In recent years there has been an increasing debate regarding the benefits of initiating erythropoietic treatment in patients with cancer when anemia is still relatively mild. To address this, a systematic review of studies was conducted that answered the following question: Is there a clinical benefit associated with early erythropoietic intervention (hemoglobin ≥ 10 g/dL) for chemotherapy-induced anemia?
A systematic review of published literature and meeting abstracts was undertaken to identify relevant studies. Data were extracted from studies meeting prespecified eligibility criteria. For outcome measures not associated with significant heterogeneity, summary measures of relative risk associated with early erythropoietic intervention were estimated using the method of Mantel and Haenszel.
Eleven studies were eligible and were included in the review. Erythropoietic treatment effectively decreased transfusion incidence and the proportion of patients with hemoglobin < 10 g/dL compared with no treatment, with relative risk reductions of 0.50 (95% confidence interval [CI], 0.43, 0.59; 7 studies, P < 0.0001) and 0.40 (95% CI, 0.19, 0.83; 4 studies, P = 0.147), respectively. The findings from both prospective studies and planned subset analyses in which early and late intervention were compared also indicated a reduction in the relative risk of both transfusions and hemoglobin < 10 g/dL after early intervention (0.55 [95% CI, 0.42, 0.73; 5 studies, P < 0.0001] and 0.44 [95% CI, 0.33, 0.57; 2 studies, P < 0.0001], respectively).
Chemotherapy-induced anemia (CIA) is a frequent complication that results from myelosuppressive chemotherapy. The resulting fatigue is associated with substantial physical, emotional, psychological, and social consequences, affecting almost all aspects of daily life.1 This is particularly true for older patients, who may experience an increased mortality, functional dependence, cognitive abnormalities, falls, and cerebrovascular and cardiovascular abnormalities as a result of their anemia.2 Anemia has been classified into five grades of increasing severity.3, 4 While the incidence of Grades 3 and 4 anemia (hemoglobin < 8 g/dL) has been reported to range from 50–60% in patients with lymphomas or lung, gynecologic, and genitourinary carcinomas receiving chemotherapy, the incidence of Grade 1 (hemoglobin ≥ 10 g/dL to < 12 g/dL) or 2 (hemoglobin ≥ 8 g/dL to hemoglobin < 10 g/dL) anemia appears to be even higher across many of the major solid tumors.5
Clinical practice guidelines for the management of CIA, issued jointly by the American Society of Clinical Oncology (ASCO) and the American Society of Hematology (ASH),6 have recommended the use of erythropoietic agents in patients with hemoglobin ≤ 10 g/dL. These recommendations were based on a systematic review of the literature (from 1985 to 1999) on the effectiveness of erythropoietic proteins for the treatment of CIA that included 22 randomized controlled trials (RCTs) in 1927 patients7 (Table 1). In contrast, the guidelines made no definitive recommendation for patients with Grade 1 anemia. Instead, the committee deferred to the treating physician to decide, based on clinical circumstances, whether to use recombinant human erythropoietin (rHuEPO) immediately or to wait until hemoglobin declined to 10 g/dL. This recommendation was based on the judgment that the existing evidence in support of early intervention, based on only three studies,8–10 was equivocal.
Table 1. American Society of Clinical Oncology, American Society of Hematology (ASCO/ASH) and National Comprehensive Cancer Network (NCCN) Guidelines for the Management of Chemotherapy-Induced Anemia (CIA)
Hb ≤ 10 g/dL (clinical decision if < 12 g/dL and > 10 g/dL)
Hb ≤ 11 g/dL
Objective of treatment
To maintain Hb level at or near 12 g/dL
To maintain optimal Hb level (11 g/dL to 12 g/dL)
Cessation of treatment
In the absence of a Hb response, even after dose escalation, discontinue therapy after 6 to 8 weeks
In the absence of a Hb response, even after dose escalation, discontinue therapy
Hb target > 12 g/dL
No recommendation was made to target Hb concentrations above 12 g/dL
No recommendation was made to target Hb concentrations above 12 g/dL
The ASCO/ASH committee also judged that there was insufficient evidence to suggest a health-related quality of life (HRQOL) benefit associated with erythropoietic treatment. However, retrospective analyses and prospective studies published since then have demonstrated that anemia correction as a result of erythropoietic therapy is associated with improved HRQOL outcomes.11–14 Some of these studies have also shown that early intervention leads to greater improvements in HRQOL measures compared with late intervention.
