Efficacy of Recombinant Human Erythropoietin in Critically Ill Patients:  A Randomized Controlled Trial FREE

Blood transfusion, an integral part of clinical practice for most of the last century, has been looked upon as relatively "risk free" and with obvious clinical benefit.¹ A dramatic change in thinking occurred in the early 1980s, when concerns about transfusion-related infections, particularly those caused by hepatitis C and the human immunodeficiency virus (HIV), prompted a reevaluation of the risks of allogeneic transfusion. Although advances in transfusion medicine have greatly decreased the risk of viral transmission during blood transfusion, other concerns now drive the debate over transfusion practice and have led to a reexamination of the approach to red blood cell (RBC) transfusion.

Critically ill patients typically receive multiple RBC transfusions.^2- 4 Recent data from both the United States and Western Europe demonstrate that between 35% and 50% of all patients admitted to ICUs today receive on average almost 5 RBC units during their ICU stay.^5- 6 However, the view of RBC transfusion as risk free is no longer tenable. In addition to well-described transfusion complications, recent studies have raised the issue of immunosuppression related to allogeneic blood transfusion,^1,7- 10 as well as concerns regarding the age of RBCs transfused.^11- 12 Adding to the controversy about risk-benefit ratio for RBC transfusion are recent data showing that an aggressive RBC transfusion strategy may decrease the likelihood of survival in selected subgroups of critically ill adults.¹³ Accordingly, limiting exposure to allogeneic RBC transfusions would be advantageous in critically ill patients.

Production of RBCs by the bone marrow is impaired in critically ill patients, and this phenomenon contributes to both the development and the persistence of anemia. Critically ill patients tend to be anemic early in their ICU course and hemoglobin levels fall during the ICU stay.^5- 6 The anemia associated with critical illness is probably fundamentally similar to the anemia of chronic inflammatory disease.¹⁴ A major feature of the anemia of critical illness is a failure of circulating erythropoietin concentrations to increase appropriately in response to physiologic stimuli.^15- 19 Thus, we hypothesized that treatment with pharmacological doses of recombinant human erythropoietin (rHuEPO) might decrease exposure to allogeneic blood and raise the hemoglobin level in critically ill patients.

Recently, a small, randomized, placebo-controlled trial of critically ill patients who received a combination of daily administration of rHuEPO for 5 days followed by every-other-day administration reported an almost 50% reduction in the number of RBC transfusions.²⁰ Despite receiving fewer RBC transfusions, patients receiving rHuEPO had a significantly greater increase in hematocrit. The efficacy of rHuEPO in this trial has raised the question whether rHuEPO would also be effective at reducing transfusions in a larger, more diverse critically ill population. Furthermore, in other clinical settings a weekly rHuEPO dose of 40 000 units has been shown to be as effective as more frequent dosing regimens.^21- 24 Our study was designed to assess the efficacy of a weekly dosing schedule of rHuEPO in reducing the exposure to allogeneic RBCs in a large, diverse group of critically ill patients.

This study was a prospective, randomized, double-blind, placebo-controlled, multicenter trial conducted at 65 US medical centers between December 1998 and June 2001 (study group members are listed at the end of this article). Approval of the study was obtained from the institutional review committee at each participating institution and written informed consent was obtained from each patient (or surrogate). Each institutional review committee determined who could qualify as a patient surrogate for the purpose of giving consent at their institution. The study was monitored for safety by an independent data and safety monitoring board, which met 11 times during the course of the study; the stopping rule was a mortality difference of P<.001 (DSMB listed at the end of this article). The study objective was to determine if administration of rHuEPO to critically ill patients admitted to the ICU would reduce the occurrence of any RBC transfusion as well as reduce the cumulative number of RBC units transfused.

The study was designed by the principal investigators with input from the data coordinating center and was reviewed by the study sponsor. Patient enrollment was done at each site and supervised by the data coordinating center. Randomization and data analysis were done by the data coordinating center. The principal investigators and manuscript committee, with assistance from the data coordinating center, interpreted the data and were responsible for the manuscript. The final manuscript was reviewed by the study sponsor. The manuscript committee determined the final manuscript content and had full access to all data.

