Recombinant human erythropoietin (EPO) is an effective

Transcription

1 Septicemia in Patients with ESRD Is Associated with Decreased Hematocrit and Increased Use of Erythropoietin Allen R. Nissenson,* Michelle L. Dylan, Robert I. Griffiths, Hsing-Ting Yu, and Robert W. Dubois *David Geffen School of Medicine, University of California Los Angeles, Los Angeles, California; Cerner Health Insights, Beverly Hills, California; and Health Economics Consulting, Craftsbury, Vermont Septicemia, a common complication in chronic dialysis patients, may be an important factor in erythropoietin (EPO) hyporesponsiveness, because it is a form of inflammation. The quantitative impact of septicemia on EPO requirements has not been studied. The purpose of this study was to analyze patterns of EPO use and levels of anemia among patients who had ESRD and were hospitalized with septicemia. Using United States Renal Data System data, septicemia admissions were identified in patients with first ESRD service from 1996 to Mean EPO dosage and hematocrit (Hct) level were analyzed from 2 mo before until 3 mo after admission and compared with patients who were hospitalized with acute myocardial infarction (AMI) and patients with no hospitalizations. A total of 4640 hospitalized patients were included in the analysis: 3975 for septicemia and 665 for AMI. In both groups, mean Hct declined significantly in the month of admission and increased in the second month after admission. At all time points, both groups had significantly lower Hct levels compared with the nonhospitalized group. Mean EPO dosage increased, most rapidly in the month after admission. EPO use was highest in the septicemia group. Hospitalization with septicemia is associated with worsening anemia and increasing EPO use. This also was observed for patients who were hospitalized with AMI, suggesting that acute intercurrent illness plays an important role in EPO hyporesponsiveness. Strategies to prevent septicemia are important not only to decrease clinical morbidity but also to conserve EPO usage and thus contain the costs of care for this complex patient population. Clin J Am Soc Nephrol 1: , doi: /CJN Recombinant human erythropoietin (EPO) is an effective treatment for anemia that is associated with ESRD, and many patients can achieve adequate hemoglobin (Hb) levels with its use (1 4). However, a significant number of patients with ESRD fail to respond adequately to this therapy. In some patients, this incomplete response is due to iron deficiency (5,6). Despite adequate iron supplementation, certain dialysis patients still may require unexpectedly high dosages of EPO (so-called EPO hyporesponsiveness ) (4,7 9). Several studies have shown that inflammation causes EPO hyporesponsiveness (5,10 12). In these studies, hyporesponsiveness was observed in patients with chronic heart failure, various inflammatory diseases, or vascular access infections. It is logical to postulate that septicemia, a common inflammatory event in dialysis patients (13 15), also would cause EPO hyporesponsiveness. Until now, an empiric assessment of this relationship has not been performed. The purpose of this study was to analyze patterns of EPO use and level of anemia among patients who had ESRD and were hospitalized with septicemia. Received September 27, Accepted January 25, Published online ahead of print. Publication date available at Address correspondence to: Dr. Allen R. Nissenson, David Geffen School of Medicine, University of California Los Angeles, 200 UCLA Medical Plaza, Suite , Los Angeles, CA Phone: ; Fax: ; anissenson@mednet.ucla.edu Materials and Methods Study Design and Patient Selection This retrospective cohort study used the United States Renal Data System (USRDS). Hospitalizations were identified on the basis of Medicare hospital claims using the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) and/or Diagnosis Related (DRG) (16) codes for discharge diagnoses. This study compared patterns of EPO use and hematocrit (Hct) among patients who were hospitalized with septicemia with two comparison groups: Patients who were admitted with another acute illness (acute myocardial infarction [AMI]) and patients who were not recently hospitalized for any acute clinical event. Inpatient septicemia admissions (ICD-9-CM 038.xx) were identified with a first ESRD service date between January 1, 1996, and September 30, Admissions were included when (1) the dialysis modality at admission was recorded as hemodialysis, (2) Medicare was the primary payer for the admission, (3) the admission occurred after the date of first ESRD service, and (4) the dialysis monthly billing period ended before or during the hospitalization (to ensure that the outpatient EPO treatment for