULTS Overall, 66,207 patients met study entry requirements and contributed data for 278,357 ironexposure/follow-up periods (Fig S1). Individuals who received ferric gluconate were more likely to be black and to reside in the South (Table 1). The groups were fairly similar with respect to many laboratory values and comorbid conditions. However, ferric gluconate users tended to have a slightly higher prevalence of some comorbid conditions and were more likely to have had any infection in the past month (12.2% vs 11.5%) and a vascular access infection in the past 6 months (14.7% vs 12.2%). Ferric gluconate users were less likely to have diabetes (49.3% vs 52.7%), but more likely to have a hemodialysis catheter (31.2% vs 121
Brookhart et al Table 1. Characteristics of Exposure Periods by Exposure Group Ferric Gluconate (n 5 31,726; 11.4%)
Iron Sucrose (n 5 136,233; 48.9%)
No Iron Use (n 5 110,398; 39.7%)
Mean age, y Female sex Race White Black
45.9 45.0
46.0 49.5
46.9 46.4
45.0 51.3
49.5 44.0
46.4 46.3
Medicaid Region Midwest Northeast South West
51.9
51.3
52.7
13.8 15.4 58.5 11.8
16.5 11.5 47.4 24.1
16.2 12.6 49.2 21.8
ESRD cause Diabetes Glomerulonephritis Hypertension
43.0 13.1 31.7
45.1 12.5 31.1
42.4 13.5 31.8
4.1 6 4.2 27.0 6 7.1 31.2 3.8 6 0.4 12.1 6 1.2 27.4 6 11.2 536 6 356 240 6 256 291 6 278 102 6 107 97 6 106 6.4 0.6 6 1.9 12.2
4.0 6 4.2 27.0 6 6.8 23.9 3.9 6 0.4 12.3 6 1.3 28.3 6 11.2 493 6 305 245 6 275 294 6 288 93 6 94 90 6 94 6.0 0.6 6 1.9 11.5
4.7 6 4.5 26.4 6 6.5 23.4 3.9 6 0.4 12.2 6 1.3 34.7 6 15.6 764 6 541 90 6 230 060 74 6 85 78 6 89 6.2 0.6 6 1.8 10.4
10.1 9.4 14.7 49.3 10.1 3.6 18.2 8.6 5.0
9.7 10.5 12.2 52.7 9.9 3.2 17.0 8.5 4.9
9.4 9.2 10.3 49.0 9.5 3.1 15.0 8.6 4.0
Characteristic
Dialysis vintage, y BMI, kg/m2 Hemodialysis catheter use Baseline albumin, g/dL Baseline hemoglobin, g/dL Index TSAT, % Baseline ferritin, mg/L Iron dose during last month of baseline, mg Iron dose during exposure period, mg EPO dose at baseline, 1,000 U/mo EPO dose during exposure period, 1,000 U/mo Blood transfusion during baseline period No. of hospital days during past montha Infection during past montha Comorbid conditions during 6-mo baseline period Pneumonia Sepsis Vascular access infection Diabetes Ischemic stroke Myocardial infarction COPD, asthma Cancer GI bleeding
Note: N 5 278,357. Unless otherwise indicated, values for categorical variables are given as percentage; for continuous variables, as mean 6 standard deviation. Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; EPO, epoetin alfa; ESRD, end-stage renal disease; GI, gastrointestinal; TSAT, transferrin saturation. a The month before the start of the exposure period.
23.9%). Mean iron dose during the exposure period did not vary by formulation type and was 291 mg for ferric gluconate and 294 mg for iron sucrose. For the adjusted analysis, we computed propensity scores for the overall population receiving iron. In Figure S2, we present the estimated densities of the propensity scores in each treatment group. There was a high degree of overlap between propensity score distributions across all subgroups, suggesting good comparability between treatment groups.17 Counts of events, total person-years of follow-up, and unadjusted and adjusted 90-day RDs for all outcomes are presented 122
in Table 2. Event rates were similar across treatment groups; the only notable difference was observed with infection-related death, for which ferric gluconate was associated with a decreased rate (relative rate, 0.78). In the adjusted estimate of the absolute 90-day RD, we observed slightly lower risk associated with ferric gluconate across all outcomes, but RDs were small in magnitude, ranging from 20.1% to 20.6%, corresponding to 1 to 6 fewer events over the 90-day followup period per 1,000 patients treated. We present the inverse probability of treatment-weighted survival curves for all outcomes in Fig S3. Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients Table 2. The 90-Day Outcome Rates, Rate Ratios, and Risk Differences, Ferric Gluconate Versus Iron Sucrose
Outcome
All-cause death
Treatment Group
IS FG Infection-related IS hospitalization FG Infection-related IS mortality FG Composite infection IS outcome FG Hospitalized stroke IS FG Acute MI IS FG CV death IS FG Composite CV IS outcome FG
Person Time, y
Events
Event Rate (per 100 person-y)
32,749 7,627 31,778 7,394 32,749 7,627 31,778 7,394 32,523 7,582 32,584 7,590 32,749 7,627 32,414 7,554
6,102 1,468 8,510 2,062 698 126 8,769 2,106 1,692 355 1,181 281 2,473 596 4,933 1,153
18.6 19.2 26.8 27.9 1.7 2.1 27.6 28.5 5.2 4.7 3.6 3.7 7.6 7.8 15.2 15.3
Unadjusted Relative Rate (95% CI)
90-d Unadjusted RD (95% CI)a
90-d Adjusted RD (95% CI)a
—
—
1.03 (0.98 to 1.09)
0.1 (20.1 to 0.4)
20.3 (21.0 to 0.3)
1.04 (0.99 to 1.10)
0.3 (0.0 to 0.6)
20.6 (21.6 to 0.2)
— —
0.78 (0.64 to 0.94)
—
1.03 (0.98 to 1.09)
—
0.90 (0.80 to 1.01)
—
— —
— — —
20.1 (20.2 to 0.0) 20.1 (20.3 to 0.0)
—
0.2 (20.1 to 0.5)
—
—
20.6 (21.7 to 0.1)
—
20.1 (20.3 to 0.0) 20.2 (20.4 to 0.0)
—
—
1.02 (0.89 to 1.17)
0.0 (20.1 to 0.2)
20.1 (20.3 to 0.1)
1.03 (0.95 to 1.13)
0.1 (20.1 to 0.2)
20.1 (20.4 to 0.1)
1.00 (0.94 to 1.07)
0.0 (20.2 to 0.3)
20.3 (21.0 to 0.1)
—
—
—
Abbreviations: CI, confidence interval; CV, cardiovascular; FG, ferric gluconate; IS, iron sucrose; MI, myocardial infarction; RD, risk difference. a RDs estimated using inverse probability treatment-weighted Kaplan-Meier functions with bootstrap CIs.
The adjusted 90-day RDs and CIs for all-cause mortality and the composite infection outcome for all subgroups are presented in forest plots in Figs 2 and 3. Forest plots for the remaining outcomes are presented in Figs S4 to S9. We observed a slightly decreased risk for both infection-related mortality (RD, 20.3%; 95% CI, 20.5% to 0.0%) and infection-related hospitalization (RD, 21.5%; 95% CI, 22.3% to 20.6%) among patients with a hemodialysis catheter who were treated with ferric gluconate. There were several additional subgroups in which ferric gluconate was associated with decreased risk for infection-related mortality. Results for the ferric gluconate versus iron sucrose comparisons using alternative infection definitions (Fig S10) indicated that ferric gluconate was associated with decreased use of IV antibiotics overall (RD, 22.5%; 95% CI, 24.2% to 21.2%) and among patients with a hemodialysis catheter (RD, 24.5%; 95% CI, 25.7% to 23.4%). The composite infection results within the group of patients with a hemodialysis catheter were robust to the addition of covariates beyond our a priori–specified multivariable models (Fig S11). In analyses stratified on past iron use and catheter status, we observed similar associations between formulation choice and risk for the composite infection outcome (Fig S12). In Fig 4, we present cumulative risk curves of infection-related hospitalization or mortality associated with bolus versus maintenance dosing stratified by formulation type and catheter use. Graphs of the differences between cumulative risk curves and 95% CIs of the differences are provided in Fig S13. These analyses showed that bolus dosing was associated with Am J Kidney Dis. 2016;67(1):119-127
increased 90-day risk among both ferric gluconate and iron sucrose users, with the risk particularly pronounced among patients with hemodialysis catheters. In Figs S14 and S15, we present contrasts in risk for the composite infection outcome between maintenance iron and no iron and bolus iron and no iron, overall and by catheter use, respectively.
