Boxes indicate interquartile range (IQR); horizontal lines, median. ESA indicates erythropoiesis-stimulating agent. Each center contributed up to 2 observations per period (1 per year). For 1999-2000, there were 7387 observations; 2001-2002, 8063 observations; 2003-2004, 8740 observations; and 2005-2006, 9351 observations.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Brookhart MA, Schneeweiss S, Avorn J, Bradbury BD, Liu J, Winkelmayer WC. Comparative Mortality Risk of Anemia Management Practices in Incident Hemodialysis Patients. JAMA. 2010;303(9):857–864. doi:10.1001/jama.2010.206
Context Controversy exists about optimal management of anemia in end-stage renal disease.
Objective To compare the mortality risk of different dialysis center–level patterns of anemia management.
Design, Setting, and Patients Using data from Medicare's end-stage renal disease program (1999-2007), we characterized each US dialysis center's annual anemia management practice by estimating its typical use of erythropoiesis-stimulating agents (ESAs) and intravenous iron in hemodialysis patients within 4 hematocrit categories. We used Cox proportional hazards regression to correlate center-level patterns of ESA and iron use with 1-year mortality risk in 269 717 incident hemodialysis patients.
Main Outcome Measure One-year all-cause mortality.
Results Monthly mortality rates were highest in patients with hematocrit less than 30% (mortality, 2.1%) and lowest for those with hematocrit of 36% or higher (mortality, 0.7%). After adjustment for baseline case-mix differences, dialysis centers that used larger ESA doses in patients with hematocrit less than 30% had lower mortality rates than centers that used smaller doses (highest vs lowest dose group: hazard ratio [HR], 0.94; 95% confidence interval [CI], 0.90-0.97). Centers that administered iron more frequently to patients with hematocrit less than 33% also had lower mortality rates (highest vs lowest quintile, HR, 0.95; 95% CI, 0.91-0.98). However, centers that used larger ESA doses in patients with hematocrit between 33% and 35.9% had higher mortality rates (highest vs lowest quintile, HR, 1.07; 95% CI, 1.03-1.12). More intensive use of both ESAs and iron was associated with increased mortality risk in patients with hematocrit of 36% or higher. These findings persisted across a range of secondary analyses.
Conclusions Greater ESA and iron use were associated with decreased mortality risk at lower hematocrit levels, in which mortality rates are the highest. Although the overall mortality rate was lower at higher hematocrit levels, elevated mortality risk was associated with greater use of ESAs and iron in these patients.
Appropriate use of erythropoiesis-stimulating agents (ESAs) and intravenous iron can effectively manage the anemia of chronic kidney disease and end-stage renal disease (ESRD),1-3 but several randomized trials have reported an increased risk of mortality and cardiovascular events in patients treated to achieve higher hematocrit levels.2-4 The earlier of these reports prompted the US Food and Drug Administration in March 2007 to issue a black box warning for all ESAs recommending that they be used at the lowest level necessary to prevent transfusions.5 Because these trials were conducted in selected patient populations, many of whom had chronic kidney disease but not ESRD, controversy remains about appropriate management of anemia in ESRD.6-8
We sought to address this evidence gap by conducting a study of the safety of ESA and iron regimens as they are prescribed in routine practice in a large, unselected population of ESRD patients. Standard nonexperimental studies of anemia management are subject to strong confounding by indication bias because ESA and iron doses are adjusted frequently in response to indications that are incompletely ascertained in available data.9 Information on ESA and iron use is also unavailable during the frequent hospitalizations experienced by dialysis patients, further complicating epidemiologic analysis.10
To address these problems, we adopted a fundamentally different analytic approach. We studied the potential natural experiment created by differences between dialysis centers in their anemia management practice.11,12 Our approach was motivated by (1) evidence suggesting that there is great variation among dialysis centers in the protocols used to make ESA and iron treatment decisions and (2) the assumption that patients are assigned to centers in a way that effectively randomizes them to different anemia management protocols. This approach is conceptually similar to most of the major trials of ESAs, where patients are not randomized to receive a specific dose of an ESA but are randomized to different treatment protocols dictating how ESAs and iron doses are to be titrated to achieve a desired hematocrit level.
