Chapter  30:  Uses of Epoetin for Anemia in Oncology: Evidence Report/Technology Assessment Number 30

Anemia Primarily Due to Cancer Therapy and Anemia Due to Malignant Disease

The first group of patients are those being treated for malignancy with chemotherapy, radiation, or chemotherapy and radiation. The second group of patients comprises those who would be anemic whether or not they were receiving treatment for their malignancy. All patients in these studies had nonmyeloid hematologic malignancies (multiple myeloma, non-Hodgkin's lymphoma, chronic lymphocytic leukemia) or myelodysplastic syndrome (MDS).

Anemia Resulting from Bone Marrow Ablation Prior to Stem-Cell Transplantation (Allogeneic or Autologous)

Virtually all patients undergoing stem-cell transplantation require red blood cell transfusion, but epoetin may reduce the duration of anemia or the number of units transfused. Allogeneic and autologous stem-cell transplantation are reviewed separately.

Disruption of Hematopoiesis in Anemia

Anemia is a deficiency in the concentration of RBCs (also termed "erythrocytes") or Hb that occurs when the equilibrium between red cell loss and production is disturbed. The principal function of erythrocytes is to transport oxygen (bound to Hb) from the lungs, where the oxygen tension is high, to the organs and tissues, where it is low (Bunn, 1994a; Spivak, 1994). Thus, the adequacy of tissue oxygenation depends on sufficiency of the red cell mass. RBCs also transport carbon dioxide from the tissues back to the lungs so that it can be eliminated. When excess blood loss, decreased red cell survival (hemolysis), or decreased red cell production disrupts the normal equilibrium between RBC loss and production, anemia and tissue hypoxia result.

Normal hematopoiesis is regulated by erythropoietin, a glycoprotein hormone produced primarily in the kidney in adults; the other site of production, the liver, is the major production site in the fetus (Faulds and Sorkin, 1989). In a dose-dependent manner, erythropoietin stimulates the proliferation of committed erythroid progenitor cells (burst-forming and colony-forming units-erythroid [BFU-E and CFU-E]), maintains cell viability during erythroid differentiation, and thus participates in regulating development of mature erythrocytes (Bunn, 1994a; Faulds and Sorkin, 1989). The usual baseline concentration of endogenous serum erythropoietin in individuals with a normal hematocrit is approximately 4 to 30 mU/mL, depending on altitude (McEvoy, 1999; Spivak, 1994). Hypoxia, or decreased oxygen tension in the blood, is the stimulus for increased erythropoietin production and release (Bosi, Vannucchi, Grossi, et al., 1991). Most patients who are anemic will demonstrate a surge of erythropoietin production in response to the severity of the anemia (Bosi, Vannucchi, Grossi, et al., 1991). In patients with hypoxia, severe anemia from blood loss, or aplastic anemia, the erythropoietin concentration can increase up to 1,000 times greater than baseline (McEvoy, 1999).

Anemia is a common condition, with an incidence of 1.5 percent in the general population (Denton, Diamond, Matloff, et al., 1994). Some disorders that produce anemia, such as nutritional deficiency states and premenopausal menorrhagia, are uncomplicated and correctable with iron replacement. Other causes of anemia, including renal dysfunction, hyperviscosity, inflammation, sequestration, infection, neoplasia, and chemotherapy for cancer, may be associated with decreased erythropoietin production (Moliterno and Spivak, 1996).

Anemia in Cancer Patients

The NCI estimates that approximately 8.2 million Americans alive today have a history of cancer (American Cancer Society. 1999). Some of these individuals can be considered cured, whereas others still have evidence of cancer and may be undergoing treatments. About 1,221,800 new cancer cases are expected in the United States in 1999 (Landis, Murray, Bolden, et al., 1999).

Table 1. Grading systems for anemia

Table 1. Grading systems for anemia

	Toxicity grading system, g/dL hemoglobin
Severity	WHO	NCI and cooperative oncology groups1
Grade 0 (WNL) 2	>11.0	WNL
Grade 1 (mild)	9.5-10.9	10.0-WNL
Grade 2 (moderate)	8.0-9.4	8.0-10.0
Grade 3 (serious/severe)	6.5-7.9	6.5-7.9
Grade 4 (life-threatening)	<6.5	<6.5

1

Cooperative oncology groups: Eastern Cooperative Oncology Group; Southwest Oncology Group; Cancer and Leukemia Group B; Gynecologic Oncology Group.

2

WNL hemoglobin values are 12.0-16.0 g/dL for women and 14.0-18.0 for men.