DISCUSSION We conducted a large-scale comparative study of the safety of different IV iron formulations administered in usual-care settings to typical patients undergoing maintenance hemodialysis in the United States. Although we examined a large number of outcomes and subgroups, we found only a few differences between the 2 formulations. Their use was associated with nearly identical risks for myocardial infarction, cardiovascular mortality, and all-cause mortality. Risks for these events were statistically similar within the overall population, as well as within all patient subgroups considered. The lack of an association between formulation choice and short-term risk for cardiovascular events is consistent with a recent study that found no association between any iron exposure and short-term risk for cardiovascular events.18 We found evidence suggesting a modestly increased risk for infection outcomes among patients exposed to iron sucrose versus ferric gluconate. In the overall population, the differences in infection risk between formulations were statistically nonsignificant or small in magnitude. However, among patients with a hemodialysis catheter, a subgroup of patients known to be particularly prone to infection, we observed differences 123
Brookhart et al Iron Ferric % Risk Sucrose Gluconate Difference Subgroup Cohort (N) Cohort (N) (95% CI) Overall 136233 31726 −0.3 (−1.0, 0.3) Female 62607 14556 −0.1 (−0.5, 0.3) Male 73626 17170 −0.4 (−1.4, 0.6) Black 59909 16282 −0.5 (−1.3, 0.2) Non Black 76324 15444 0.1 (−0.3, 0.6) Hemodialysis Catheter 32576 9907 −0.4 (−0.9, 0.2) Fistula or Graft 103657 21819 −0.2 (−1.0, 0.5) Recent Infection 15673 3859 −0.2 (−1.2, 0.8) Hemoglobin <= 10g/dL 6432 1585 −0.5 (−2.1, 1.1) Hemoglobin in 10−12g/dL 51569 13040 −0.2 (−0.6, 0.3) Hemoglobin >12g/dL 78232 17101 −0.3 (−1.1, 0.5) TSat <= 20% 32957 8887 −0.3 (−0.9, 0.3) TSat in 20−50% 98540 21759 −0.3 (−1.1, 0.4) TSat >50% 4736 1080 0.4 (−1.2, 1.7) Ferritin <= 200ng/mL 17874 2420 −0.7 (−1.5, 0.2) Ferritin in 200−500ng/mL 115074 28600 −0.3 (−1.1, 0.4) Ferritin >500ng/mL 60790 16366 0.0 (−0.3, 0.3) Vintage <= 1 year 39943 9260 0.2 (−0.4, 0.6) Vintage in 1−3 years 40437 9347 0.1 (−0.4, 0.6) Vintage > 3 years 55853 13119 −0.8 (−1.8, 0.3) No Iron Use in Previous Month 22173 5732 0.1 (−0.6, 0.7) −2
−1.5
−1
−0.5
0
0.5
1
1.5
Figure 2. The 90-day risk for all-cause mortality, ferric gluconate versus iron sucrose, by subgroup. Risk differences were estimated using inverse probability of treatment-weighted Kaplan-Meier functions with bootstrap confidence intervals (CIs). Abbreviation: TSat, transferrin saturation.
in both infection-related hospitalization and infectionrelated mortality, with higher rates in patients exposed to iron sucrose. In a sensitivity analysis, we also observed greater use of antibiotics among patients treated with iron sucrose. The differences between the
formulations were found despite the higher prevalence of catheters in the ferric gluconate group. Concerns over iron and infection in the chronic kidney disease population were raised many years ago as more patients began receiving iron as adjuvant
% Risk Ferric Iron Difference Sucrose Gluconate (95% CI) Cohort (N) Cohort (N) Subgroup 31726 −0.6 (−1.7, 0.1) 136233 Overall 14556 −0.2 (−0.6, 0.4) 62607 Female 17170 −0.9 (−2.2, 0.3) 73626 Male Black 59909 16282 −1.3 (−2.9, 0.2) 15444 0.0 (−0.5, 0.4) 76324 Non Black 9907 −1.5 (−2.4, −0.6) Hemodialysis Catheter 32576 103657 21819 −0.2 (−1.1, 0.6) Fistula or Graft 3859 −0.9 (−2.2, 0.6) 15673 Recent Infection 1.0 (−0.9, 2.8) 1585 Hemoglobin <= 10g/dL 6432 13040 −0.2 (−0.7, 0.3) 51569 Hemoglobin in 10−12g/dL 17101 −1.0 (−2.2, 0.2) Hemoglobin >12g/dL 78232 32957 8887 −0.1 (−0.9, 0.6) TSat <= 20% 21759 −0.8 (−1.9, 0.2) 98540 TSat in 20−50% 0.3 (−1.2, 2.0) 1080 TSat >50% 4736 17874 2420 −1.4 (−2.6, −0.3) Ferritin <= 200ng/mL 28600 −0.5 (−1.6, 0.4) 115074 Ferritin in 200−500ng/mL 16366 −0.2 (−0.7, 0.2) 60790 Ferritin >500ng/mL 0.0 (−0.7, 0.5) 9260 39943 Vintage <= 1 year 9347 −0.2 (−0.9, 0.4) Vintage in 1−3 years 40437 55853 13119 −1.3 (−3.0, 0.3) Vintage > 3 years 5732 −0.4 (−1.2, 0.4) 22173 No Iron Use in Previous Month −2
−1
0
1
2
Figure 3. The 90-day risk for the composite infection outcome, ferric gluconate versus iron sucrose, by subgroup. Risk differences were estimated using inverse probability of treatment-weighted Kaplan-Meier functions with bootstrap confidence intervals (CIs). Abbreviation: TSat, transferrin saturation. 124
Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients
Ferric Gluconate Subgroup
Iron Sucrose Subgroup
0.15
0.10
Cumulative Risk
0.05
Treatment Group
0.00 Ferric Gluconate, Catheter Subgroup
Iron Sucrose, Catheter Subgroup
Maintenance Bolus Dosing
0.15
0.10
0.05
0.00 30
60
90
30
60
90
Days Figure 4. Bolus versus maintenance dosing and cumulative risk for the composite infection outcome, stratified by formulation type and catheter use. Cumulative risk was estimated using inverse probability of treatment-weighted Kaplan-Meier functions.