Because data are not available on the anemia management protocols of individual dialysis centers, we were unable to directly estimate the mortality risk associated with different protocols. Instead, we associated characteristics of the observed anemia management practice in each dialysis center with mortality risk among patients initiating hemodialysis at the center. Using recent data from Medicare's ESRD program, we estimated an annual anemia management profile for every US dialysis center. The profile consisted of a set of predicted ESA and iron treatments given to patients with varying degrees of anemia. We then correlated the predicted treatments with 1-year mortality risk among patients initiating hemodialysis at each center.
Our analysis was based on data from the United States Renal Data System (USRDS). The USRDS contains detailed data on all patients in Medicare's ESRD program, including information collected at dialysis initiation (reported on the Medical Evidence Form) describing demographics, primary cause of ESRD, clinical data (eg, body mass index), and certain laboratory measurements (eg, serum albumin and hematocrit levels). In addition, the USRDS contains all Medicare Parts A and B claims that include information on diagnoses and procedures recorded for all hospitalizations and outpatient visits. The USRDS also contains data on total monthly ESA doses, which must be submitted with the final hematocrit laboratory recorded during the month. Most ESA use was epoetin alfa; however, there was a small amount of darbepoetin alfa use in the latter years of our study. We converted darbepoetin alfa to units of epoetin alfa using the equimolar dose conversion ratio of 200:1 (1 μg of darbepoetin = 200 units of epoetin). We identified intravenous iron administration using Healthcare Common Procedure Coding System codes from the Medicare Parts A and B claims. The Brigham and Women's Hospital Institutional Review Board approved this research.
From the USRDS, we identified all patients who began receiving maintenance hemodialysis between January 1, 1999, and August 31, 2006, and had no history of cancer indicated on the Medical Evidence Form (CMS-2728). Our study cohort consisted of a 50% random sample of all incident patients in this population. The remaining 50% of patients were placed in a “training sample” that was used to construct the anemia management profile for each US dialysis center.
We estimated the anemia management profile using data on ESA doses and iron administrations given to the patients in the training sample. By doing so, we ensured that the patients used to characterize the dialysis center's anemia management practice would not also be used in our assessment of outcomes.
As required by Medicare, the final hematocrit value of the month must be submitted with each claim for ESA reimbursement. We paired each ESA administration with the hematocrit laboratory value from the previous month; ie, the hematocrit level that gave rise to the ESA dosing. If no hematocrit was identified prior to a given ESA administration, that monthly ESA claim was not included in the analysis. Because ESA exposure captured in the USRDS is almost entirely from outpatient treatment at dialysis centers, we did not include months in which a patient spent 5 or more days in the hospital. We also did not include months that occurred after a patient switched to peritoneal dialysis or received a transplant. We converted the total ESA administered during the index month into units per day by dividing the total units of ESA administered during the month by the days in the month minus the time the patient spent in hospital during the month. We allowed each patient to contribute multiple observations to the analysis data set but used only ESA dosing and iron administration data that occurred 6 months after the start of dialysis.
Dialysis Center Anemia Management Profiles. For each dialysis center, we estimated a center-level ESA dosing profile (a characterization of the center's typical use of ESAs) using a set of linear random-effects models. To adjust for basic aspects of dialysis center case mix, we included fixed effects for race (white, black, American Indian, or other, as reported to the Centers for Medicare & Medicaid Services by the dialysis centers), age, sex, dialysis center business status (profit vs nonprofit), and cause of ESRD in each model. Other clinical variables were adjusted for in a later stage of the analysis. To capture center-specific ESA dosing practices, we included a center-level random effect that we assumed to be normally distributed. A separate model was fit for each calendar year and hematocrit group (classified as <30%, 30%-32.9%, 33%-35.9%, or ≥36%). This allowed the center's ESA dosing profile to change each year. Using the fitted models, predicted ESA doses for each dialysis center during each calendar year were generated for each hematocrit range. For each dialysis center, this set of predicted values represented estimates of typical ESA doses given to patients across the hematocrit groups. The models were fit using PROC MIXED in SAS software, version 9.1 (SAS Institute Inc, Cary, North Carolina).