The relative impact of Grade 1 anemia on patient fatigue and overall HRQOL remains a controversial topic within the oncology community, instigating much debate regarding the merits of initiating erythropoietic treatment when anemia is still relatively mild, and prompting wide variations in practice and reimbursement patterns across the U.S. The purpose of this systematic review is to first determine whether there is substantial evidence to suggest a clinical benefit for initiating erythropoietic therapy in cancer patients with hemoglobin ≥ 10 g/dL versus no treatment. Second, this study addresses whether there is a comparative clinical benefit associated with treating mild anemia versus waiting until hemoglobin falls below 10 g/dL before intervening.
MATERIALS AND METHODS
We performed a literature search of the Medline and PreMedline databases (http://www.ncbi.nlm.nih.gov/pubmed/) from 1999 to June 15, 2004 using the Boolean search string: “erythropoietin AND cancer AND anemia NOT myelodysplastic.” The search was limited to clinical trials in human subjects that were reported in English. References from identified articles were also reviewed for additional citations. We also conducted a search for relevant abstracts and associated poster presentations delivered at ASCO and ASH conferences from 1999 to the last-held conference (2004 and 2003, respectively).
Criteria for Study Selection
Studies were selected for review if all of the following criteria were met: they were designed as RCTs in patients with CIA; a stated objective was to evaluate the ability of erythropoietic intervention to prevent the onset or worsening of CIA; a baseline hemoglobin concentration ≥ 10 g/dL was required for eligibility; the studies included one or more of the following endpoints: incidence of transfusion, changes in hemoglobin, hemoglobin response, or HRQOL. In addition, studies reporting combined, planned subset analyses based on baseline hemoglobin, or retrospective analyses of clinical trial data, were included if the study objectives incorporated an assessment of the effect of baseline hemoglobin on transfusion incidence, hemoglobin change, hemoglobin response, or HRQOL. For studies reporting on the same dataset, the most recent citations available were used.
Studies were separated into those that evaluated the effects of treating mild anemia compared with no erythropoietic intervention, and those that assessed the clinical benefits of treating mild anemia versus waiting until hemoglobin < 10 g/dL before treating. Because of the relatively small individual study sample sizes and somewhat divergent results across studies, a formal metaanalysis was planned in order to provide the most accurate and reliable estimate of treatment effect when all available data were considered.
Study-by-study heterogeneity was estimated based on the Q statistic. The hypothesis that the studies are all drawn from a population of studies with the same effect size is rejected if Q exceeds the upper 100 (1-α) percentile of the chi-square distribution. An inconsistency index (I2) was calculated as an estimate of the proportion of variation in estimates due to heterogeneity, rather than between-study variation. The I2 was estimated by the method of Altman as (H2 – 1)/H2 where H2 = Q/(k-1) and K is the degrees of freedom.15
After assessment for heterogeneity, an analysis was undertaken providing summary estimates of relative risk (RR) for transfusion and hemoglobin < 10 g/dL using the method of Mantel and Haenszel.16 The standardized mean difference in hemoglobin change was estimated by the method of Cohen.17 The combined estimates represent the weighted sum of the individual estimates where the weights are the reciprocal of the variance or the interstudy-adjusted variance of the estimates depending on the model applied. Fixed effects models were used to estimate summary measures when no significant heterogeneity was found across studies. Under the fixed effects model, we assume that all studies come from a common population, i.e., if the sample size in each study were infinite, the true effect size would be the same in all studies. Assuming this model, the only source of variation is random error; therefore, the standard error will approach zero as the sample size becomes large, revealing the true effect size. Alternatively, random effects models were used to estimate summary measures when significant heterogeneity was observed for outcomes across reporting studies. Under the random effects model, the samples are assumed to be from different populations with different true effect sizes. The true effect may differ between studies because of differences in patient populations, treatment variation, or because outcome measures differ from one study to the next. Therefore, two sources of variation are assumed: random error and variation due to real differences between the populations, treatment, or outcomes. The standard error of the effect size estimates will approach zero as the sample size within studies or the number of studies in the analysis increase. However, the differences in the true effect sizes between studies will persist.