Adapted from Groopman and Itri, 1999.

A recent literature review catalogued the incidence and severity of anemia in patients by type of malignancy, with information on chemotherapy treatments and regimens (Groopman and Itri, 1999). The incidence of Grade 3 or 4 anemia varied considerably. For example, the range was 10 to 80 percent in studies of patients with advanced colorectal, breast, or ovarian cancer. Overall, Grade 1 or 2 anemia was much more common than Grade 3 or 4. Another recent report found that, in ptients with nonmyeloid malignancies receiving cytotoxic chemotherapy, those with lymphomas, lung tumors, and ovarian or genitourinary tumors experienced the highest incidence (50 to 60 percent) of Grade 3 or 4 anemia (Ludwig and Fritz, 1998b). Patients treated with nephrotoxic agents such as cisplatin may be susceptible to anemia as a consequence of reduced endogenous production of erythropoietin (Wood and Hrushesky, 1995).

Anemia, regardless of its underlying cause, reduces the oxygen-carrying capacity of the blood (Armitage, 1998; Koeller, 1998; Lee, 1999a, 1999b; Ludwig and Fritz, 1998a; Moliterno and Spivak, 1996). Physiologic adaptive mechanisms such as changes in heart and respiration rates, cardiac output, venous return, peripheral resistance, and oxygen affinity of hemoglobin in the tissues generally compensate for anemia of milder severity and prevent the development of most symptoms. Whether the onset of anemia is gradual or rapid is likely to affect the ability of these adaptive mechanisms to compensate for the decline in Hb and to prevent development of symptoms.

Table 2. Signs and symptoms of anemia in cancer patients (more...)

Table 2. Signs and symptoms of anemia in cancer patients

Body System	Mild Anemia	Severe Anemia
Cardiovascular	Tachycardia Palpitations on exertion	Tachycardia Palpitations at rest Cardiac enlargement Systolic ejection murmur
Pulmonary	Dyspnea on exertion	Dyspnea at rest
Central nervous system		Dizziness and vertigo Headaches Irritability Impaired cognition Difficulty sleeping Difficulty concentrating
Gastrointestinal		Anorexia Nausea Indigestion
Genitourinary		Menstrual problems Male impotence Loss of libido
Skin		Pallor Low skin temperature
General	Mild fatigue	Severe fatigue Exercise intolerance

Adapted from Ludwig and Fritz, 1998b; Koeller 1998.

Differential Diagnosis and Treatment of Anemia in Cancer Patients

Anemia in cancer patients can be caused by the underlying disease, its treatment, or causes unrelated to the malignancy. Treatment of anemia requires a thorough differential diagnosis of the underlying cause or causes (Armitage, 1998). An evaluation may include some or all of the following measurements: Hb and hematocrit levels; RBC indices; reticulocyte counts; iron indices, including serum iron level, total iron-binding capacity (TIBC), percent transferrin saturation (TSAT), and serum ferritin; Coomb's test; vitamin B₁₂ and folate concentrations; and a test for occult blood in stool. Blood smears also can reveal characteristic profiles of various anemias associated with malignancy.

Mechanisms and Classification of Anemia

Different types of anemia exhibit characteristic changes in size and hemoglobin content of erythrocytes (Lee, 1999a). The initial classification into macrocytic, microcytic, and normocytic anemias is based on erythrocyte indices such as mean corpuscular volume (MCV). Additional indices such as mean corpuscular hemoglobin (MCH) and mean corpuscular hemoglobin concentration (MCHC) may be less useful, although readily available with most automated cell counters. Erythropoietic response and reticulocyte count can be useful to indicate whether the underlying disorder affects the bone marrow.

Loss of RBCs by internal hemorrhage often occurs in advanced stages of certain solid tumors (e.g., colorectal carcinoma or other gastrointestinal malignancies; some genitourinary tumors), and if undetected, almost always causes iron deficiency (Armitage, 1998; Bunn, 1994b; Lee, 1999a, 1999b, 1999c). Chronic blood loss is typically accompanied by a decrease in the size of RBCs (microcytic anemia, detectable by decreases in MCV) and in their hemoglobin content (hypochromic anemia, detectable by decreases in MCH). Iron deficiency in cancer patients can be unrelated to malignancy, for example, chronic bleeding resulting from peptic ulcer or menorrhagia or inadequate iron intake.