For each center, we estimated typical use of intravenous iron (proportion of patients treated) using a set of logistic random-effects models that included race, age, sex, cause of ESRD, and dialysis center profit status as fixed effects and a normally distributed center-level random effect. Separate models were likewise fit for each calendar year and hematocrit range. Using the fitted models, predicted probabilities of iron administration for each dialysis center during each calendar year were generated for each hematocrit group. For each dialysis center, this set of predicted values represented the estimated probability that a patient with a given hematocrit would receive supplemental intravenous iron. These models were fit by PROC GLIMMIX using a quasi-likelihood approach maximized with the Nelder-Mead simplex algorithm.13
Association of Dialysis Center Anemia Management Profile and 1-Year All-Cause Mortality Risk. In our study cohort, we evaluated the 1-year all-cause mortality risk associated with dialysis center ESA and intravenous iron management practices. The statistical analysis was based on a Cox proportional hazards regression model of 1-year mortality. Follow-up began 60 days after the start of dialysis. Censoring occurred at the end of 1 year of follow-up, administrative end of follow-up (August 31, 2007), loss to follow-up, change to peritoneal dialysis, or renal transplantation.
We categorized predicted center-level ESA and intravenous iron use into quintiles for each hematocrit range and created indicator variables for these categories that were entered into the model, with the lowest quintile of ESA and iron use for each hematocrit range taken to be the reference category. The Cox models were stratified across calendar year and patient age (broken into 5-year age groups). This explicitly creates risk sets composed of patients of the same age, who began dialysis the same year and were receiving dialysis for the same length of time. We also made multivariable adjustments for various comorbid conditions as reported on the Medical Evidence Form (eg, history of stroke, myocardial infarction, heart failure), cause of ESRD, geographic region indicator variables (to account for regional differences in outcomes and medical practice), zip code–level measures of socioeconomic status, and several dialysis center−level variables. We computed standard errors that appropriately accounted for the clustering of patients within facility.14
Secondary Analyses. Associations between center-level practices and mortality risk could be confounded by patient characteristics or other aspects of care correlated with the center's anemia management practice. To assess this possibility, we compared our results that were adjusted only for the stratification variables with those from the full multivariable-adjusted model. To test the validity of the anemia management profile, we examined how well it predicted the actual management of anemia in patients from the outcomes sample as well as month-to-month changes in hematocrit. The details of this analysis are provided in the eAppendix.
To further assess the robustness of our findings, we conducted the following secondary analyses: (1) were laxed the definition of censoring, considering just administrative end of follow-up and loss to follow-up as censoring events; (2) we restricted the analysis to non–hospital-based facilities; (3) we included adjustment for baseline serum albumin, body mass index, and estimated glomerular filtration rate for the subsample of patients for whom these variables were available; (4) we restricted the analysis to larger facilities (those with >200 billed claims for ESAs during the year); and (5) we conducted analyses within different periods (1999-2002 and 2003-2006) to look for evidence of an era effect.
We analyzed data on approximately 4500 dialysis units in the United States. For each unit, we created a unique anemia management profile for each calendar year. The Figure, A, shows the median predicted center-level ESA dose for 33 541 annual anemia management profiles. The plot reveals a secular trend in ESA dosing, with centers using more ESA in the later years of the study, particularly among patients with lower hematocrit. The most variability in center-level dosing occurs in patients with hematocrit less than 30%. Many centers are treating these patients with less than 3500 units/d, whereas others are using more than 6000 units/d of ESAs. For patients with hematocrit of 36% or higher, predicted center-level ESA doses were considerably lower (1000-2000 units/d). The Figure, B, depicts the distribution of center-level intravenous iron use as a monthly rate by calendar year. The graph reveals a modest secular trend toward greater use of intravenous iron in later years: centers in 2005-2006 used iron in about 62% to 68% of patient-months (depending on hematocrit level), whereas centers in 1999-2000 used iron in 52% to 58% of patient-months (depending on hematocrit level).
The characteristics of the 269 717 patients in the outcomes sample are presented in Table 1. Patients had a mean age of 63 years, with 129 055 (48%) having diabetes and 68 996 (26%) having hypertension as the cause of ESRD. Comorbidities were common: 90 005 (33%) had a history of heart failure, 65 694 (24%) had a history of ischemic heart disease, and 25 719 (10%) had a history of stroke or transient ischemic attack. The patients studied were mostly treated at for-profit dialysis centers (78%) and at centers that were not connected to a hospital (90%).
During follow-up, 60 993 patients (22.6%) died, 5662 (2.1%) were censored by a switch to peritoneal dialysis, and 7478 (2.8%) were censored by transplantation. We found that mortality rates were highest in months following hematocrit levels less than 30% (mortality, 2.1%) and then decreased monotonically: for hematocrit levels of 30% to 32.9%, mortality was 1.3%; for 33% to 35.9%, it was 0.9%; and for 36% or higher, it was 0.7%.