All patients admitted to either a medical, surgical, or a medical/surgical ICU in each of the 65 participating institutions, who remained in the ICU for at least 2 days, were evaluated for study eligibility prior to ICU day 3 (study day 1). Inclusion criteria included stay in the ICU for 3 days; age at least 18 years; hematocrit less than 38%; and provision of signed informed consent. Exclusion criteria included renal failure with dialysis; uncontrolled hypertension; new onset or uncontrolled seizures; acute burns; pregnancy or lactation; acute ischemic heart disease; acute gastrointestinal bleeding; prior treatment with rHuEPO; participation in another research protocol; and expected ICU discharge within 48 hours of ICU day 2. Patients who met entry criteria and who gave informed consent were randomized and entered into the study on ICU day 3 (study day 1). Randomization was stratified by site and entailed use of computer-generated random numbers (Figure 1).

Study drug (40 000 units of rHuEPO [Procrit; Ortho Biotech Products, LP, Bridgewater, NJ]) or a placebo identical in appearance was administered by subcutaneous injection on ICU day 3 and continued once weekly for patients who remained in the hospital, for a total of 3 doses (study days 1, 7, and 14). Patients who remained in the ICU on study day 21 received a fourth dose. Syringes were prepared either in the hospital's pharmacy or in the patient care area. Study drug was withheld if the hematocrit was 38% prior to the scheduled administration. All patients were to be followed up for 28 days following randomization (study day 28).

Patients received oral iron (liquid preparation), at least 150 mg/d of elemental iron, either orally or via nasogastric tube beginning on study day 1 (ICU day 3), unless they could not tolerate oral feeding. Parenteral iron was to be given to patients demonstrating an inadequate response to oral iron (transferrin saturation <20% and a decrease of serum ferritin to <100 ng/mL [<225 pmol/L]).

The need for RBC transfusion was determined by each patient's physician. The following transfusion guideline was established for the study: no RBC transfusion if the hemoglobin level was at least 9 g/dL or the hematocrit concentration was at least 27%, unless there was a specific clinical indication (active bleeding, ischemia, or other); RBC transfusion for a hemoglobin level less than 9 g/dL or a hematocrit concentration less than 27% was at a physician's discretion. There was no hemoglobin level or hematocrit concentration that mandated an RBC transfusion. Transfusion indication, pretransfusion hemoglobin level, and pretransfusion hematocrit concentration were recorded for each RBC transfusion. In addition, patients were monitored for all adverse events associated with either drug administration or RBC transfusion.

The primary efficacy end point was transfusion independence, assessed by comparing the percentage of patients in each treatment group who received any RBC transfusion between study days 1 and 28.

Secondary efficacy end points identified prospectively were cumulative RBC units transfused per patient through study day 28; cumulative mortality through study day 28; change in hemoglobin from baseline; and time to first transfusion or death. Additional data recorded included ICU length of stay, hospital length of stay, and days receiving mechanical ventilation.

Baseline demographic, diagnostic, and laboratory data were obtained at randomization. Acute Physiology and Chronic Health Evaluation (APACHE) II scores were based on data obtained within the first 24 hours after ICU admission. Admitting diagnosis was selected from a list of diagnostic categories; all that applied were noted. Comorbidities identified from the medical history included cardiac disease, chronic pulmonary disease, diabetes mellitus, hypertension, malignancy, peripheral vascular disease, primary hematologic disease, and thromboembolic disease.

Adverse events were assessed daily. Laboratory data were obtained weekly, within 24 hours prior to the weekly administration of study drug, and at study completion. Patients discharged from the hospital prior to study day 28 had final laboratory data obtained within 7 days of study day 28. Mechanical ventilation was recorded on a daily basis, as was a patient's presence in the ICU.