treatment to ESAs.19-23 The part that iron plays in bacterial growth and infection risk has biological plausibility and has been shown in animal models,20,24-26 but there are relatively few studies of iron and infections in large chronic kidney disease populations. One study found that frequent iron administration is associated with increased risk for infection-related mortality in ESRD.27 Another recent report found that bolus (ie, repletion) dosing of iron is associated with increased risk for infection-related hospitalizations compared with maintenance dosing.15 Finally, a small clinical study determined that cumulative iron exposure is associated with increased risk for bacteremia.28 Our study raises the possibility that the infection risk from IV iron administration may differ by the IV iron formulation agent used. The putative differences in infection risk between ferric gluconate and iron sucrose could be explained by pharmacokinetic differences between the 2 iron formulations. Ferric gluconate is taken up quickly by the reticuloendothelial system and thus is eliminated rapidly from plasma.7 In contrast, iron sucrose is taken up more slowly and thus has a longer plasma half-life.7 These differences in pharmacology suggest that iron Am J Kidney Dis. 2016;67(1):119-127
sucrose may result in more iron in circulation, both free and bound to transferrin. Two clinical studies have reported abundant non–transferrin-bound iron after iron sucrose administration.29,30 Increasing the amount of iron in circulation may result in differences in both safety and effectiveness. A recent report using the same data found that iron sucrose was slightly more effective than ferric gluconate at increasing hemoglobin levels and iron indexes.10 Also, a small clinical study found that iron sucrose administration resulted in transferrin oversaturation in some patients and the growth of Staphylococcus epidermidis in vitro that was proportional to the amount of free iron in the sample.31 Despite some biological plausibility, the associations that we have observed between iron sucrose and infection risk must be interpreted cautiously. Given that our study was nonexperimental, it is susceptible to bias resulting from residual confounding. Although we have included a wide range of variables in our propensity score model, there may be some uncontrolled differences between patient groups. Our results may also have been confounded by uncontrolled aspects of medical practice. For example, facilities that 125
Brookhart et al
tended to use more ferric gluconate may also have used other practices that could alter infection risk. Our study may have been vulnerable to this bias because some patients in our study came from a smaller dialysis organization that had recently merged with the larger chain. We observed some difference in formulation preference between the groups of units, which may have been correlated with coding practices or unobserved aspects of practice or patient case-mix. Because our outcomes and covariate definitions are based on International Classification of Diseases, Ninth Revision codes and cause-of-death reporting, they may have been misclassified to some degree, possibly in a differential manner. Finally, given the large number of associations examined, some falsepositive associations would be expected. We did not attempt to shrink subgroup effects in any way to account for this issue. Our study had some limitations beyond those related to internal validity. First, our study focused on short-term effects. Some of the hypothesized effects of iron on cardiovascular risk resulting from free radical damage to the vasculature would be a function of longer term cumulative exposure.8 Second, our study design necessitated survival until 9 months after dialysis therapy initiation; thus, our results may not generalize to patients new to dialysis therapy. The limitations of our study are counterbalanced by 2 key strengths. First, by focusing on shorter term associations of a well-defined exposure with covariates defined prior to exposure and outcomes ascertained after the exposure assessment period ended, we removed or diminished many sources of bias typical in nonexperimental studies of longitudinal exposures (eg, immortal person-time bias, selection bias, and timevarying confounding). Second, we used detailed data, which let us control for essential clinical and laboratory variables and also capture important events outside the clinic (eg, hospitalization and death). In summary, we found that the 2 most commonly used iron formulations were similar with respect to most clinical outcomes. However, we found increased risk for infection associated with iron sucrose versus ferric gluconate among patients with a hemodialysis catheter. Given the nonexperimental nature of our study and the many outcomes and subgroup effects examined, this finding should motivate additional studies using different data or employing different research designs.
ACKNOWLEDGEMENTS Some of the data reported here have been supplied by the USRDS. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the US government or the USRDS. The authors thank DaVita Clinical Research, which had 126
no role in the design or implementation of this study or on the decision to publish, for providing data for this study. Support: This project was funded under contract no. HHSA290200500401 Task Order #5 from the Agency for Healthcare Research and Quality, US Department of Health and Human Services. Dr Winkelmayer was supported by grant R01 DK090181 from the National Institutes of Health (NIH), NIDDK. He also receives research and salary support through the endowed Gordon A. Cain Chair in Nephrology at Baylor College of Medicine. Dr Brookhart receives investigator-initiated research funding from the NIH (R01 AG042845, R21 HD080214, and R01 AG023178) and through contracts with the Patient Centered Outcomes Research Institute. The study sponsors did not have any role in study design; collection, analysis, and interpretation of data; writing the report; or the decision to submit the report for publication. Financial Disclosure: Dr Brookhart has received investigatorinitiated grant support from Amgen and served as a scientific advisor for Pfizer, Merck, Amgen, and Rockwell Medical, but has not accepted personal compensation for this service (honoraria received by institution or donated). Drs Freburger, Kshirsagar, and Ellis have received investigator-initiated grant support from Amgen. Dr Kshirsagar served on a Fresenius Advisory Board and Dr Ellis has received investigator-initiated grant support from Merck and the UNC Center for Pharmacoepidemiology. Dr Winkelmayer has served as an advisor or consultant to Amgen, Astra-Zeneca, Bayer, Keryx, Medgenics, Medtronic, MitsubishiTanabe, and Rockwell Pharma. Contributions: Research idea and study design: MAB, JKF, AVK, WCW; data acquisition: MAB; data analysis/interpretation: MAB, ARE, LW, JKF, WCW; statistical analysis: MAB, ARE, LW; supervision or mentorship: MAB, AVK, WCW. Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved. MAB takes responsibility that this study has been reported honestly, accurately, and transparently; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
SUPPLEMENTARY MATERIAL Table S1: Study outcomes. Table S2: Definition of covariates. Table S3: Definition of subgroups. Figure S1: Flowchart of patients through study. Figure S2: Propensity score distributions, overall and by treatment group. Figure S3: Inverse-probability of treatment-weighted survival curves. Figure S4: 90-d risk of hospitalized infection, by subgroup. Figure S5: 90-d risk of infection-related mortality, by subgroup. Figure S6: 90-d risk of acute MI, by subgroup. Figure S7: 90-d risk of hospitalized stroke, by subgroup. Figure S8: 90-d risk of CV-related death, by subgroup. Figure S9: 90-d risk of composite CV outcome, by subgroup. Figure S10: Sensitivity of results to definition of infection outcomes. Figure S11: Sensitivity of composite infection outcome to model specification, overall and in patients with HD catheter. Figure S12: Cumulative RD of composite infection outcome, stratified by iron use at baseline and catheter use. Figure S13: Cumulative RD of composite infection outcome, bolus vs maintenance dosing, stratified by formulation type and catheter use. Figure S14: 90-d risk of composite infection outcome, bolus vs no iron, by subgroup. Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients Figure S15: 90-d risk of composite infection outcome, maintenance dosing vs no iron, by subgroup. Note: The supplementary material accompanying this article (http://dx.doi.org/10.1053/j.ajkd.2015.07.026) is available at www. ajkd.org
REFERENCES 1. US Renal Data System. USRDS 2009 Annual Data Report: Atlas of End-Stage Renal Disease in the United States. Bethesda, MD: US Renal Data System, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 2009. 2. Collins AJ, Li S, St Peter W, et al. Death, hospitalization, and economic associations among incident hemodialysis patients with hematocrit values of 36 to 39%. J Am Soc Nephrol. 2001;12(11): 2465-2473. 3. Fishbane S, Maesaka JK. Iron management in end-stage renal disease. Am J Kidney Dis. 1997;29(3):319-333. 4. Besarab A, Coyne DW. Iron supplementation to treat anemia in patients with chronic kidney disease. Nat Rev Nephrol. 2010;6(12):699-710. 5. Faich G, Strobos J. Sodium ferric gluconate complex in sucrose: safer intravenous iron therapy than iron dextrans. Am J Kidney Dis. 1999;33(3):464-470. 6. Fletes R, Lazarus JM, Gage J, Chertow GM. Suspected iron dextran-related adverse drug events in hemodialysis patients. Am J Kidney Dis. 2001;37(4):743-749. 7. Danielson BG. Structure, chemistry, and pharmacokinetics of intravenous iron agents. J Am Soc Nephrol. 2004;15(suppl 2): S93-S98. 8. Aronoff GR. Safety of intravenous iron in clinical practice: implications for anemia management protocols. J Am Soc Nephrol. 2004;15(suppl 2):S99-S106. 9. Brewster UC. Intravenous iron therapy in end-stage renal disease. Semin Dial. 2006;19(4):285-290. 10. Kshirsagar AV, Freburger J, Ellis A, Wang L, Winkelmayer W, Brookhart MA. The comparative short-term effectiveness of iron dosing and formulations in us hemodialysis patients. Am J Med. 2012;126(6):541.e1-541.e14. 11. Coppol E, Shelly J, Cheng S, Kaakeh Y, Shepler B. A comparative look at the safety profiles of intravenous iron products used in the hemodialysis population (February). Ann Pharmacother. 2011;45(2):241-247. 12. US Renal Data System. Researcher’s Guide to the USRDS Database: 2010 ADR Edition. http://www.usrds.org/2010/rg/A_ intro_sec_1_10.pdf. Accessed September 11, 2015. 13. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8-27.