Table 2 summarizes the multivariable-adjusted associations between anemia management practices and mortality. Centers that used larger doses of ESAs in patients with hematocrit less than 30% achieved lower mortality rates (highest vs lowest quintile of predicted dose: hazard ratio [HR], 0.94; 95% confidence interval [CI], 0.90-0.97). We observed no association between mortality and predicted center ESA dose in patients with a hematocrit between 30% and 32.9%. However, mortality rates were increased in centers that used larger ESA doses in patients with hematocrit between 33% and 35.9% (highest vs lowest quintile of predicted dose: HR, 1.07; 95% CI, 1.03-1.12) and in those with hematocrit of 36% or higher (highest vs lowest quintile of predicted dose: HR, 1.11; 95% CI, 1.07-1.15). We observed decreased mortality in centers that used iron more frequently in patients with hematocrit less than 30% (highest vs lowest quintile of monthly rate of iron administration: HR, 0.97; 95% CI, 0.94-0.99) and in patients with hematocrit between 30% and 32.9% (highest vs lowest quintile of monthly rate of iron administration: HR, 0.95; 95% CI, 0.91-0.98). We also observed increasing mortality rates in centers that used iron more frequently in patients with hematocrit levels of 36% or higher (highest vs lowest quintile of monthly rate of iron administration: HR, 1.07; 95% CI, 1.02-1.13).
In secondary analyses, we found that the estimated effects were substantively unchanged when we removed all covariates from the Cox model (eAppendix and eTable 1), when we added baseline laboratory and clinical variables to the Cox model (eTable 2), when censoring was redefined (eTable 3), when we restricted the analysis to non–hospital-based facilities (eTable 4), and when we restricted the analysis to large dialysis centers (eTable 5). When we stratified by time period, we observed similar results during the years 1999-2002 (eTable 6), but in 2003-2006 the apparent benefit of iron in patients with low hematocrit was slightly attenuated (eTable 7). We also found that the center-level anemia management profile was strongly related to ESA dosing and iron use decisions and month-to-month changes in hematocrit. In this analysis, we observed increased center-level use of ESAs associated with increased hematocrit across all hematocrit categories and increased use of iron associated with modestly increased hematocrit in patients with hematocrit of 30% or higher (eTable 8). These results support the validity of the analysis.
In a large cohort of incident US hemodialysis patients, we assessed the 1-year mortality risk associated with different dialysis center–level patterns of ESA and intravenous iron use. After adjustment for a range of potential confounding factors, we found that certain patterns of ESA and iron use by dialysis centers were associated with altered mortality risk among incident patients at those centers.
We found elevated risk at centers that use larger (vs smaller) doses of ESAs in patients with hematocrit levels of 33% or higher and at centers that use more iron (vs less) in patients with hematocrit levels of 36% or higher. Several major randomized controlled trials of ESAs have found increased risk associated not directly with dose but with protocols that target normal or near-normal hematocrit levels. Besarab et al2 studied ESRD patients with congestive heart failure or ischemic heart disease and found that those treated to achieve a “normalized” hematocrit of 42% had an increased risk of death or myocardial infarction relative to patients treated to achieve a hematocrit of 30% (relative risk, 1.3; 95% CI, 0.9-1.9). Singh et al3 studied 1432 patients with chronic kidney disease and found increased risk of a composite end point of death and cardiovascular events in patients treated to achieve a hemoglobin level of 13.5 g/dL compared with 11.3 g/dL (HR, 1.34; 95% CI, 1.03-1.74). Pfeffer et al4 studied 4038 patients with diabetes, anemia, and chronic kidney disease and found that those treated to achieve a target hemoglobin level of 13.0 g/dL had an increased risk of stroke relative to those treated with placebo with rescue ESA treatment at a hemoglobin level of 9.0 g/dL or lower (HR, 1.92; 95% CI, 1.38-2.68). Under the reasonable assumption that greater center use of ESAs and iron in patients with mild anemia is reflective of higher hematocrit targets in the center, our study is consistent with trials and suggests that high hematocrit targets are problematic in the wider population of dialysis patients.