A sample size of 1300 was calculated to provide at least 90% power to detect an absolute treatment difference of 10% with respect to the primary efficacy end point, percentage of patients receiving any RBC transfusion. All patients were followed up for 28 days, unless death occurred earlier. Analysis of outcomes was on an intent-to-treat basis.

The primary efficacy end point was evaluated using a 2-sided Fisher exact test. Patients not receiving transfusion at the time of study withdrawal or lost to follow-up were considered not transfused for this analysis. In addition, a second analysis was done in which all patients withdrawn or lost to follow-up were considered transfused.

Secondary analyses were specified in the original study protocol. To compare the number of RBC units transfused per patient in the 2 study groups, we used the Mann-Whitney test. Transfusion rate, expressed as RBC units transfused per day alive while in the study, was determined by dividing the total number of RBC units transfused for each group by the total number of days alive for the patients in that group (sum of the number of days alive for each individual patient in the group, maximum of 28 for a patient surviving to study day 28).

Mechanical ventilation was analyzed as ventilator-free days, defined as the number of days a patient was alive and did not receive mechanical ventilation through study day 28. Patients were included whether mechanical ventilation was initiated on or after study day 1. Patients who were reventilated were also included. Patients were excluded only if they withdrew from the study, were lost to follow-up, or had a missing ventilator stop date with a hospital discharge date prior to study day 28. Reventilation was analyzed only for those patients who were candidates for reventilation, ie, discontinued initial mechanical ventilation prior to study day 28 and who did not die on the day of or the day after discontinuation of mechanical ventilation. Ventilator-free days were compared using the Mann-Whitney test. Similarly, ICU length of stay was analyzed as ICU-free days, defined as the number of days a patient was alive and not in the ICU through study day 28. Patients were excluded only if they withdrew from the study or were lost to follow-up. Readmission to the ICU was analyzed only for those patients who were candidates for readmission, ie, discharged from their initial ICU stay prior to study day 28 and did not die on their last day in the ICU. Number of ICU-free days was compared using the Mann-Whitney test.

Several survival analyses of time to event (death, first transfusion, first transfusion or death, ICU readmission, ventilation, reventilation) were performed. Kaplan-Meier survival curves for the 2 groups were compared using the log-rank test.

Logistic regression was performed to adjust the odds ratio (OR) for RBC transfusion. Covariates entered included age, sex, diagnostic categories, comorbidities, APACHE II score, baseline hemoglobin, baseline iron, baseline erythropoietin level, and baseline serum creatinine. Two statistical approaches, Cox semiparametric modeling assuming proportional hazards and recursive partitioning, were used to explore the association of mortality with treatment and other variables.

To assess changes in laboratory values from baseline to final value, analysis of covariance was used with baseline value and number of days between baseline and final value as covariates.

Results are presented as mean (SD) unless otherwise indicated. All statistical tests were 2-tailed at the .05 significance level, except for the tests of treatment-by-covariate interactions, which were at the .10 significance level. Analyses were conducted with SAS OnlineDoc version 8 (SAS Institute Inc, Cary, NC).

In addition to the main analysis of all patients, several subgroups of patients were analyzed separately for both percentage of patients undergoing transfusion and mortality. Although these subgroups were not prospectively identified, they were identified prior to knowledge of treatment assignment and prior to the locking of the database and the conduct of the final data analysis. The DSMB identified 3 mutually exclusive admitting diagnostic categories for safety monitoring during the study: trauma; surgical, nontrauma; and medical, nonsurgical, nontrauma. These 3 admitting diagnostic categories were determined independently of the baseline admitting diagnoses. Four other subgroups identified were identical to the subgroups studied by Hebert et al¹³ to allow for comparison of our findings with theirs: age younger than 55 years; age 55 years or older; APACHE II score 20 or lower; and APACHE II score higher than 20.