Am J Kidney Dis. 2016;67(1):119-127
14. Wood SN. Thin plate regression splines. J R Stat Soc B. 2003;65(1):95-114. 15. Brookhart MA, Freburger JK, Ellis AR, Wang L, Winkelmayer WC, Kshirsagar AV. Infection risk with bolus versus maintenance iron supplementation in hemodialysis patients. J Am Soc Nephrol. 2013;24(7):1151-1158. 16. R: A Language and Environment for Statistical Computing [computer program]. Vienna, Austria: R Foundation for Statistical Computing; 2014. 17. Westreich D, Cole SR. Invited commentary: positivity in practice. Am J Epidemiol. 2010;171(6):674-677; discussion 678-681. 18. Kshirsagar AV, Freburger JK, Ellis AR, Wang L, Winkelmayer WC, Brookhart MA. Intravenous iron supplementation practices and short-term risk of cardiovascular events in hemodialysis patients. PLoS One. 2013;8(11):e78930. 19. Besarab A, Frinak S, Yee J. An indistinct balance: the safety and efficacy of parenteral iron therapy. J Am Soc Nephrol. 1999;10(9):2029-2043. 20. Cieri E. Does iron cause bacterial infections in patients with end stage renal disease? ANNA J. 1999;26(6):591-596. 21. Fishbane S. Review of issues relating to iron and infection. Am J Kidney Dis. 1999;34(4)(suppl 2):S47-S52. 22. Hoen B. Iron and infection: clinical experience. Am J Kidney Dis. 1999;34(4)(suppl 2):S30-S34. 23. Patruta SI, Horl WH. Iron and infection. Kidney Int Suppl. 1999;69:S125-S130. 24. Jurado RL. Iron, infections, and anemia of inflammation. Clin Infect Dis. 1997;25(4):888-895. 25. Maynor L, Brophy DF. Risk of infection with intravenous iron therapy. Ann Pharmacother. 2007;41(9):1476-1480. 26. Zager RA, Johnson AC, Hanson SY. Parenteral iron therapy exacerbates experimental sepsis. Kidney Int. 2004;65(6):2108-2112. 27. Collins AJ, Ebben J, Ma JZ, Xia H. Iron dosing patterns and mortality [abstract]. J Am Soc Nephrol. 1998;9:250A. 28. Sirken G, Raja R, Rizkala AR. Association of different intravenous iron preparations with risk of bacteremia in maintenance hemodialysis patients. Clin Nephrol. 2006;66(5):348-356. 29. Kooistra MP, Kersting S, Gosriwatana I, et al. Nontransferrin-bound iron in the plasma of haemodialysis patients after intravenous iron saccharate infusion. Eur J Clin Invest. 2002;32(suppl 1):36-41. 30. Rooyakkers TM, Stroes ES, Kooistra MP, et al. Ferric saccharate induces oxygen radical stress and endothelial dysfunction in vivo. Eur J Clin Invest. 2002;32(suppl 1):9-16. 31. Parkkinen J, von Bonsdorff L, Peltonen S, GronhagenRiska C, Rosenlof K. Catalytically active iron and bacterial growth in serum of haemodialysis patients after i.v. iron-saccharate administration. Nephrol Dial Transplant. 2000;15(11):1827-1834.
127
A
nemia is common among patients with end-stage renal disease (ESRD)1 and is associated with increased morbidity, mortality, and risk for hospitalization.2 The anemia seen in these patients is primarily caused by impaired production of endogenous renal erythropoietin. It is worsened by depletion of iron reserves caused by hemodialysis-related blood loss and poor intestinal absorption of iron.3 Erythropoiesisstimulating agents are the primary treatment for the anemia of ESRD. The iron deficiency is addressed
through administration of intravenous (IV) iron formulations and, much less effectively, oral iron supplementation.4 Currently, there are 6 formulations of IV iron available on the US market: low-molecular-weight iron dextran, high-molecular-weight iron dextran, iron sucrose, sodium ferric gluconate, ferumoxytol, and ferric carboxymaltose. High-molecular-weight iron dextran has been linked to hypersensitivity reactions, including anaphylaxis,5,6 and carries a black box advisory on its
From the 1Department of Epidemiology, UNC Gillings School of Global Public Health UNC Chapel Hill; 2Cecil G. Sheps Center for Health Services Research, University of North Carolina, Chapel Hill, NC; 3Selzman Institute for Kidney Health, Section of Nephrology, Department of Medicine, Baylor College of Medicine, Houston, TX; and 4University of North Carolina Kidney Center, UNC School of Medicine, Chapel Hill, NC. Received February 2, 2014. Accepted in revised form July 15, 2015. Originally published online September 15, 2015. Because an author of this article is an editor for AJKD, the peerreview and decision-making processes were handled entirely by an
Associate Editor (Michel Jadoul, MD) who served as Acting Editor-in-Chief. Details of the journal’s procedures for potential editor conflicts are given in the Information for Authors & Journal Policies. Address correspondence to M. Alan Brookhart, PhD, Department of Epidemiology, 2105F McGavran-Greenberg, CB#7435, Chapel Hill, NC 27599. E-mail: [email protected] 2016 by the National Kidney Foundation, Inc. Published by Elsevier Inc. All rights reserved. 0272-6386 http://dx.doi.org/10.1053/j.ajkd.2015.07.026
Am J Kidney Dis. 2016;67(1):119-127
119
Brookhart et al
label. Consequently, there is currently little use of iron dextran in the US dialysis population. Ferumoxytol was approved by the US Food and Drug Administration in 2009 and is not widely used; ferric carboxymaltose was approved in 2013, but remains labeled for use in non2dialysis-dependent chronic kidney disease. Thus, the iron formulations most commonly used in the US ESRD population are sodium ferric gluconate and iron sucrose. Sodium ferric gluconate and iron sucrose preparations are iron-carbohydrate complexes or colloids of varying molecular size with a spheroid ironcarbohydrate core. Because of differences in core size, carbohydrate chemistry, and molecular weight, the 2 formulations have different pharmacokinetic and pharmacodynamic properties.7 Despite concerns about oxidative stress and infection risk associated with the use of IV iron,8,9 pharmacologic differences between the different iron formulations,7 and apparent differences in effectiveness,10 there is currently little information about the comparative safety of ferric gluconate versus iron sucrose.11 To address this evidence gap, we conducted a large-scale comparative safety study of ferric gluconate versus iron sucrose in a cohort of patients undergoing maintenance hemodialysis.
antibiotic use. The USRDS is a national data system funded by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) that collects, analyzes, and distributes information about the treatment of ESRD. USRDS includes data from the Medical Evidence Report Form, the Medicare Enrollment database, the ESRD Death Notification Form, and the standard analytic files, which contain final action claims data.12 These data were used to obtain information about demographic characteristics, health care use (eg, hospitalizations and outpatient care), and additional clinical characteristics (eg, comorbid conditions). We examined clinical data from 2 years (2004-2005) during which the dialysis provider used both ferric gluconate and iron sucrose. After 2005, the dialysis provider used mostly iron sucrose.
METHODS
We first identified center-based outpatient hemodialysis patients defined as follows: individuals who had undergone at least 9 months of maintenance dialysis (which accounted for the 6-month baseline period and an additional 3 months from the initiation of dialysis therapy to allow for stability in the Centers for Medicare & Medicaid Services claims processing),12 received hemodialysis in a dialysis facility, were covered by Medicare Parts A and B, and had at least one TSAT measurement during January 30, 2004, to November 30, 2005 (the November 30 date was chosen to allow for the 1-month exposure period and at least one day of followup). Patients were excluded if they had polycystic kidney disease because anemia management may be considerably different in these patients. TSAT records were excluded if: (1) iron dextran or both ferric gluconate and iron sucrose were delivered in the exposure period; (2) the duration of Part A claims prior to index was insufficient (ie, ,120 days of Part A claims), suggesting
Data Sources The data used for this study came from the clinical research database of a large dialysis provider and the US Renal Data System (USRDS), which were merged at the patient level. The dialysis provider owns and manages more than 1,500 outpatient dialysis facilities located throughout the United States in urban, rural, and suburban areas. Their clinical database captures detailed clinical, laboratory, and treatment data for patients receiving care at all their dialysis units. All data are collected using standardized electronic health record systems. For this study, we used the clinical data to obtain detailed information about iron formulation use and dosing, erythropoiesis-stimulating agent use and dosing, clinical laboratory values (eg, hemoglobin, transferrin saturation [TSAT], and serum ferritin), current vascular access, and recent IV
Study Design We used a retrospective cohort design with repeated measures on iron treatments and outcomes. The index date of the 1-month exposure assessment period was anchored on a TSAT measurement because this information is used to guide iron administration. The 6-month baseline period prior to the TSAT measurement was used to identify potential confounders and effect modifiers. Patients were followed up for 3 months after the exposure period. Eligible patients could contribute multiple exposure/follow-up periods (Fig 1). Our design was implicitly attempting to mimic a randomized trial in which patients prescribed a 1-month course of iron were randomly assigned to receive the iron as either ferric gluconate or iron sucrose and then followed up for 90 days, with further iron treatment decisions left to the discretion of the physician.