We found decreased mortality risk in centers that used larger doses of ESAs in patients with hematocrit levels less than 30% and in centers that administered iron more frequently in patients with hematocrit less than 33%. Experimental research has found that treatment with ESAs in appropriately selected patients reduces the need for transfusions1 and improves cardiac ischemia15,16 and left ventricular hypertrophy.15,17 Additionally, epidemiological research has linked severity of anemia to increased rates of hospitalizations and mortality.18-21 Although no randomized studies of ESAs or iron have reported a mortality benefit associated with ESA or iron use, none of the trials were designed to provide information about the effects of aggressive compared with conservative treatment of severe anemia. The existing trials were designed only to compare the effect of treating patients to achieve normal or near-normal vs relatively lower hematocrit.
The observational studies of the effect of ESA and iron dosing on mortality risk are conflicting. Earlier studies have linked higher ESA and iron doses to mortality,22-25 but several recent observational studies, including some that attempted to address limitations of previous work, found no evidence of increased mortality risk related to ESA or iron dosing.10,26-28 However, only 1 of these studies examined dose effects within strata of hematocrit, and it observed only patients with low hematocrit, in whom it found a near-null effect associated with larger ESA doses.29
The degree to which the apparent risks of ESAs and iron are mediated through their effect on hematocrit or are a result of dose-related toxicities is unclear. Erythropoiesis-stimulating agents are known to increase platelet count and reactivity under certain circumstances,30,31 which may increase risk of thrombosis. Erythropoiesis-stimulating agents can also increase arterial pressure,32,33 possibly affecting cardiovascular risk. Iron may also have important toxic effects.34 Excessive use of iron could theoretically increase the risk of sepsis and infection-related mortality,35,36 worsen atherogenesis,37 and increase the risk of cardiovascular disease events.34,38 Iron dextran has also been linked to hypersensitivity reactions including anaphylaxis, although fatalities attributable to these events are rare.39,40 Although there are plausible biological mechanisms, it seems unlikely that altered mortality risk would be solely attributable to dose-related effects because the largest ESA doses were given to patients in the lowest hematocrit category and were associated with lower mortality rates. However, we also found that mortality rates were inversely associated with hematocrit, suggesting that high hematocrit alone is not harmful. Whatever the mechanism, our study suggests that greater use of ESAs and iron in patients with higher hematocrit is problematic.
The estimates reported herein are similar to intention-to-treat estimates; they describe the effect of a treatment practice on mortality rates in all patients at the center. The expected mortality rate in a dialysis center depends on the average risk among the patients beginning dialysis there and the product of the HRs that represent the center's anemia management practice. To assess public health relevance, the composite relative hazard can be converted into an approximate change in absolute risk. For example, assuming a first-year mortality rate in the hemodialysis population of about 23%, a relative decrease in hazard of 5% corresponds to a decrease of approximately 1 death per 100 patients during the first year of treatment. Similarly, a 10% relative increase in hazard corresponds to an increase of approximately 2 deaths per 100 patients.
The analytic approach used in this study was motivated by evidence suggesting that standard epidemiologic methods would be subject to strong confounding by indication and information bias.10 By adopting a center-level analytic approach, we attempted to estimate treatment effects using differences in anemia management practices between dialysis centers as a natural experiment. Other studies of patients with ESRD have used facility-level measures of treatment patterns as a proxy for actual exposure when unmeasured confounding may be intractable.29,41,42 This approach has also been used in studies of inpatient procedures43-45 and prescription medications.46,47
Despite their connection to a natural experiment, these analyses are still observational in nature and can result in incorrect inferences.48 For example, the center-level anemia management profile could be confounded by differences in patient case mix. This could cause the anemia management profile to be more reflective of the ESA requirements of the center's case mix rather than the center's anemia management practice. For example, patients with greater ESA requirements because of higher levels of inflammation may be at greater risk of mortality.10,23,49,50 If higher center-level use of ESAs reflects the prevalence of treatment-resistant anemia in the center's case mix, center-level ESA use could be associated with mortality even if ESAs have no effect on mortality. The anemia management profile could also be spuriously associated with mortality if anemia management practices are correlated with other aspects of care that might affect mortality, such as intravenous vitamin D use or dialysis adequacy. We explored the possibility of confounding by conducting both an unadjusted analysis and an analysis that included a richer set of clinical and demographic variables than were included in the main analysis. We found nearly identical results across all analyses. We also observed similar associations within subgroups defined by center characteristics. This suggests that our anemia management profiles are not confounded by patient case mix or other aspects of medical practice.