Among the 33 685 patients screened on ICU day 2 (Figure 1), more than 70% were ineligible for the study, primarily because of expected ICU discharge within 24 to 48 hours. Of those eligible, approximately two thirds were not approached for consent to participate, in most instances because of inability to identify and contact the appropriate patient surrogate(s) during the window of time for study enrollment. Of those patients (or surrogates) asked to consent, 50% agreed.

A total of 1302 patients were enrolled in the study, with 650 randomized to receive rHuEPO and 652 to receive placebo. The 2 groups were generally comparable at enrollment with respect to baseline demographic characteristics and laboratory values as well as admitting diagnosis and comorbidities (Table 1). There were some statistically significant differences; however, the magnitudes were not clinically meaningful and multivariate analysis took these variables into account. The study drug exposure was as follows: 15% received 1 dose; 31% received 2 doses; 37% received 3 doses; and 17% received 4 doses.

The percentage of patients who received any RBC transfusion during the 28-day follow-up was significantly lower in the rHuEPO group than in the placebo group (n = 328 [50.5%] vs n = 394 [60.4%]; P<.001; OR, 0.67; 95% confidence interval [CI], 0.54-0.83). After adjustment for baseline characteristics, the effect of rHuEPO was essentially unchanged (adjusted OR, 0.65; 95% CI, 0.51-0.83). The results of an additional analysis in which all patients withdrawn or lost to follow-up were considered to have had transfusion were similar to those described above (63.3% placebo vs 53.4% rHuEPO; OR, 0.66; 95% CI, 0.53-0.83).

A Kaplan-Meier plot of the time to first transfusion indicates that a difference between the 2 treatment groups commenced near the end of the first week following randomization and increased progressively over the course of the 28-day follow-up (Figure 2A). A similar pattern occurred for the composite end point of time to first transfusion or death (Figure 2B).

The cumulative number of RBC transfusions for each treatment group over the 28-day follow-up appears in Table 2. The total number of RBC units transfused was 1590 units for rHuEPO therapy compared with 1963 units for placebo therapy. The cumulative RBC transfusions were significantly lower in patients receiving rHuEPO compared with placebo patients. This was initially assessed based on the cumulative RBC units per subject (median units per subject, 1 vs 2; P<.001). A second analysis that accounted for time at risk for transfusion demonstrated a 19% reduction in RBC units transfused per day alive (ratio of transfusion rates, 0.81; 95% CI, 0.79-0.83; P = .04). Figure 3 displays the cumulative units transfused by study day. This plot suggests that the treatment groups do not begin to differ with regard to RBC units transfused until approximately 1 week following treatment.

The mean (SD) increase in hemoglobin from baseline to final determination was significantly greater for patients who received rHuEPO (1.32 [2] g/dL vs 0.94 [1.9] g/dL for placebo; P<.001). The mean day the final hemoglobin measurement was obtained was identical for both groups at study day 23.

Transfusion practices were similar in the 2 treatment groups. The mean pretransfusion hemoglobin was 8.57 (0.96) g/dL for the placebo group and 8.53 (1.08) g/dL for the rHuEPO group. Among those patients undergoing transfusion, pretransfusion hemoglobin was similar in the 2 treatment groups for the first RBC transfusion as well as for all subsequent RBC transfusions. Similarly, 21% of patients in each group underwent transfusion at a hemoglobin level greater than 9 g/dL or hematocrit greater than 27%.

There was no significant difference in 28-day mortality between the 2 groups (14% in rHuEPO vs 15% in placebo; P = .61, Figure 4). The incidence of severe adverse events reported was comparable between the 2 treatment groups (Table 3).

Median hospital length of stay (19 days for rHuEPO vs 21 days for placebo; P = .82) and median ICU-free days (18 days for rHuEPO vs 17 days for placebo; P = .25) did not differ between groups. However, there was a higher, but not statistically significant, ICU readmission rate in the placebo group compared with the rHuEPO group (13.3% vs 9.8%, respectively; P = .07).