Cohort Identification
second observation first observation
baseline
exposure
TSAT lab
follow-up
baseline
exposure
follow-up
TSAT lab
Figure 1. Study design schematic. Abbreviations: lab, laboratory; TSAT, transferrin saturation. 120
Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients incomplete data; (3) the TSAT measurement occurred during the follow-up period for a prior eligible TSAT; or (4) there were fewer than 9 dialysis sessions in the month preceding the index date or during the exposure period (suggesting that the individual was not receiving regular in-center hemodialysis). We also excluded records with missing covariate information.
Study Variables Outcomes We examined a variety of adverse clinical outcomes related to death, infection, and cardiovascular events. These were determined by examining the Medicare death notification and inpatient and outpatient claims data. The specific definition and data source for each outcome are presented in Table S1 (provided as online supplementary material). We examined 6 outcomes: death from any cause, infection-related hospitalization, infection-related death, hospitalization for myocardial infarction, hospitalization for stroke, and cardiovascular disease–related death. In addition, we created 2 composite outcomes: cardiovascular disease–related hospitalization or death and infection-related hospitalization or death.
Exposures Iron formulations were classified as ferric gluconate or iron sucrose based on native codes in the clinical database. The primary exposure contrast of interest was ferric gluconate versus iron sucrose. We also created a no-iron category for individuals who received no iron in the 1-month exposure period. We describe characteristics of the no-iron patients, but they were not otherwise used in the analysis.
Covariates We included a wide range of covariates in our analyses. Descriptions of the covariates and the manner in which they were defined are presented in Table S2. Covariates included demographic characteristics (eg, age, sex, race, Medicaid eligibility, census region, and year), clinical characteristics (eg, cause of ESRD, dialysis vintage, body mass index, type of vascular access, and number of hospital days), laboratory and anemia management variables (baseline values for hemoglobin, ferritin, TSAT, iron dose, epoetin alfa dose, and albumin; receipt of a blood transfusion; and epoetin alfa dose during exposure period), and several comorbid condition measures based on the Elixhauser classification13 and content expertise of the investigative team. Because of the potential relationship between iron use and infections, we created 4 “history of infection” variables: history of pneumonia, sepsis, or vascular access infection during the baseline period and history of any infection or use of IV antibiotics in the last month. Due to the extensive list of comorbid conditions, we selected certain ones to include in a parsimonious model and others to add during sensitivity analyses.
Statistical Analysis To describe the treatment groups, we calculated mean values and frequencies of all covariates within each treatment group. To assess the relationship between formulation choice and outcomes, we estimated unadjusted and adjusted cumulative incidence functions using Kaplan-Meier methods. Adjustment was obtained by applying inverse probability of treatment weighting to the Kaplan-Meier estimator. Weights were estimated using a generalized additive logistic model of the treatment choice, in which all covariates in Table S1 were included as potential confounders. Age, serum albumin level, body mass index, TSAT, and hemoglobin level were entered into the propensity score model using smoothing splines.14 Patients were censored after 90 days, at loss to follow-up, or administratively on December 31, 2005. Hospitalization outcomes were further censored at loss of Medicare Parts A and B eligibility. We graphed the survival function and the Am J Kidney Dis. 2016;67(1):119-127
difference between survival functions (risk difference [RD]) for each day during the 90-day follow-up period. Confidence intervals (CIs) were estimated for the daily RDs between groups using a nonparametric clustered bootstrap based on 250 resamples. Because previous research has reported increased infection risk for bolus relative to maintenance dosing,15 we assessed the contrast between these dosing strategies among patients receiving ferric gluconate and iron sucrose separately. Bolus and maintenance dosing were defined as in the previous report.15 Separate propensity scores were estimated for this contrast. All statistical analyses were conducted using R Statistical Software, version 3.3.16
Subgroup Analysis We examined several demographic and clinical subgroups (Table S3). Individuals were categorized based on race (black and nonblack), sex, hemodialysis catheter use, history of infection in the last month of baseline, TSAT at baseline (,20%, 20%-,50%, and $50%), time since reported incidence of ESRD (,1, 1-,4, and $4 years), hemoglobin level at baseline (,10, 10-12, and .12 g/dL), and ferritin level at baseline (,200, 200-500, and .500 mg/L).
Sensitivity Analyses Because our definition of hospitalization for infection was specific to sepsis, vascular access infection, and pneumonia, in a sensitivity analysis, we considered a broader infection definition, hospitalization for infection of any major organ system, and also receipt of IV antibiotics (Table S1). We also assessed the sensitivity of our results to the specification of the propensity score model by examining how the addition of other potentially relevant covariates to our primary model affected estimates. The additional variables that we considered were cause of ESRD, calendar year, region, Medicaid eligibility, and a number of additional comorbid conditions (Table S2). In this analysis, we increased the degrees of freedom (flexibility) of the regression splines. We also conducted a sensitivity analysis in which we stratified on iron use (vs no use) in the past month. This analysis was conducted to examine whether there may be delayed effects of formulation choice on outcomes. Finally, we conducted 3 sensitivity analyses with shorter iron exposure and/or follow-up periods: (1) a 2-week exposure and 6week follow-up period, (2) a 1-week exposure and 6-week followup period, and (3) a 1-month exposure and 6-week follow-up period. An institutional review board of the University of North Carolina approved this study. The study design, analytic methods, outcomes, and covariate definitions were prespecified in a protocol submitted to the Agency for Healthcare Research and Quality, which funded this study. Some aspects of the study were modified in response to comments from journal reviewers.
RESULTS Overall, 66,207 patients met study entry requirements and contributed data for 278,357 ironexposure/follow-up periods (Fig S1). Individuals who received ferric gluconate were more likely to be black and to reside in the South (Table 1). The groups were fairly similar with respect to many laboratory values and comorbid conditions. However, ferric gluconate users tended to have a slightly higher prevalence of some comorbid conditions and were more likely to have had any infection in the past month (12.2% vs 11.5%) and a vascular access infection in the past 6 months (14.7% vs 12.2%). Ferric gluconate users were less likely to have diabetes (49.3% vs 52.7%), but more likely to have a hemodialysis catheter (31.2% vs 121
Brookhart et al Table 1. Characteristics of Exposure Periods by Exposure Group Ferric Gluconate (n 5 31,726; 11.4%)
Iron Sucrose (n 5 136,233; 48.9%)
No Iron Use (n 5 110,398; 39.7%)
Mean age, y Female sex Race White Black
45.9 45.0
46.0 49.5
46.9 46.4
45.0 51.3
49.5 44.0
46.4 46.3
Medicaid Region Midwest Northeast South West
51.9
51.3
52.7
13.8 15.4 58.5 11.8
16.5 11.5 47.4 24.1
16.2 12.6 49.2 21.8
ESRD cause Diabetes Glomerulonephritis Hypertension
43.0 13.1 31.7
45.1 12.5 31.1
42.4 13.5 31.8
4.1 6 4.2 27.0 6 7.1 31.2 3.8 6 0.4 12.1 6 1.2 27.4 6 11.2 536 6 356 240 6 256 291 6 278 102 6 107 97 6 106 6.4 0.6 6 1.9 12.2
4.0 6 4.2 27.0 6 6.8 23.9 3.9 6 0.4 12.3 6 1.3 28.3 6 11.2 493 6 305 245 6 275 294 6 288 93 6 94 90 6 94 6.0 0.6 6 1.9 11.5
4.7 6 4.5 26.4 6 6.5 23.4 3.9 6 0.4 12.2 6 1.3 34.7 6 15.6 764 6 541 90 6 230 060 74 6 85 78 6 89 6.2 0.6 6 1.8 10.4
10.1 9.4 14.7 49.3 10.1 3.6 18.2 8.6 5.0
9.7 10.5 12.2 52.7 9.9 3.2 17.0 8.5 4.9
9.4 9.2 10.3 49.0 9.5 3.1 15.0 8.6 4.0
Characteristic
Dialysis vintage, y BMI, kg/m2 Hemodialysis catheter use Baseline albumin, g/dL Baseline hemoglobin, g/dL Index TSAT, % Baseline ferritin, mg/L Iron dose during last month of baseline, mg Iron dose during exposure period, mg EPO dose at baseline, 1,000 U/mo EPO dose during exposure period, 1,000 U/mo Blood transfusion during baseline period No. of hospital days during past montha Infection during past montha Comorbid conditions during 6-mo baseline period Pneumonia Sepsis Vascular access infection Diabetes Ischemic stroke Myocardial infarction COPD, asthma Cancer GI bleeding
Note: N 5 278,357. Unless otherwise indicated, values for categorical variables are given as percentage; for continuous variables, as mean 6 standard deviation. Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; EPO, epoetin alfa; ESRD, end-stage renal disease; GI, gastrointestinal; TSAT, transferrin saturation. a The month before the start of the exposure period.