the month in which the patient was hospitalized occurred immediately before the admission). We also studied patients who were hospitalized with AMI and patients with no hospitalization to understand whether patterns of EPO dosage and level of anemia were unique to the septicemia group. Inpatient AMI admissions (DRG codes 121 to 123), excluding bypass surgery (ICD-9-CM 36.1, to 36.17, and 36.19), were identified during the same period as described above. Because patients who had bypass surgery could have experienced decreased Hct as a result of blood loss, we excluded them from the analyses. After identifying the septicemia and AMI groups, we determined mean time from start of dialysis to hospitalization and used it to define the time period for Copyright 2006 by the American Society of Nephrology ISSN: /

2 506 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 1: , 2006 sampling the nonhospitalized group. This resulted in identification of patients without a hospitalization during the 15- to 21-mo period after the date of first ESRD service. Data Basic demographic data consisted of patient age at the time of hospitalization, race, and gender. Average length of stay (LOS) was calculated for the septicemia and AMI admissions. Vascular access type was determined by identification of the Healthcare Common Procedure Coding System (17) code for the vascular access procedure that occurred closest in time before the septicemia or AMI admission or closest in time to the middle of the observation period in the nonhospitalized group (18 mo after first ESRD service). Clinical history, as recorded at the time of the first ESRD service, was examined using the Medical Evidence Form that was completed by the patient s physician. Conditions that were recorded were alcohol dependence, cancer, cardiac arrest, congestive heart failure, cerebrovascular disease, diabetes (insulin dependent), diabetes as primary or contributing cause of ESRD, drug dependence, cardiac dysrhythmia, hypertension, ischemic heart disease, myocardial infarction, inability to ambulate, inability to transfer, EPO use (predialysis), chronic obstructive pulmonary disease, and peripheral vascular disease. Cause of renal failure also was examined using the Medical Evidence Form. Charlson comorbidity index (CCI) scores, a method of classifying prognostic comorbidity, were calculated on the basis of diagnosis codes and procedures that were reported during 1 yr before index hospitalization. The Dartmouth-Manitoba version (18,19), with modifications to reflect more recently developed codes, was used to access the severity of comorbidity. Mean EPO dosage and level of anemia (defined according to Hct) were recorded from 2 mo before hospitalization to 3 mo after admission to understand the correlation of EPO dosage and level of anemia with septicemia hospitalization. EPO dosage and Hct level were determined on the basis of the dialysis center report that was submitted to Medicare with the claim. To ensure that the outpatient EPO treatment for the month in which the patient was hospitalized (t 0 ) occurred immediately before the admission, we included only the admissions that overlapped with dialysis monthly billing period. Therefore, hospitalization (t 0 ) contained two distinct periods: (1) Outpatient treatment before the admission and (2) the admission itself. As a result, exposure time to EPO varied for each patient depending on the timing of admission in the month. For example, for a person who was admitted at the end of the month, the exposure time was longer compared with a person who was admitted at the beginning of the month. Mean EPO dosage was calculated using the total dosage given during each billing period (on monthly basis) divided by the reported number of injections. Hct levels were determined from the monthly dialysis center report that was submitted to Medicare. Statistical Analyses Univariate and bivariate statistics were used to examine differences in patient demographics among septicemia, AMI, and nonhospitalized groups. Continuous variables were summarized by the mean and SD, whereas discrete data or counts were summarized using percentages. Differences in EPO dosage and Hct level between time points were calculated for the septicemia, AMI, and nonhospitalized groups. Patients with septicemia and AMI were divided further into two subgroups using their median LOS: (1) Shorter LOS (septicemia LOS 9 d; AMI LOS 6 d) and (2) longer LOS (septicemia LOS 9 d; AMI LOS 6 d). It was hypothesized that LOS might function as a proxy for severity of illness during the admission and that greater severity might have a greater impact on level of anemia and EPO usage. Differences in EPO dosage and Hct level between patients with