Median ventilator-free days did not differ between the 2 groups (22 days for rHuEPO vs 20 days for placebo; P = .27). There were also no statistically significant differences between the 2 treatment groups in either reventilation rate (16.6% for rHuEPO vs 20.5% for placebo; P = .17) or new-onset ventilation (20.8% for rHuEPO vs 24.4% for placebo; P = .38).

The subgroups analyzed were admitting diagnosis (trauma; surgical, nontrauma; and medical, nonsurgical, nontrauma); age younger than 55 years; age 55 years or older; APACHE II score 20 or lower; and APACHE II score higher than 20.

The percentage of patients undergoing transfusion and the OR for RBC transfusion for the entire group and for each subgroup are shown in Table 4. The reduction in the percentage of rHuEPO patients undergoing transfusion was consistent across all the subgroups analyzed, as well as for the range of baseline hemoglobin levels (Table 4).

Mortality in each subgroup for patients receiving rHuEPO and patients receiving placebo is shown in Table 5. Overall mortality varied widely by subgroup, with the expected increase in mortality with older age and higher APACHE II scores. There was also considerable variation in mortality across the subgroups by treatment group. However, multivariate analysis using Cox semiparametric and recursive partitioning approaches demonstrated no treatment or treatment-by-baseline variable interaction with mortality.

In a previous randomized, placebo-controlled trial conducted in 160 patients, a combination of daily administration of rHuEPO (300 U/kg) followed by administration every other day resulted in an almost 50% reduction in total RBC transfusions.²⁰ There was also a trend toward an increase in the percentage of patients receiving no RBC transfusions (transfusion independence) with rHuEPO therapy. Our study, in a much larger and more diverse group of critically ill patients, expands the findings of the prior trial. A weekly dosing schedule of 40 000 units of rHuEPO significantly increased the percentage of patients achieving transfusion independence as well as reduced the cumulative number of RBCs transfused compared with placebo patients. The RBC transfusion reduction was consistent across all of the subgroups examined, although not all comparisons were statistically significant. In both studies, despite the reduction in the number of RBC transfusions, the increase in hemoglobin level was significantly greater with rHuEPO therapy.

The reduction in total RBC units transfused and the increase in hemoglobin level were more modest in this study. This smaller transfusion effect is in part a consequence of the 2-week shorter follow-up period. If the total RBC units transfused in the 2 studies are compared at 28 days following randomization, the reduction in RBC transfusion in the earlier trial²⁰ is approximately 30%, closer to the 19% reduction seen in this trial. Therefore, it is possible that a longer follow-up period would have demonstrated a greater reduction in RBC transfusion than the 19% reduction observed. An additional contributing factor to the difference in effect size may be the decrease in total rHuEPO dose received by patients in our trial. In the current study patients received, on average, half the amount of rHuEPO that was given in the earlier trial (80 000 units vs 160 000 units).

Transfusion practice varies greatly among physicians and institutions.²⁵ Recently, a transfusion strategy in critically ill patients to maintain a hemoglobin level between 7 and 9 g/dL has been shown to be as effective as, and in some subgroups superior to, a transfusion strategy to maintain a hemoglobin level between 10 and 12 g/dL.^13,26 However, recent reports suggest that the "transfusion trigger" in most ICUs remains higher than the more conservative transfusion strategy.^5- 6 Clearly, transfusion practice could affect our results and the magnitude of any potential benefit resulting from rHuEPO therapy. In the current study, 60% of the patients in the placebo group underwent transfusion. This rate of transfusion, as well as the number of units transfused per patient, was comparable to the transfusion practice pattern observed in the prior trial as well as the restrictive group in the TRICC trial.^13,20 Similarly, the pretransfusion hemoglobin level (8.5 g/dL) was identical in both the placebo and the rHuEPO groups, and importantly, transfusion practice in the current trial is consistent with recent reports of transfusion practices in ICUs across the United States⁶ and Western Europe.⁵ Therefore, the results of this study reflect the efficacy of rHuEPO in a setting representative of the predominant transfusion practice in ICUs today.