23.9%). Mean iron dose during the exposure period did not vary by formulation type and was 291 mg for ferric gluconate and 294 mg for iron sucrose. For the adjusted analysis, we computed propensity scores for the overall population receiving iron. In Figure S2, we present the estimated densities of the propensity scores in each treatment group. There was a high degree of overlap between propensity score distributions across all subgroups, suggesting good comparability between treatment groups.17 Counts of events, total person-years of follow-up, and unadjusted and adjusted 90-day RDs for all outcomes are presented 122
in Table 2. Event rates were similar across treatment groups; the only notable difference was observed with infection-related death, for which ferric gluconate was associated with a decreased rate (relative rate, 0.78). In the adjusted estimate of the absolute 90-day RD, we observed slightly lower risk associated with ferric gluconate across all outcomes, but RDs were small in magnitude, ranging from 20.1% to 20.6%, corresponding to 1 to 6 fewer events over the 90-day followup period per 1,000 patients treated. We present the inverse probability of treatment-weighted survival curves for all outcomes in Fig S3. Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients Table 2. The 90-Day Outcome Rates, Rate Ratios, and Risk Differences, Ferric Gluconate Versus Iron Sucrose
Outcome
All-cause death
Treatment Group
IS FG Infection-related IS hospitalization FG Infection-related IS mortality FG Composite infection IS outcome FG Hospitalized stroke IS FG Acute MI IS FG CV death IS FG Composite CV IS outcome FG
Person Time, y
Events
Event Rate (per 100 person-y)
32,749 7,627 31,778 7,394 32,749 7,627 31,778 7,394 32,523 7,582 32,584 7,590 32,749 7,627 32,414 7,554
6,102 1,468 8,510 2,062 698 126 8,769 2,106 1,692 355 1,181 281 2,473 596 4,933 1,153
18.6 19.2 26.8 27.9 1.7 2.1 27.6 28.5 5.2 4.7 3.6 3.7 7.6 7.8 15.2 15.3
Unadjusted Relative Rate (95% CI)
90-d Unadjusted RD (95% CI)a
90-d Adjusted RD (95% CI)a
—
—
1.03 (0.98 to 1.09)
0.1 (20.1 to 0.4)
20.3 (21.0 to 0.3)
1.04 (0.99 to 1.10)
0.3 (0.0 to 0.6)
20.6 (21.6 to 0.2)
— —
0.78 (0.64 to 0.94)
—
1.03 (0.98 to 1.09)
—
0.90 (0.80 to 1.01)
—
— —
— — —
20.1 (20.2 to 0.0) 20.1 (20.3 to 0.0)
—
0.2 (20.1 to 0.5)
—
—
20.6 (21.7 to 0.1)
—
20.1 (20.3 to 0.0) 20.2 (20.4 to 0.0)
—
—
1.02 (0.89 to 1.17)
0.0 (20.1 to 0.2)
20.1 (20.3 to 0.1)
1.03 (0.95 to 1.13)
0.1 (20.1 to 0.2)
20.1 (20.4 to 0.1)
1.00 (0.94 to 1.07)
0.0 (20.2 to 0.3)
20.3 (21.0 to 0.1)
—
—
—
Abbreviations: CI, confidence interval; CV, cardiovascular; FG, ferric gluconate; IS, iron sucrose; MI, myocardial infarction; RD, risk difference. a RDs estimated using inverse probability treatment-weighted Kaplan-Meier functions with bootstrap CIs.
The adjusted 90-day RDs and CIs for all-cause mortality and the composite infection outcome for all subgroups are presented in forest plots in Figs 2 and 3. Forest plots for the remaining outcomes are presented in Figs S4 to S9. We observed a slightly decreased risk for both infection-related mortality (RD, 20.3%; 95% CI, 20.5% to 0.0%) and infection-related hospitalization (RD, 21.5%; 95% CI, 22.3% to 20.6%) among patients with a hemodialysis catheter who were treated with ferric gluconate. There were several additional subgroups in which ferric gluconate was associated with decreased risk for infection-related mortality. Results for the ferric gluconate versus iron sucrose comparisons using alternative infection definitions (Fig S10) indicated that ferric gluconate was associated with decreased use of IV antibiotics overall (RD, 22.5%; 95% CI, 24.2% to 21.2%) and among patients with a hemodialysis catheter (RD, 24.5%; 95% CI, 25.7% to 23.4%). The composite infection results within the group of patients with a hemodialysis catheter were robust to the addition of covariates beyond our a priori–specified multivariable models (Fig S11). In analyses stratified on past iron use and catheter status, we observed similar associations between formulation choice and risk for the composite infection outcome (Fig S12). In Fig 4, we present cumulative risk curves of infection-related hospitalization or mortality associated with bolus versus maintenance dosing stratified by formulation type and catheter use. Graphs of the differences between cumulative risk curves and 95% CIs of the differences are provided in Fig S13. These analyses showed that bolus dosing was associated with Am J Kidney Dis. 2016;67(1):119-127
increased 90-day risk among both ferric gluconate and iron sucrose users, with the risk particularly pronounced among patients with hemodialysis catheters. In Figs S14 and S15, we present contrasts in risk for the composite infection outcome between maintenance iron and no iron and bolus iron and no iron, overall and by catheter use, respectively.
DISCUSSION We conducted a large-scale comparative study of the safety of different IV iron formulations administered in usual-care settings to typical patients undergoing maintenance hemodialysis in the United States. Although we examined a large number of outcomes and subgroups, we found only a few differences between the 2 formulations. Their use was associated with nearly identical risks for myocardial infarction, cardiovascular mortality, and all-cause mortality. Risks for these events were statistically similar within the overall population, as well as within all patient subgroups considered. The lack of an association between formulation choice and short-term risk for cardiovascular events is consistent with a recent study that found no association between any iron exposure and short-term risk for cardiovascular events.18 We found evidence suggesting a modestly increased risk for infection outcomes among patients exposed to iron sucrose versus ferric gluconate. In the overall population, the differences in infection risk between formulations were statistically nonsignificant or small in magnitude. However, among patients with a hemodialysis catheter, a subgroup of patients known to be particularly prone to infection, we observed differences 123
Brookhart et al Iron Ferric % Risk Sucrose Gluconate Difference Subgroup Cohort (N) Cohort (N) (95% CI) Overall 136233 31726 −0.3 (−1.0, 0.3) Female 62607 14556 −0.1 (−0.5, 0.3) Male 73626 17170 −0.4 (−1.4, 0.6) Black 59909 16282 −0.5 (−1.3, 0.2) Non Black 76324 15444 0.1 (−0.3, 0.6) Hemodialysis Catheter 32576 9907 −0.4 (−0.9, 0.2) Fistula or Graft 103657 21819 −0.2 (−1.0, 0.5) Recent Infection 15673 3859 −0.2 (−1.2, 0.8) Hemoglobin <= 10g/dL 6432 1585 −0.5 (−2.1, 1.1) Hemoglobin in 10−12g/dL 51569 13040 −0.2 (−0.6, 0.3) Hemoglobin >12g/dL 78232 17101 −0.3 (−1.1, 0.5) TSat <= 20% 32957 8887 −0.3 (−0.9, 0.3) TSat in 20−50% 98540 21759 −0.3 (−1.1, 0.4) TSat >50% 4736 1080 0.4 (−1.2, 1.7) Ferritin <= 200ng/mL 17874 2420 −0.7 (−1.5, 0.2) Ferritin in 200−500ng/mL 115074 28600 −0.3 (−1.1, 0.4) Ferritin >500ng/mL 60790 16366 0.0 (−0.3, 0.3) Vintage <= 1 year 39943 9260 0.2 (−0.4, 0.6) Vintage in 1−3 years 40437 9347 0.1 (−0.4, 0.6) Vintage > 3 years 55853 13119 −0.8 (−1.8, 0.3) No Iron Use in Previous Month 22173 5732 0.1 (−0.6, 0.7) −2
−1.5
−1
−0.5
0
0.5
1
1.5
Figure 2. The 90-day risk for all-cause mortality, ferric gluconate versus iron sucrose, by subgroup. Risk differences were estimated using inverse probability of treatment-weighted Kaplan-Meier functions with bootstrap confidence intervals (CIs). Abbreviation: TSat, transferrin saturation.