shorter versus longer hospital stays were examined. T tests (20) were conducted to determine differences in EPO dosage and Hct levels between groups and time points. Changes in Hct levels and EPO dosages over time were tested for significance using ANOVA (21) with repeated measures. SAS statistical software 8.2 (SAS Institute, Cary, NC) was used to conduct the analyses. Results Characteristics of the Sample A total of 4640 hospitalized patients (3975 septicemia and 665 AMI) and 103,397 nonhospitalized patients were included in the analysis. Characteristics of this study population are shown in Table 1. The three groups were similar in terms of demographics, although the AMI group was significantly older. Catheter was the most common type of vascular access for the three groups. The septicemia group had a significantly longer LOS than the AMI group (12.2 versus 7.5; P ). The CCI score was found to be higher in the AMI group (7.0). However, the CCI includes myocardial infarction (Charlson weight 1), which skews the comparison. Diabetes (type 2) was the leading cause of renal failure and was followed closely by hypertension. In terms of clinical conditions that were recorded at the time of first ESRD service, the septicemia and AMI groups seemed similar. The nonhospitalized group was less likely to have congestive heart failure, diabetes (insulin dependent), chronic obstructive pulmonary disease, and peripheral vascular disease compared with the septicemia group. Level of Anemia and Patterns of EPO Usage Level of Anemia. Mean Hct decreased in the month of hospitalization and increased in the second month after hospitalization in both the AMI and septicemia groups (Figure 1). Changes in mean Hct over time within the groups were significant (P ), with the greatest impact in both groups occurring in the month after hospitalization. Mean Hct values were significantly lower in the septicemia group than in the AMI group until 2 mo after hospitalization (P 0.05). In the nonhospitalized group, Hct was relatively stable over time and was significantly higher than in the septicemia and AMI groups during the entire observation period (P 0.05). The pattern of change in Hct across time points did not differ between the two hospitalized groups. Mean Hct was significantly lower in the longer LOS septicemia group than in the other groups at each time point (P 0.05; Figure 2). Differences between longer and shorter LOS in Hct level were significant in the septicemia (P 0.05) but not in the AMI group. It seems that LOS has greater impact on Hct in the septicemia group than in the AMI group. EPO Usage. Mean EPO dosage increased over time for the septicemia and AMI groups, with the largest increase occurring 1 mo after hospitalization (Figure 3). EPO dosage slowly decreased over time in the nonhospitalized group. Changes in EPO use across time points within the septicemia and AMI groups were significant (P ), but they were NS within

3 Clin J Am Soc Nephrol 1: , 2006 Patterns of EPO Use and Hematocrit Associated with Septicemia 507 Table 1. Demographics of study populations a Characteristic Septicemia AMI P Nonhospitalized n % n % n % P Compared with Septicemia P Compared with AMI Race Native American , Asian , black 1, , white 2, , unknown other , total 3, , Gender female 1, , male 2, , total 3, , Vascular access b fistula , catheter 2, , graft , total 2, , Mean SD Mean SD Mean SD Age (yr) LOS (d) Charlson Comorbidity Index score Clinical conditions c alcohol dependence , cancer , cardiac arrest , congestive heart failure 1, , cerebrovascular disease , diabetes, insulin dependent , diabetes, primary or contributing 1, , drug dependence cardiac dysrhythmia , hypertension 3, , ischemic heart disease 1, , myocardial infarction , inability to ambulate , inability to transfer , EPO use predialysis 1, , chronic obstructive pulmonary disease , peripheral vascular disease , Cause of renal failure b diabetes, type 2 adult onset 1, , renal disease due to hypertension , diabetes, type 1 juvenile onset , glomerulonephritis , cause uncertain , other reason d , a AMI, acute myocardial infarction; EPO, erythropoietin; LOS, length of stay b Extensive missing data. c Recorded on medical evidence form at start of dialysis. d Includes all causes of renal failure with a frequency of 3%. the nonhospitalized group. EPO use was significantly higher in the septicemia group than in the AMI group from the time of hospitalization through 2 mo after hospitalization (P 0.05). Differences between the septicemia and AMI groups in EPO dosage changes across time points were significant only in the month before hospitalization, with a mean increase of 164 units in the septicemia group and a mean decrease of 155 units in the AMI group (P ).