In view of questions that have been raised regarding the safety and efficacy of RBC transfusions,¹³ does a reduction in RBC transfusions with rHuEPO therapy lead to better clinical outcomes? In the present study there were no significant differences in morbidity or mortality observed between the 2 groups. Clearly, the current study does not have the power to identify small differences in clinical outcomes among subgroups.

As in the prior study,²⁰ rHuEPO therapy did not increase serious clinical events. As with clinical outcomes, even a study as large as the current one does not have the power to identify less common adverse events. Recently, the occurrence of pure red-cell aplasia associated with the presence of antierythropoietin antibodies was reported in a small number of patients with chronic renal failure treated with rHuEPO administered subcutaneously.^27- 28 This phenomenon was not observed in our trial, although our follow-up time was short.

The individuals studied represent a diverse group of critically ill patients from multiple ICUs across the United States. The major reason for ineligibility for the study was early discharge from the ICU. The intent of the study was to focus on "long-term" ICU patients, those with the highest transfusion burden.³ The 30% of patients found eligible for the study is consistent with other reports regarding the number of patients remaining in the ICU for longer than 1 week.^2,5- 6 Only one third of eligible patients were approached for consent, reflecting the difficulty of performing research in a critically ill patient population.^29- 30 Regulations governing who can serve as a surrogate to provide consent for participation in a research study when patients are unable to consent for themselves are becoming stricter. This is a particularly important issue for research in the ICU, where patients often are not able to provide consent for themselves. The difficulty in identifying and contacting the appropriate surrogate in the relatively short time frame prior to randomization was responsible for the majority of the instances in which consent was not requested. There did not appear to be any systematic exclusion of patients other than for the reasons delineated in the protocol, but no information about excluded patients was collected to assess whether the group studied was representative.

The cost of rHuEPO is approximately $400 for each 40 000-unit dose while the cost of a unit of RBCs is generally in the $300 to $400 range. The average patient received 2 or 3 doses of rHuEPO ($800-$1200) and avoided approximately 1 unit of RBCs ($300-$400). However, the reduction in the number of RBC units transfused (19%) may itself be beneficial, independent of any additional clinical outcome effects. For example, not transfusing "unnecessary" units of RBCs avoids the morbidity and mortality directly associated with each RBC unit transfused (ie, transfusion reactions, transfusion-related infection) as well as the potential for medical errors associated with the transfusion process itself.^31- 32 The avoidance of unnecessary RBC transfusions would also save a resource that is becoming increasingly scarce. Establishing whether rHuEPO therapy is cost-effective will involve the consideration of all these factors as well as any additional benefit in clinical outcome achieved.

Could similar benefit be achieved by changing transfusion practice? Clearly, as demonstrated by Hebert et al,¹³ changing transfusion practice will substantially impact the number of RBC units transfused. However, even in the restrictive group of the study by Hebert et al, two thirds of the patients received at least 1 RBC transfusion, a rate of transfusion similar to that observed in the placebo group in the current study. Therefore, the potential to reduce transfusion with the use of rHuEPO remains at even more restrictive transfusion thresholds. Key in identifying critically ill patients who will be most likely to benefit from rHuEPO is selecting the more "long-term" critically ill patient (ie, longer than 1 week length of stay), who represent 25% to 30% of critically ill patients.^3,5- 6

In conclusion, weekly therapy with 40 000 units of rHuEPO in critically ill patients results in a significant reduction in their exposure to allogeneic RBC transfusion. Despite receiving fewer RBC transfusions, patients treated with rHuEPO achieve a higher hemoglobin level, consistent with the hypothesis that the anemia of critical illness is an underproduction anemia characterized in part by a relative erythropoietin deficiency.^14- 15 No differences in clinical outcomes were demonstrated between the rHuEPO and placebo groups. Therefore, while it is clear that rHuEPO treatment reduces RBC transfusions in critically ill patients, further study is necessary to establish whether this reduction in RBC transfusions will also result in improved clinical outcomes for some critically ill patients.