Hypothesis testing on summary effect estimates was based on a z-statistic with estimates of standard error and 95% confidence intervals (CIs) provided for all individual studies, as well as the summary overall effect estimate. Results are presented as Forest plots where appropriate, with effect estimates and 95% CIs presented for each individual study, and a summary measure and CI presented for all studies combined. For studies with zero events for one group, a modified estimator of relative risk was used adding 0.5 to the number of events and nonevents for each group within the study. Such an amended estimator has been shown to be well-behaved in terms of bias and mean-square error with an asymptotic normal distribution around the true RR.18, 19
All of the prospective studies included in this review were randomized in design; however, only the Vansteenkiste et al.20 and the Kunikane et al.21 studies were blinded. In addition, all of the full-length reports provided reasons and accompanying numbers for patient withdrawals and dropouts.
By using our search strategy, 69 references were collected, seven of which met the criteria described in Materials and Methods and were further evaluated.10, 20–25 Five abstracts met selection eligibility criteria and were included in the review.26–30 The authors of four of these abstracts posted presentations on the respective conference websites.27–30 Initially, we identified studies that evaluated the clinical outcomes associated with treating mild anemia compared with no treatment.
Are There Clinical Benefits Associated with Initiating Erythropoietic Therapy in Patients with Hemoglobin Concentration ≥ 10 g/dL versus No Treatment?
The clinical benefits associated with treating mild anemia were first suggested by two studies published before 1999 that were included as part of the metaanalysis used to develop the ASH/ASCO guidelines.7 To determine whether these early observations were substantiated by subsequent studies, and to evaluate whether there now exists a substantial body of evidence in favor of early intervention, we included these studies in our analysis, in addition to the five studies we identified using our search criteria.10, 20, 24, 27, 30
Table 2 summarizes the transfusion outcomes of all seven studies. All were prospective studies: six compared the effects of treatment with epoetin alfa versus no treatment or best supportive care,8–10, 24, 27, 30 while one study was a double-blind, placebo-controlled trial of darbepoetin alfa.20 The results indicated that transfusion rates decreased with erythropoietic therapy, and were significantly lower than control in all but one study.8 The proportion of patients requiring transfusion across the seven RCTs was 22.5% (95% CI, 19.6–25.7%) in patients randomized to receive rHuEPO and 41.3% (37.5–45.3%) among control patients. No significant heterogeneity was observed for this outcome across studies (P = 0.1064) with an I2 of 0.43. A fixed effects model was used to calculate the weighted summary RR of a transfusion associated with rHuEPO treatment across the seven studies (RR 0.50 [0.43, 0.59]; P < 0.0001; Fig. 1A).
Table 2. Randomized Controlled Trials to Evaluate Early Erythropoietic Intervention in Patients with Chemotherapy-Induced Anemia: Effect on Transfusion Incidence
Across all seven studies, hemoglobin levels improved with erythropoietic treatment, thereby paralleling the increase in transfusion independence (Table 3). Improved hemoglobin response was also observed in a small (n = 72) RCT of rHuEPO (epoetin beta) in nonsmall cell lung cancer patients with CIA.21 A significant dose-dependent increase in the change in mean hemoglobin concentration from baseline was observed after 8 weeks of treatment in patients who had received rHuEPO 200 U/kg or 100 U/kg compared with those who had not received treatment (200 U/kg arm: 1.72 g/dL [standard error 0.26]; 100 U/kg arm: 0.89 g/dL [0.43]; untreated control arm: −0.65 g/dL [0.30]; P < 0.01 vs. control for both treatment arms).
Table 3. Summary of Hemoglobin Outcomes from Early Intervention Studies of Erythropoietic Agents in Patients with Chemotherapy-Induced Anemia (CIA)
The proportion of patients experiencing hemoglobin < 10 g/dL across the four reporting RCTs was 21.2% (15.9–27.8%) in patients randomized to receive rHuEPO and 52.6% (45.2–60.0%) among control patients. Significant heterogeneity was observed for this outcome across studies (Q = 12.00, P = 0.007) with an I2 of 0.75. A random effects model was used to estimate the summary RR of a hemoglobin decline to < 10 g/dL. As shown in Figure 1B, the weighted summary RR of hemoglobin < 10 g/dL associated with rHuEPO treatment across these studies was 0.40 (0.19, 0.83; P = 0.0147). Given that these data suggest clinical benefits to treating mild anemia compared with no treatment, we subsequently identified studies that compared clinical outcomes associated with erythropoietic intervention at different baseline hemoglobin categories (< 10 g/dL vs. ≥ 10 g/dL).