4 508 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 1: , 2006 Figure 1. Mean hematocrit (Hct) over time by study population. 1, The difference between the septicemia and acute myocardial infarction (AMI) groups is significant (P 0.05); 2, the difference between the septicemia and nonhospitalized groups is significant (P 0.05); 3, the difference between the AMI and nonhospitalized groups is significant (P 0.05). Figure 3. Mean erythropoietin (EPO) dosage over time by study population. 1, The difference between the septicemia and AMI groups is significant (P 0.05); 2, the difference between the septicemia and nonhospitalized groups is significant (P 0.05); 3, the difference between the AMI and nonhospitalized groups is significant (P 0.05). Long versus short LOS had a greater impact on EPO use in the septicemia group than in the AMI group (P 0.05 after hospitalization; Figure 4). Mean EPO dosage increases were similar across shorter and longer LOS groups for those who were hospitalized with septicemia, but the patients with longer LOS consistently required higher dosages over time, particularly around the hospitalization. For both shorter and longer LOS AMI groups, EPO dosage was lower than both the shorter and longer LOS septicemia groups (except at the beginning and the end of the observation period). The difference in change in EPO dosage, measured by the between time periods, was significant only for the septicemia group during the time period from hospitalization to 1 mo after hospitalization, with the shorter LOS septicemia group having a of 382 units and the longer LOS septicemia group having a of 977 units (P ). Discussion This study confirmed our hypothesis that septicemia is associated with EPO hyporesponsiveness. We observed that in patients with ESRD and with septicemia, Hct levels fell and EPO dosage increased in the time period surrounding hospitalization. In general, patterns of EPO use and Hct were found Figure 2. Mean Hct over time by length of stay (LOS) in the septicemia and AMI groups. 1, The difference between the septicemia shorter and longer LOS groups is significant (P 0.05). to be similar for patients with septicemia and with AMI, and both groups differed significantly from the nonhospitalized control subjects. Within-group changes of EPO use differed significantly between septicemia and AMI groups only in the month before hospitalization. Although the difference between time points in Hct levels and EPO dosages were similar for the septicemia and AMI groups, we observed that absolute values of Hct levels were significantly lower for the septicemia group from 2 mo before hospitalization (P 0.05), and EPO dosages were significantly higher from hospitalization through 2 mo after hospitalization. These findings suggest that hospitalization for an acute illness affects the level of anemia and EPO usage and that an infection may have a greater impact on both. To explore this relationship further and test for a dosage response, we hypothesized that if EPO resistance were related to an acute illness, then a more severe illness might have greater impact than a less severe one. Using median LOS as a proxy for disease severity, we found that septicemia patients with longer LOS had substantially lower Hct levels and higher EPO dosages compared with the shorter LOS group. The differences in Hct levels and EPO dosages between the longer and shorter LOS AMI groups were less pronounced. Perhaps LOS does not reflect illness severity as accurately in this latter population. Our findings add to previously published work in this area. Several studies suggested that either chronic or acute inflammation leads to inadequate EPO response in hemodialysis patients, which was shown by measuring markers of inflammation (e.g., levels of serum C-reactive protein) (5,10 12). An important finding of our study was the strong association between EPO hyporesponsiveness and the specific cause of inflammation (i.e., septicemia). One retrospective chart audit of 65 patients concluded that hemodialysis patients experience a significant and prolonged decrease in Hb levels and an increase in EPO requirements after hospitalization (22). However, they did not find that inflammation had a significant impact on either Hb levels or EPO dosage. One possible confounding factor may be that more patients in the noninflammatory group underwent surgery during hospitalization, suggesting that the