Two other causes of anemia, acute blood loss and hemolysis, are usually accompanied by an elevated number of reticulocytes (reticulocytosis) and little if any change in the size of RBCs (normocytic anemia). Hemolytic anemia in patients with malignancies may occur by either autoimmune or microangiopathic mechanisms (Armitage, 1998; Beguin, 1996; Frenkel, Bick, and Rutherford, 1996; Moliterno and Spivak, 1996; Rosse and Bunn, 1994). Autoimmune hemolytic anemia is relatively uncommon in cancer patients but occurs most often in those with chronic lymphocytic leukemia (CLL) or lymphoma, possibly as an adverse effect of fludarabine treatment. Microangiopathic hemolytic anemia (also referred to as thrombotic microangiopathy), which involves intravascular hemolysis caused by red cell fragmentation, is rarely caused by malignancy, except occasionally in patients with widely disseminated metastases (Foerster, 1999).

In cancer patients, several conditions can result in a normocytic anemia without reticulocytosis (Armitage, 1998; Beguin, 1996; Moliterno and Spivak, 1996). One example is overgrowth of the marrow with malignant cells; this occurs most often in leukemia, but also is possible in multiple myeloma, lymphoma, and prostate, breast, and small cell lung cancer (Abels, 1992; Frenkel, Bick, and Rutherford, 1996; Rappeport and Bunn, 1994). Myelofibrosis and other nonmalignant processes that replace the normal complement of marrow cells also are possible causes of normocytic anemia without reticulocytosis. Another cause is drug- or radiation-induced aplastic anemia, often associated with alkylating agents. Bone marrow biopsy may be necessary to determine the cause of normocytic anemia without reticulocytosis.

Anemia that results in larger than normal RBCs (macrocytic anemia) can occur as a result of nutritional deficiency or inadequate absorption of vitamin B₁₂ and/or folate, after treatment with certain antimetabolites used in cancer (e.g., methotrexate, mercaptopurine, cytosine arabinoside) or viral diseases (e.g., zidovudine), and in patients who drink alcohol (Abels, 1992; Armitage, 1998; Bunn, 1994b; Lee, 1999b; Moliterno and Spivak, 1996). In these instances, the anemia is termed "megaloblastic," is a consequence of impaired DNA synthesis, and results in the presence of cells (megaloblasts) with characteristic alterations of chromatin. Macrocytic anemia also may occur without the appearance of megaloblasts and without impaired DNA synthesis. Examples of such macrocytic anemias include those resulting from myelodysplastic syndromes and some hepatic diseases and in patients who have undergone splenectomy.

Anemia Due to Chronic Disease

Patients with malignant diseases, as well as those with many infectious and inflammatory diseases and in some instances those with trauma, may have "anemia of chronic disease" (Abels, 1992; Armitage, 1998; Beguin, 1996; Bunn, 1994b; Frenkel, Bick, and Rutherford, 1996; Henry 1996; Koeller, 1998; Means, 1999; Moliterno and Spivak, 1996). Anemia due to chronic disease may occur as either a microcytic or normocytic anemia and may be accompanied by a normal or reduced number of reticulocytes. The serum concentration of iron, total iron binding capacity, and transferrin saturation are generally decreased, whereas serum ferritin and bone marrow iron stores are normal or increased. The RBC life span is shortened. Elevated production of cytokines (possibly including interleukin-1 and interleukin-6 [IL-6], tumor necrosis factor, and interferons beta and gamma) that inhibit erythropoiesis and reduce endogenous production of erythropoietin may play a role in the etiology of anemia of chronic disease. Serum erythropoietin levels are lower than expected for the degree of anemia when compared with those measured in iron-deficient patients with an equivalent severity of anemia (Miller, Jones, Piantadosi, et al., 1990). This observation provides a rationale for use of erythropoietin to treat the anemia of chronic disease.

The myelodysplastic syndromes (MDSs) are clonal disorders of hematopoiesis characterized by cytopenias and dysfunctional blood cells (Lowenthal and Marsden, 1997). MDS may be associated with severe neutropenia, transfusion-dependent anemia, and, less commonly, transformation to acute leukemia (Stein, Abels, and Krantz, 1991). The French, American, British (FAB) classification of myelodysplastic syndromes includes five categories of anemias: refractory anemia (RA), refractory anemia with ringed sideroblasts (RARS), refractory anemia with excess of blasts (RAEB), refractory anemia with excess of blasts in transformation (RAEB-T), and chronic myelomonocytic leukemia (CMML) (Rappeport and Bunn, 1994). MDS patients do not in general have a deficiency of erythropoietin but do have an inappropriately low endogenous erythropoietic response to their anemia, consistent with anemia of chronic disease (Stein, Abels, and Krantz, 1991).