in both infection-related hospitalization and infectionrelated mortality, with higher rates in patients exposed to iron sucrose. In a sensitivity analysis, we also observed greater use of antibiotics among patients treated with iron sucrose. The differences between the
formulations were found despite the higher prevalence of catheters in the ferric gluconate group. Concerns over iron and infection in the chronic kidney disease population were raised many years ago as more patients began receiving iron as adjuvant
% Risk Ferric Iron Difference Sucrose Gluconate (95% CI) Cohort (N) Cohort (N) Subgroup 31726 −0.6 (−1.7, 0.1) 136233 Overall 14556 −0.2 (−0.6, 0.4) 62607 Female 17170 −0.9 (−2.2, 0.3) 73626 Male Black 59909 16282 −1.3 (−2.9, 0.2) 15444 0.0 (−0.5, 0.4) 76324 Non Black 9907 −1.5 (−2.4, −0.6) Hemodialysis Catheter 32576 103657 21819 −0.2 (−1.1, 0.6) Fistula or Graft 3859 −0.9 (−2.2, 0.6) 15673 Recent Infection 1.0 (−0.9, 2.8) 1585 Hemoglobin <= 10g/dL 6432 13040 −0.2 (−0.7, 0.3) 51569 Hemoglobin in 10−12g/dL 17101 −1.0 (−2.2, 0.2) Hemoglobin >12g/dL 78232 32957 8887 −0.1 (−0.9, 0.6) TSat <= 20% 21759 −0.8 (−1.9, 0.2) 98540 TSat in 20−50% 0.3 (−1.2, 2.0) 1080 TSat >50% 4736 17874 2420 −1.4 (−2.6, −0.3) Ferritin <= 200ng/mL 28600 −0.5 (−1.6, 0.4) 115074 Ferritin in 200−500ng/mL 16366 −0.2 (−0.7, 0.2) 60790 Ferritin >500ng/mL 0.0 (−0.7, 0.5) 9260 39943 Vintage <= 1 year 9347 −0.2 (−0.9, 0.4) Vintage in 1−3 years 40437 55853 13119 −1.3 (−3.0, 0.3) Vintage > 3 years 5732 −0.4 (−1.2, 0.4) 22173 No Iron Use in Previous Month −2
−1
0
1
2
Figure 3. The 90-day risk for the composite infection outcome, ferric gluconate versus iron sucrose, by subgroup. Risk differences were estimated using inverse probability of treatment-weighted Kaplan-Meier functions with bootstrap confidence intervals (CIs). Abbreviation: TSat, transferrin saturation. 124
Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients
Ferric Gluconate Subgroup
Iron Sucrose Subgroup
0.15
0.10
Cumulative Risk
0.05
Treatment Group
0.00 Ferric Gluconate, Catheter Subgroup
Iron Sucrose, Catheter Subgroup
Maintenance Bolus Dosing
0.15
0.10
0.05
0.00 30
60
90
30
60
90
Days Figure 4. Bolus versus maintenance dosing and cumulative risk for the composite infection outcome, stratified by formulation type and catheter use. Cumulative risk was estimated using inverse probability of treatment-weighted Kaplan-Meier functions.
treatment to ESAs.19-23 The part that iron plays in bacterial growth and infection risk has biological plausibility and has been shown in animal models,20,24-26 but there are relatively few studies of iron and infections in large chronic kidney disease populations. One study found that frequent iron administration is associated with increased risk for infection-related mortality in ESRD.27 Another recent report found that bolus (ie, repletion) dosing of iron is associated with increased risk for infection-related hospitalizations compared with maintenance dosing.15 Finally, a small clinical study determined that cumulative iron exposure is associated with increased risk for bacteremia.28 Our study raises the possibility that the infection risk from IV iron administration may differ by the IV iron formulation agent used. The putative differences in infection risk between ferric gluconate and iron sucrose could be explained by pharmacokinetic differences between the 2 iron formulations. Ferric gluconate is taken up quickly by the reticuloendothelial system and thus is eliminated rapidly from plasma.7 In contrast, iron sucrose is taken up more slowly and thus has a longer plasma half-life.7 These differences in pharmacology suggest that iron Am J Kidney Dis. 2016;67(1):119-127
sucrose may result in more iron in circulation, both free and bound to transferrin. Two clinical studies have reported abundant non–transferrin-bound iron after iron sucrose administration.29,30 Increasing the amount of iron in circulation may result in differences in both safety and effectiveness. A recent report using the same data found that iron sucrose was slightly more effective than ferric gluconate at increasing hemoglobin levels and iron indexes.10 Also, a small clinical study found that iron sucrose administration resulted in transferrin oversaturation in some patients and the growth of Staphylococcus epidermidis in vitro that was proportional to the amount of free iron in the sample.31 Despite some biological plausibility, the associations that we have observed between iron sucrose and infection risk must be interpreted cautiously. Given that our study was nonexperimental, it is susceptible to bias resulting from residual confounding. Although we have included a wide range of variables in our propensity score model, there may be some uncontrolled differences between patient groups. Our results may also have been confounded by uncontrolled aspects of medical practice. For example, facilities that 125
Brookhart et al
tended to use more ferric gluconate may also have used other practices that could alter infection risk. Our study may have been vulnerable to this bias because some patients in our study came from a smaller dialysis organization that had recently merged with the larger chain. We observed some difference in formulation preference between the groups of units, which may have been correlated with coding practices or unobserved aspects of practice or patient case-mix. Because our outcomes and covariate definitions are based on International Classification of Diseases, Ninth Revision codes and cause-of-death reporting, they may have been misclassified to some degree, possibly in a differential manner. Finally, given the large number of associations examined, some falsepositive associations would be expected. We did not attempt to shrink subgroup effects in any way to account for this issue. Our study had some limitations beyond those related to internal validity. First, our study focused on short-term effects. Some of the hypothesized effects of iron on cardiovascular risk resulting from free radical damage to the vasculature would be a function of longer term cumulative exposure.8 Second, our study design necessitated survival until 9 months after dialysis therapy initiation; thus, our results may not generalize to patients new to dialysis therapy. The limitations of our study are counterbalanced by 2 key strengths. First, by focusing on shorter term associations of a well-defined exposure with covariates defined prior to exposure and outcomes ascertained after the exposure assessment period ended, we removed or diminished many sources of bias typical in nonexperimental studies of longitudinal exposures (eg, immortal person-time bias, selection bias, and timevarying confounding). Second, we used detailed data, which let us control for essential clinical and laboratory variables and also capture important events outside the clinic (eg, hospitalization and death). In summary, we found that the 2 most commonly used iron formulations were similar with respect to most clinical outcomes. However, we found increased risk for infection associated with iron sucrose versus ferric gluconate among patients with a hemodialysis catheter. Given the nonexperimental nature of our study and the many outcomes and subgroup effects examined, this finding should motivate additional studies using different data or employing different research designs.