5 Clin J Am Soc Nephrol 1: , 2006 Patterns of EPO Use and Hematocrit Associated with Septicemia 509 Figure 4. Mean EPO over time by LOS in septicemia and AMI groups. 1, The difference between the septicemia shorter and longer LOS groups is significant (P 0.05). noninflammatory group might have been sicker than the inflammatory group. To our knowledge, no studies have examined the impact of disease severity on EPO hyporesponsiveness. Comparisons of characteristics among the study populations were conducted to understand the composition of the three study groups. We found that the septicemia, AMI, and nonhospitalized groups were similar in terms of race, gender, vascular access, age, LOS, CCI, and cause of renal failure. A previous study showed that female gender was associated with EPO hyporesponsiveness (23). However, the results applied only for women with menses (aged 25 to 44). Our study population was aged 60 yr on average, and no significant differences between men and women were observed in EPO use and level of anemia. Some differences in EPO use and level of anemia were found during the observation period when white individuals were compared with nonwhite individuals. The magnitude of each change showed statistical significance but did not reach clinical importance. The septicemia and AMI groups were similar in terms of clinical profiles. The AMI group was more likely to have ischemic heart disease and myocardial infarction than the septicemia group, as expected. Overall, these hospitalized groups had similar demographic and clinical profiles. This makes the AMI group a good comparator. Similar patterns of EPO use and level of anemia were found for the two hospitalized groups. The nonhospitalized group seemed similar to the AMI group in terms of clinical profile but was less severely ill than the septicemia group. Distinct patterns of EPO use and level of anemia were found between the septicemia group and the nonhospitalized control group, which confirmed our hypothesis that septicemia is associated with EPO hyporesponsiveness. As with all studies, there are limitations to our findings and to our approach. First, the Hct and EPO dosage data in the USRDS are available only at monthly intervals as required by Medicare for reimbursement purposes. Although these data likely are accurate (because they are critical to reimbursement), they lack the granularity that might result from more frequent points of observation. Nevertheless, we do not believe that more frequent points of observation would change the overall conclusions of our study. Second, the USRDS does not contain Hct values that were measured during the hospitalization or dosages of EPO that might be provided. Given that the period of observation spans several months, we also do not believe that data from the hospital itself would fundamentally change the findings. Moreover, our restriction to billing periods that ended before or during the hospitalization seems to alleviate any potential problems. Despite these possible limitations, the results should have generalizability because our analysis was conducted among a very large sample of hemodialysis patients who had ESRD and were hospitalized for septicemia. Our approach may underestimate the differences in EPO dosage before and after septicemia hospitalization, as the average EPO usage was not adjusted for patient weight or for whether the patient received transfusions. Patients with ESRD tend to lose weight as a result of poor intake of protein and calories during the course of chronic kidney disease. In addition, if patients received transfusions, then changes in EPO dosage would underestimate the impact of inflammation on EPO hyporesponsiveness. Conclusion This study provides the first quantitative assessment of EPO usage and level of anemia in patients who had ESRD and