Anemia Due to Cancer Treatment

Patients with malignant diseases may experience a direct toxic effect of treatment on either the bone marrow stem-cell compartment or on the renal cells that produce erythropoietin (Abels, 1992; Armitage, 1998; Beguin, 1996; Moliterno and Spivak, 1996). In a dose-dependent manner, radiation and many chemotherapeutic agents may reduce the pool of stem cells, induce temporary or permanent damage to their capacity to proliferate and/or differentiate, or both. Although treatment-induced neutropenia and/or thrombocytopenia are more common and usually more severe than anemia, anemia severe enough to warrant transfusion also occurs (Aledort and Mohandas, 1996).

Red Blood Cell Transfusion

Red blood cell transfusion is the usual treatment for patients with anemia accompanied by clinically significant symptoms. It is also the most rapid means of raising the RBC count and is invariably effective in the absence of continued bleeding or ongoing hemolysis. Although the frequencies of adverse effects from RBC transfusions are small, they can be serious or life-threatening (Aledort and Mohandas, 1996; Ludwig and Fritz, 1998a; Schroeder, 1999; Valeri, Crowley, and Loscalzo, 1998;). The risks include hemolytic transfusion reactions, nonhemolytic febrile transfusion reactions, iron overload, hypervolemia, transmission of viral or (less often) bacterial infectious diseases, transfusion-related graft-versus-host disease (GVHD), transfusion-related acute lung injury, and immunosuppression. Recent estimates of the frequency for these and other adverse effects of RBC transfusion range from as much as 0.5 percent to 1 percent for circulatory overload or nonhemolytic febrile transfusion reactions to as little as 1 in 1,000,000 for bacterial contamination or 1 in 450,000 to 660,000 for transmission of HIV-1 (Simon, Alverson, AuBuchon, et al., 1998).

Epoetin

Recombinant human erythropoietin or "epoetin" was developed in the 1980s, after the human gene responsible for its production was cloned and expressed in vitro (Jacobs, Shoemaker, and Rudersdorf, 1985; Lin, Suggs, Lin, et al., 1985). The 165-amino acid mature recombinant protein is identical to the endogenous hormone with respect to its peptide sequence, and it has identical biologic activity (Amgen, Inc., 1999; Faulds and Sorkin, 1989; Ortho Biotech, Inc., 1999).

The recombinant epoetin has been produced in two forms: alfa and beta (Halstenson, Macres, Katz, et al., 1991). Each differs from the other and from the endogenous form principally in the nature and composition of the carbohydrate chains attached to the peptide (McEvoy, 1999). The two recombinant forms may differ in their pharmacokinetic properties (Halstenson, Macres, Katz, et al., 1991). However, only the alfa form has been approved for marketing in the United States by the U.S. Food and Drug Administration (FDA) (McEvoy, 1999).

Epoetin cannot be administered orally; it is administered either by subcutaneous (sc) or intravenous (iv) injection (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). In most studies on patients with malignancies, the drug has been administered as an iv bolus three times weekly. The drug also has been administered parenterally via other routes convenient to the clinical situation (e.g., via central lines used for infusion of chemotherapy). Also, the drug may be self-administered by the patient (Amgen, Inc., 1999; Ortho-Biotech, Inc., 1999).

Trials in patients with end-stage renal disease provide evidence that the drug is cleared more rapidly after intravenous than after subcutaneous administration (Albitar, Meulders, Hammoud, et al., 1995; Besarab, Flaherty, Erslev, et al., 1992; Canaud, Bennhold, Delons, et al., 1995; Kaufmann and Reda, 1996; Paganini, Eschbach, Lazarus, et al., 1995; Virot, Janin, Guillaumie, et al., 1996). Slower clearance suggests a longer duration of exposure to biologically effective concentrations at a given dose when the drug is administered subcutaneously rather than intravenously.

For cancer patients undergoing chemotherapy, the FDA-approved labeling states that the recommended starting dose is 150 U/kg sc three times weekly (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). If the hematocrit exceeds 40 percent (Hb >13.4 g/dL), the drug should be withheld until the hematocrit falls to 36 percent (Hb <13); the drug dose should be decreased by 25 percent when therapy is resumed (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). If the initial drug dose induces a very rapid response (e.g., an increase of four or more points in any 2-week period), the dose should be reduced. If hematocrit response is not satisfactory after 8 weeks, the dose can be increased up to 300 U/kg three times weekly (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). If patients fail to respond to 300 U/kg three times weekly, it is unlikely that they will respond to higher doses (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999).