Are There Clinical Benefits Associated with Initiating Erythropoietic Therapy in Patients with Mild Anemia (≥ 10 g/dL) versus Hemoglobin < 10 g/dL?
Planned subset analyses25, 31 of Phase 3 RCT data of rHuEPO and darbepoetin alfa in the CIA setting11, 20 evaluated the relationship between baseline hemoglobin concentration and transfusion outcomes. Both Phase 3 trials had prospectively specified analyses of hematologic outcomes by baseline hemoglobin category: > 10.5 g/dL versus ≤ 10.5 g/dL in the Littlewood et al. trial11 and ≥ 10 g/dL versus < 10 g/dL in the Vansteenkiste et al. trial.20 Although neither was designed or powered to detect a significant difference in transfusion outcomes between early and late intervention, each study showed a substantial reduction in transfusion requirements among patients treated early compared with those treated late. This initial finding was confirmed in three RCTs identified in our review26, 28, 29 that directly compared the outcomes of early versus late erythropoietic intervention (Table 4).
Table 4. Randomized Trials of Erythropoietic Agents to Evaluate Early versus Late Intervention: Effect on Transfusion Incidence
Hb: hemoglobin; SE: standard error; rHuEPO: recombinant human erythropoietin.
Hb entry criteria in g/dL
≥ 10 but ≤ 12
≥ 11 but < 15
≥ 10.5 but ≤ 12.0
Baseline Hb in g/dL Mean (SD)
11.1 (SE 0.7)
11.2 (SE 0.7)
No. patients treated
Across all five studies the proportion of patients requiring transfusion was 14.3% (11.4–17.7%) in those randomized to receive early treatment and 25.6% (22.2–29.2%) among those receiving late treatment. Again, no significant heterogeneity was observed across studies (P = 0.518) with an I2 of zero. The weighted summary RR of a transfusion associated with erythropoietic treatment across these studies, using a fixed effects model, was 0.55 (0.42, 0.73; P = 0.0001; Fig. 2A). In aggregate, these combined analyses provide support for the hypothesis that mildly anemic patients who receive erythropoietic intervention experience a significant reduction in transfusion requirements compared with later intervention.
The proportion of patients with CIA experiencing hemoglobin < 10 g/dL in the two reporting RCTs28, 29 (Table 5) of early versus late erythropoietic treatment was 23.4% (18.1–29.7%) in those randomized to receive early treatment and 54.1% (47.3–60.8%) among those whose treatment was delayed. No significant heterogeneity was detected in these studies (P = 0.727). As shown in Figure 2B, the weighted summary RR of hemoglobin < 10 g/dL associated with early treatment across these studies was 0.44 (0.33, 0.57; P < 0.0001).
Table 5. Summary of Hemoglobin Outcomes Associated with Early versus Late Erythropoietic Intervention in Patients with Chemotherapy-Induced Anemia (CIA)
In addition to the evaluation of the potential benefit of erythropoietic therapy on hematologic measures, several recent studies have also studied the impact of the timing of treatment initiation and HRQOL (Table 6).
Table 6. Summary of Health-Related Quality of Life (HRQOL) Outcomes from Early Intervention Studies of Erythropoietic Agents in Patients with Chemotherapy-Induced Anemia (CIA)
Study sample size (n)
Hb inclusion criterion
Hb: hemoglobin; rHuEPO: recombinant human erythropoietin.
Retrospective analysis of 2 nonrandomized, single-arm, open-label, community-based trials32, 33
Relation between Hb and HRQOL: Significant positive correlation between Hb levels and LASA overall HRQOL scores in both trials (r = 0.25 and 0.29, respectively; P < 0.01 × 2). Similar correlation between Hb and FACT-Anemia scores was observed (r = 0.27; P < 0.01 × 2).