were hospitalized for septicemia. Our findings suggest that intercurrent events in patients with ESRD have a clinically important impact on the patients ability to produce red blood cells. These clinical consequences also result in the additional economic cost of increasing the EPO dosage, which seems to persist for a substantial amount of time. Strategies to prevent septicemia (or AMI) are important not only to decrease clinical morbidity but also to optimize treatment with EPO and thus constrain the costs of care. Acknowledgments Research funding for the study was provided by Nabi Biopharmaceuticals to design the study, conduct analysis, interpret the data, and prepare the manuscript. A.R.N. 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. These data were accepted for poster presentation at the American Society of Nephrology Meeting in Philadelphia, PA, November 10, References 1. The US Recombinant Human Erythropoietin Predialysis Study : Double-blind, placebo-controlled study of the therapeutic use of recombinant human erythropoietin for anemia associated with chronic renal failure in predialysis patients. Am J Kidney Dis 18: 50 59, IV. NKF-K/DOQI clinical practice guidelines for anemia of chronic kidney disease: Update Am J Kidney Dis 37: S182 S238, Nissenson AR, Strobos J: Iron deficiency in patients with renal failure. Kidney Int Suppl 69: S18 S21, Tong EM, Nissenson AR: Erythropoietin and anemia. Semin Nephrol 21: , Drueke T: Hyporesponsiveness to recombinant human erythropoietin. Nephrol Dial Transplant 16[Suppl 7]: 25 28, Kalantar-Zadeh K, Hoffken B, Wunsch H, Fink H, Kleiner M, Luft FC: Diagnosis of iron deficiency anemia in renal

6 510 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 1: , 2006 failure patients during the post-erythropoietin era. Am J Kidney Dis 26: , Kalantar-Zadeh K, Don BR, Rodriguez RA, Humphreys MH: Serum ferritin is a marker of morbidity and mortality in hemodialysis patients. Am J Kidney Dis 37: , Thamer M, Richard CM, Klinkmann J, Ivanovich P, Lang G, Cotter DJ: Use of clinical guidelines for treatment of anemia among hemodialysis patients. Artif Organs 24: 91 94, Van de Vyver FL, Visser WJ, D Haese PC, De Broe ME: Iron overload and bone disease in chronic dialysis patients. Nephrol Dial Transplant 5: , Kalantar-Zadeh K, McAllister CJ, Lehn RS, Lee GH, Nissenson AR, Kopple JD: Effect of malnutrition-inflammation complex syndrome on EPO hyporesponsiveness in maintenance hemodialysis patients. Am J Kidney Dis 42: , Del Vecchio L, Pozzoni P, Andrulli S, Locatelli F: Inflammation and resistance to treatment with recombinant human erythropoietin. J Ren Nutr 15: , Stenvinkel P: The role of inflammation in the anaemia of end-stage renal disease. Nephrol Dial Transplant 16[Suppl 7]: 36 40, Abbott KC, Agodoa LY: Etiology of bacterial septicemia in chronic dialysis patients in the United States. Clin Nephrol 56: , Powe NR, Jaar B, Furth SL, Hermann J, Briggs W: Septicemia in dialysis patients: Incidence, risk factors, and prognosis. Kidney Int 55: , Foley RN, Guo H, Snyder JJ, Gilbertson DT, Collins AJ: Septicemia in the United States dialysis population, 1991 to J Am Soc Nephrol 15: , DRG Expert 2004: A Comprehensive Reference to the DRG Classification System, 20th Ed., Salt Lake City, Ingenix, Inc., HCPCS Level II Expert 2004, 15th Ed., Salt Lake City, Ingenix, Inc., Roos LL, Sharp SM, Cohen MM, Wajda A: Risk adjustment in claims-based research: The search for efficient approaches. J Clin Epidemiol 42: , Romano PS, Roos LL, Jollis JG: Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: Differing perspectives. J Clin Epidemiol 46: , Gossett WS: The probable error of a mean. Biometrika 6: 1 25, Fisher RA: Studies in crop variation. I. An examination of the yield of dressed grain from Broadbalk. J Agric Sci 11: , Yaqub MS, Leiser J, Molitoris BA: Erythropoietin requirements increase following hospitalization in end-stage renal disease patients. Am J Nephrol 21: , Di Iorio BR, Stellato D, De Santo NG, Cirillo M: Association of gender and age with erythropoietin resistance in hemodialysis patients: Role of menstrual status. Blood Purif 22: , 2004