Center-level analyses may be consistent with different hypotheses about patient-level treatment effects. For example, we observed higher mortality rates in centers that used larger doses of ESAs in patients with hematocrit levels in the range of 33% to 35.9%. Centers using more ESAs in these patients could either be attempting to keep patients who are poorly responsive in the target range or may be treating all patients to a hematocrit higher than 36%. Our analysis is unable to distinguish which of these 2 practices might be potentially harmful.
Our study was limited in that we studied the incident dialysis patient population who make up only 20% of the overall hemodialysis population but have the highest mortality rates.51 Therefore, our results may not completely generalize to prevalent patients. Our study of iron use was subject to 2 additional limitations. First, during the period of our study, the data could be used reliably only to identify whether iron had been given, not the dose of iron administered. Second, measures of iron availability, the primary indications for iron therapy, are not reported to the Centers for Medicare & Medicaid Services. Therefore, high or low center-level iron use may not reflect overuse or underuse, respectively. Additional studies using more reliable iron dosing information and measures of iron availability are warranted to further evaluate any potential risks or benefits related to iron dosing.
In conclusion, we found evidence of decreased mortality risk associated with greater use of ESAs and more frequent use of iron at lower hematocrit levels where mortality is the highest. While lower overall mortality risk occurs at higher hematocrit levels, elevated mortality risk was associated with greater use of ESAs and iron in these patients. Further observational and experimental studies are needed to help identify optimal treatment algorithms for both ESAs and iron that maximize clinical benefit while minimizing adverse outcomes.
Corresponding Author: Wolfgang C. Winkelmayer, MD, ScD, Division of Nephrology, Stanford University School of Medicine, 780 Welch Rd, Ste 106, Palo Alto, CA 94304 (email@example.com).
Author Contributions: Dr Brookhart had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Brookhart, Schneeweiss, Avorn, Bradbury, Winkelmayer.
Acquisition of data: Brookhart, Winkelmayer.
Analysis and interpretation of data: Brookhart, Schneeweiss, Avorn, Bradbury, Liu, Winkelmayer.
Drafting of the manuscript: Brookhart, Winkelmayer.
Critical revision of the manuscript for important intellectual content: Brookhart, Schneeweiss, Avorn, Bradbury, Liu, Winkelmayer.
Statistical analysis: Brookhart, Schneeweiss, Bradbury, Liu, Winkelmayer.
Obtained funding: Brookhart, Bradbury.
Study supervision: Brookhart, Schneeweiss, Avorn, Winkelmayer.
Financial Disclosures: Dr Brookhart reports having received investigator-initiated grant support from Amgen. He has participated, without receiving honoraria, in advisory boards of Amgen. Dr Schneeweiss reports that he is principal investigator of the Brigham and Women's Hospital DEcIDE Center on Comparative Effectiveness Research, funded by the Agency for Healthcare Research and Quality, and of the Harvard-Brigham Drug Safety and Risk Management Research Contract, funded by the US Food and Drug Administration. Dr Schneeweiss is a paid member of scientific advisory boards for HealthCore and ii4sm and has received consulting fees from WHISCON, RTI Health Solutions, the Lewin Group, and HealthCore. Dr Bradbury reports that he is an employee of Amgen Inc. Dr Winkelmayer reports receiving investigator-initiated grants from Amgen and GlaxoSmithKline. He has participated in advisory boards for AMAG Pharmaceuticals, Amgen, Roche, Genzyme, and Fresenius. No other disclosures were reported.
Funding/Support: This work was supported by an investigator-initiated contract from Amgen to Brigham and Women's Hospital that placed no restrictions on publications. Dr Brookhart is supported by a career development award from the National Institute on Aging (grant AG-027400). Dr Winkelmayer has received support including a Scientist Development Grant from the American Heart Association, a Norman S. Coplon Extramural Research Program Award from Satellite Healthcare Inc, support from the National Institutes of Health for unrelated projects.
Role of the Sponsor: The funding organization, Amgen, had no formal role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation of the manuscript. The contract gave the company 60 days to review the manuscript but placed no restrictions on publication. Dr Bradbury, as an employee of Amgen, provided feedback on the design, analysis, presentation, and interpretation of results, but final decisions about all aspects of the research project were made by the research team at Brigham and Women's Hospital. The final content of the manuscript was not subject to the approval of the sponsor.
Disclaimer: Data reported herein were supplied by the USRDS. Interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the US government.
This article was corrected online for typographical errors on 4/20/2010.