Relation between incremental changes in Hb levels and HRQOL scores: A nonlinear relationship was observed between Hb and HRQOL up to Hb = 12 g/dL. Beyond this, subsequent incremental increases in Hb corresponded with smaller magnitude of HRQOL improvement.
Retrospective analysis of data from a nonrandomized, open-label, community-based trial
Breast cancer receiving anthracycline (± taxane) chemotherapy
Not reported Baseline Hb: 10 g/dL to 14 g/dL
Significant positive correlation observed between Hb and HRQOL in all domains tested (P < 0.0001 in each domain). The greatest increases in HRQOL scores occurred when Hb increased from 11.0 to 12 g/dL. A clinically meaningful improvement in overall HRQOL corresponded to rise in Hb from 10 to 12 g/dL.
99 randomized to early intervention arm; 102 randomized to late intervention arm
Early intervention: ≥ 10.5 g/dL but ≤ 12 g/dL Late intervention: ≤ 10 g/dL
Mean change in FACT-Fatigue subscale score was higher in early vs. late intervention group at the end of the test period (week 13; 1.5 vs −0.8, respectively) and comparable for both arms at the end of the treatment period (week 22; 1.5 vs. 1.8, respectively)
Crawford et al.22 conducted a retrospective analysis of data from two community-based studies of rHuEPO32, 33 in 4382 anemic cancer patients to determine the incremental change in HRQOL associated with a 1-g/dL increase in hemoglobin level. Two instruments were used to assess HRQOL—the Linear Analog Scale Assessment (LASA) and the Functional Assessment of Cancer Therapy – Anemia (FACT-Anemia). The analysis demonstrated a significant positive correlation between high hemoglobin levels and high LASA and FACT-Anemia scores (r = 0.25 and 0.29, respectively; P < 0.01). A longitudinal analysis of the relation between incremental changes in hemoglobin and HRQOL scores revealed the greatest improvement in HRQOL occurring when hemoglobin levels increased from 11 g/dL to 12 g/dL. The authors concluded that an increase in hemoglobin concentration correlated with HRQOL improvement in patients with CIA receiving erythropoietic therapy across the clinically relevant hemoglobin range of 8–14 g/dL.
Consistent with these results, Gralow et al.34 also reported a significant positive correlation between increased hemoglobin levels and improvements in energy, ability to perform daily activities, and overall HRQOL (P < 0.0001 in each case) among patients with breast cancer and CIA. Although that study did not report a hemoglobin inclusion criterion and was therefore not eligible for inclusion in our review, a longitudinal analysis demonstrated that an increase in hemoglobin levels from 10–12 g/dL was associated with clinically meaningful improvements in HRQOL. Furthermore, the authors also found that the greatest improvements in LASA scores, across the three domains, were coincident with a hemoglobin rise from 11 to 12 g/dL.
Straus et al.26 published results of an open-label RCT that compared the effectiveness of rHuEPO when administered early to patients who had mild anemia (defined as hemoglobin ≥ 10 g/dL but ≤ 12 g/dL) or after patients became moderately anemic (hemoglobin < 9 g/dL). Effectiveness was measured in terms of hemoglobin response, HRQOL, healthcare resource utilization, and patient productivity. Compared with patients whose treatment was delayed, patients who were treated early had significantly higher scores in various HRQOL assessments, spent fewer days in bed (52.2% vs. 3.1%, respectively; P = 0.017), and experienced fewer days of restricted activity (41.6% vs. 12.2%, respectively; P = 0.042). HRQOL outcomes were not reported in the Crawford et al. study28 but were shown in the Rearden et al. study29 to be better among patients treated early versus late at the end of the test period.
The combined analyses presented here strongly suggest a clinical benefit associated with initiating erythropoietic treatment at hemoglobin > 10 g/dL. Compared with placebo or no treatment, erythropoietic intervention at hemoglobin > 10 g/dL significantly reduced the relative risk of a transfusion and a hemoglobin decline below 10 g/dL. Moreover, the findings from both prospective studies26, 28, 29 and planned subset analyses25, 31 in which early and late intervention were compared also indicated a reduction in the RR of both transfusions and a hemoglobin decline below 10 g/dL after early intervention. Studies examining the impact of erythropoietic treatment on HRQOL22, 26, 29, 34 have demonstrated that early intervention not only ameliorates anemia symptoms but also offers significant improvements in HRQOL and productivity compared with initiating treatment only when anemia becomes more severe. It should be noted that there are several limitations associated with this literature review. The studies that directly compared early and late intervention have only been presented at meetings to date and have yet to appear in peer-reviewed journals. The comparative studies, although randomized, were open-label and thus may be subject to biases inherent in a nonblinded study, at least with respect to HRQOL outcomes.