ACKNOWLEDGEMENTS Some of the data reported here have been supplied by the USRDS. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the US government or the USRDS. The authors thank DaVita Clinical Research, which had 126
no role in the design or implementation of this study or on the decision to publish, for providing data for this study. Support: This project was funded under contract no. HHSA290200500401 Task Order #5 from the Agency for Healthcare Research and Quality, US Department of Health and Human Services. Dr Winkelmayer was supported by grant R01 DK090181 from the National Institutes of Health (NIH), NIDDK. He also receives research and salary support through the endowed Gordon A. Cain Chair in Nephrology at Baylor College of Medicine. Dr Brookhart receives investigator-initiated research funding from the NIH (R01 AG042845, R21 HD080214, and R01 AG023178) and through contracts with the Patient Centered Outcomes Research Institute. The study sponsors did not have any role in study design; collection, analysis, and interpretation of data; writing the report; or the decision to submit the report for publication. Financial Disclosure: Dr Brookhart has received investigatorinitiated grant support from Amgen and served as a scientific advisor for Pfizer, Merck, Amgen, and Rockwell Medical, but has not accepted personal compensation for this service (honoraria received by institution or donated). Drs Freburger, Kshirsagar, and Ellis have received investigator-initiated grant support from Amgen. Dr Kshirsagar served on a Fresenius Advisory Board and Dr Ellis has received investigator-initiated grant support from Merck and the UNC Center for Pharmacoepidemiology. Dr Winkelmayer has served as an advisor or consultant to Amgen, Astra-Zeneca, Bayer, Keryx, Medgenics, Medtronic, MitsubishiTanabe, and Rockwell Pharma. Contributions: Research idea and study design: MAB, JKF, AVK, WCW; data acquisition: MAB; data analysis/interpretation: MAB, ARE, LW, JKF, WCW; statistical analysis: MAB, ARE, LW; supervision or mentorship: MAB, AVK, WCW. Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved. MAB takes responsibility that this study has been reported honestly, accurately, and transparently; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
SUPPLEMENTARY MATERIAL Table S1: Study outcomes. Table S2: Definition of covariates. Table S3: Definition of subgroups. Figure S1: Flowchart of patients through study. Figure S2: Propensity score distributions, overall and by treatment group. Figure S3: Inverse-probability of treatment-weighted survival curves. Figure S4: 90-d risk of hospitalized infection, by subgroup. Figure S5: 90-d risk of infection-related mortality, by subgroup. Figure S6: 90-d risk of acute MI, by subgroup. Figure S7: 90-d risk of hospitalized stroke, by subgroup. Figure S8: 90-d risk of CV-related death, by subgroup. Figure S9: 90-d risk of composite CV outcome, by subgroup. Figure S10: Sensitivity of results to definition of infection outcomes. Figure S11: Sensitivity of composite infection outcome to model specification, overall and in patients with HD catheter. Figure S12: Cumulative RD of composite infection outcome, stratified by iron use at baseline and catheter use. Figure S13: Cumulative RD of composite infection outcome, bolus vs maintenance dosing, stratified by formulation type and catheter use. Figure S14: 90-d risk of composite infection outcome, bolus vs no iron, by subgroup. Am J Kidney Dis. 2016;67(1):119-127
Comparative Safety of Iron Formulations in HD Patients Figure S15: 90-d risk of composite infection outcome, maintenance dosing vs no iron, by subgroup. Note: The supplementary material accompanying this article (http://dx.doi.org/10.1053/j.ajkd.2015.07.026) is available at www. ajkd.org
REFERENCES 1. US Renal Data System. USRDS 2009 Annual Data Report: Atlas of End-Stage Renal Disease in the United States. Bethesda, MD: US Renal Data System, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 2009. 2. Collins AJ, Li S, St Peter W, et al. Death, hospitalization, and economic associations among incident hemodialysis patients with hematocrit values of 36 to 39%. J Am Soc Nephrol. 2001;12(11): 2465-2473. 3. Fishbane S, Maesaka JK. Iron management in end-stage renal disease. Am J Kidney Dis. 1997;29(3):319-333. 4. Besarab A, Coyne DW. Iron supplementation to treat anemia in patients with chronic kidney disease. Nat Rev Nephrol. 2010;6(12):699-710. 5. Faich G, Strobos J. Sodium ferric gluconate complex in sucrose: safer intravenous iron therapy than iron dextrans. Am J Kidney Dis. 1999;33(3):464-470. 6. Fletes R, Lazarus JM, Gage J, Chertow GM. Suspected iron dextran-related adverse drug events in hemodialysis patients. Am J Kidney Dis. 2001;37(4):743-749. 7. Danielson BG. Structure, chemistry, and pharmacokinetics of intravenous iron agents. J Am Soc Nephrol. 2004;15(suppl 2): S93-S98. 8. Aronoff GR. Safety of intravenous iron in clinical practice: implications for anemia management protocols. J Am Soc Nephrol. 2004;15(suppl 2):S99-S106. 9. Brewster UC. Intravenous iron therapy in end-stage renal disease. Semin Dial. 2006;19(4):285-290. 10. Kshirsagar AV, Freburger J, Ellis A, Wang L, Winkelmayer W, Brookhart MA. The comparative short-term effectiveness of iron dosing and formulations in us hemodialysis patients. Am J Med. 2012;126(6):541.e1-541.e14. 11. Coppol E, Shelly J, Cheng S, Kaakeh Y, Shepler B. A comparative look at the safety profiles of intravenous iron products used in the hemodialysis population (February). Ann Pharmacother. 2011;45(2):241-247. 12. US Renal Data System. Researcher’s Guide to the USRDS Database: 2010 ADR Edition. http://www.usrds.org/2010/rg/A_ intro_sec_1_10.pdf. Accessed September 11, 2015. 13. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8-27.
Am J Kidney Dis. 2016;67(1):119-127
14. Wood SN. Thin plate regression splines. J R Stat Soc B. 2003;65(1):95-114. 15. Brookhart MA, Freburger JK, Ellis AR, Wang L, Winkelmayer WC, Kshirsagar AV. Infection risk with bolus versus maintenance iron supplementation in hemodialysis patients. J Am Soc Nephrol. 2013;24(7):1151-1158. 16. R: A Language and Environment for Statistical Computing [computer program]. Vienna, Austria: R Foundation for Statistical Computing; 2014. 17. Westreich D, Cole SR. Invited commentary: positivity in practice. Am J Epidemiol. 2010;171(6):674-677; discussion 678-681. 18. Kshirsagar AV, Freburger JK, Ellis AR, Wang L, Winkelmayer WC, Brookhart MA. Intravenous iron supplementation practices and short-term risk of cardiovascular events in hemodialysis patients. PLoS One. 2013;8(11):e78930. 19. Besarab A, Frinak S, Yee J. An indistinct balance: the safety and efficacy of parenteral iron therapy. J Am Soc Nephrol. 1999;10(9):2029-2043. 20. Cieri E. Does iron cause bacterial infections in patients with end stage renal disease? ANNA J. 1999;26(6):591-596. 21. Fishbane S. Review of issues relating to iron and infection. Am J Kidney Dis. 1999;34(4)(suppl 2):S47-S52. 22. Hoen B. Iron and infection: clinical experience. Am J Kidney Dis. 1999;34(4)(suppl 2):S30-S34. 23. Patruta SI, Horl WH. Iron and infection. Kidney Int Suppl. 1999;69:S125-S130. 24. Jurado RL. Iron, infections, and anemia of inflammation. Clin Infect Dis. 1997;25(4):888-895. 25. Maynor L, Brophy DF. Risk of infection with intravenous iron therapy. Ann Pharmacother. 2007;41(9):1476-1480. 26. Zager RA, Johnson AC, Hanson SY. Parenteral iron therapy exacerbates experimental sepsis. Kidney Int. 2004;65(6):2108-2112. 27. Collins AJ, Ebben J, Ma JZ, Xia H. Iron dosing patterns and mortality [abstract]. J Am Soc Nephrol. 1998;9:250A. 28. Sirken G, Raja R, Rizkala AR. Association of different intravenous iron preparations with risk of bacteremia in maintenance hemodialysis patients. Clin Nephrol. 2006;66(5):348-356. 29. Kooistra MP, Kersting S, Gosriwatana I, et al. Nontransferrin-bound iron in the plasma of haemodialysis patients after intravenous iron saccharate infusion. Eur J Clin Invest. 2002;32(suppl 1):36-41. 30. Rooyakkers TM, Stroes ES, Kooistra MP, et al. Ferric saccharate induces oxygen radical stress and endothelial dysfunction in vivo. Eur J Clin Invest. 2002;32(suppl 1):9-16. 31. Parkkinen J, von Bonsdorff L, Peltonen S, GronhagenRiska C, Rosenlof K. Catalytically active iron and bacterial growth in serum of haemodialysis patients after i.v. iron-saccharate administration. Nephrol Dial Transplant. 2000;15(11):1827-1834.
127