The results of our combined analysis support the recommendations of the NCCN, who recently updated its practice guidelines for managing cancer and treatment-related anemia.35 The NCCN panel recommended that clinicians consider initiating erythropoietic intervention in cancer patients with hemoglobin ≤ 11 g/dL, based on recent evidence demonstrating the hematologic and HRQOL benefits of erythropoietic intervention at this hemoglobin level. The guidelines further state that optimal management of anemia is achieved when hemoglobin concentration is maintained between 11 g/dL and 12 g/dL for the longest duration throughout treatment.
The bulk of patient-experience data gathered to date indicate that erythropoietic agents are well tolerated and effective if used for the treatment of anemia. However, concern was recently raised by two clinical trial reports of epoetin alfa36 and epoetin beta37 in cancer patients in which a decrease in survival among rHuEPO-treated patients was observed. The first, a large study of breast carcinoma patients, reported an increased mortality for patients receiving epoetin alfa as a result of a higher incidence of disease progression and an increase in the number of thrombotic and vascular events.36 The second was a study of head-and-neck carcinoma patients receiving epoetin beta and concurrent radiotherapy, in which a shorter median duration of locoregional progression-free survival was observed for the treatment group.37 These studies were not included in our systematic review as they did not meet our inclusion criteria: the full details of the breast carcinoma study have not been published36 and the Henke et al. study evaluated patients receiving radiotherapy, not chemotherapy.37 It is of note that the observed adverse outcomes may have arisen in part from the initiation of treatment in nonanemic patients (hemoglobin > 12 g/dL)36 and allowing hemoglobin levels to rise to 15 g/dL during the study period.37 Unfortunately, several limitations associated with both studies have prevented firm conclusions regarding causality to be reached.38–40 In the breast carcinoma study, important prognostic factors for survival were not assessed or documented while the study was ongoing. For the Henke et al.37 study, an imbalance in baseline prognostic factors between the treatment and control groups may have affected the outcome. Regardless of the limitations or interpretation of these studies, concerns regarding the safety of erythropoietic therapy should not be ignored, particularly with respect to the normalization of hemoglobin concentration and the rapidity of hemoglobin increase. Additional randomized controlled studies are under way to determine whether a risk exists.
In general, no obvious differences in adverse events between the treatment and control arms of the studies evaluated in this systematic review have been reported in the literature. However, given the potential adverse effects in treating nonanemic patients, it remains prudent to limit erythropoietic therapy to patients with demonstrable anemia, i.e., hemoglobin < 12 g/dL. Consistent with evidence-based guidelines, hemoglobin targets above this should not be considered in routine clinical practice.
The results of our analysis, and the recent safety concerns, together with the NCCN recommendations, suggest a need on the part of clinical researchers in the field to consider aligning endpoints in future studies with current clinical goals. Appropriately powered studies should prospectively assess transfusion reduction and HRQOL improvement in specific patient populations. Furthermore, if hemoglobin concentration over time is evaluated as a surrogate endpoint for these clinical outcomes, it must be shown to correlate with clinically meaningful improvements in transfusions and anemia-related symptoms. Thus, one endpoint that may be appropriate to evaluate is the time to achieve the NCCN-recommended hemoglobin target concentration of 12 g/dL. Another potential endpoint worthy of consideration might be to measure the effectiveness of erythropoietic agents in maintaining the hemoglobin concentration between 11–12 g/dL for the longest duration during treatment. It would also be of interest to determine the relative cost effectiveness of early intervention followed by maintenance of hemoglobin levels within the clinically relevant target range, compared with late intervention. Such studies would provide useful information to healthcare policy makers as they consider how to best use existing resources while optimizing patient benefit.
The authors thank Christine Gatchalian, Ph.D., and Sadie Whittaker, Ph.D. for assistance with the preparation of the article.