Epoetin alfa is FDA-approved in the United States and marketed under the trade names Epogen^® (Amgen, Inc., Thousand Oaks, CA) and Procrit^® (Ortho Biotech, Inc., Raritan, NJ). Both trade products are derived from the same source and are identical in composition (McEvoy, 1999).

Table 3. FDA-approved uses of epoetin alfa and dosing (more...)

Table 3. FDA-approved uses of epoetin alfa and dosing recommendations¹

FDA-Approved Labeling	Starting Dose	Adjustment 2	Comment
"Indicated in the treatment of anemia associated with chronic renal failure, including patients on dialysis (end-stage renal disease) and patients not on dialysis"	50-100 U/kg 3 times weekly (iv or sc)	Reduce dose if: Hct approaches 36% or Hct increases >4 points in any 2-week period. Increase dose if: Hct does not increase by 5-6 points after 8 weeks of therapy and Hct is below the suggested range.	Predialysis patients with symptomatic anemia considered for therapy should have a hematocrit less than 30%. Transferrin saturation should be at least 20% and ferritin should be at least 100 ng/mL. Hct should be monitored twice weekly until it has stabilized in the target range and the maintenance dose has been established. Suggested target hematocrit range: 30-36%; however, the labeling states that "at the physician's discretion, the suggested target hematocrit range may be expanded to achieve maximal patient benefit."
"Indicated for the treatment of anemia related to therapy with zidovudine in HIV-infected patients"	100 U/kg 3 times weekly (iv or sc) for 8 weeks	After attainment of the desired response (i.e., reduced transfusion requirements or increased Hct), the dose should be titrated to maintain the response based on factors such as the change in zidovudine dosage, and the presence of intercurrent infections or inflammatory episodes. If the Hct exceeds 40%, the dose should be discontinued until the Hct drops to 36%. The dose should be reduced by 25% when the treatment is resumed and then titrated to maintain the desired Hct.	Endogenous serum erythropoietin concentration should be <500 mU/mL. Weekly total zidovudine dose should be <4.2 g Hct should be monitored once weekly until it has stabilized in the target range and then measured periodically thereafter.
"Indicated in the treatment of anemia in patients with non-myeloid malignancies where anemia is due to the effect of concomitantly administered chemotherapy.	150 U/kg sc 3 times weekly	If Hct exceeds 40%, the drug should be withheld until the Hct falls to 36%; the drug dose should be decreased by 25% when therapy is resumed. If the initial drug dose induces a very rapid response (e.g., an increase of 4 or more points in any 2-week period), the dose should be reduced. If Hct response is not satisfactory after 8 weeks, the dose can be increased up to 300 U/kg 3 times weekly.	Patients must be receiving the concomitant chemotherapy for a minimum of 2 months. Therapy not indicated for the treatment of anemia in cancer patients due to other factors such as iron or folate deficiencies, hemolysis, or GI bleeding; these should be managed appropriately. Although no specific serum erythropoietin concentration can be stipulated above which patients would be unlikely to respond, treatment of patients with "grossly elevated" serum concentrations (e.g., >200 mU/mL) is not recommended. If patients fail to respond to 300 U/kg 3 times weekly, it is unlikely that they will respond to higher doses. Hct should be monitored once weekly until it has stabilized in the target range and then measured periodically thereafter.
"Indicated in the treatment of anemic patients(hemoglobin >10 to <13 g/dL) scheduled to undergo elective, noncardiac, nonvascular surgery to reduce the need for allogeneic blood transfusions… indicated for patients at high risk for perioperative transfusions with significant, anticipated blood loss."	300 U/kg daily sc for 10 days prior to surgery, the day of surgery, and 4 days after surgery. Also, the drug may be given at a dosage of 600 U/kg sc once weekly at 21, 14, and 7 days before surgery, plus a fourth dose the day of surgery.	N/A	Not indicated for anemic patients who are willing to donate autologous blood. The safety of the perioperative use of the drug has been studied only in patients who are receiving anticoagulant prophylaxis. All patients should receive adequate iron supplementation; iron therapy should be initiated no later than the beginning of epoetin treatment and should continue throughout the therapy course.

1

All information from Ortho Biotech and Amgen FDA-approved labeling.

2

Maintenance doses should be titrated to response in all cases.

• The treatment of anemia associated with chronic renal failure (including patients on dialysis [end-stage renal disease] and patients not on dialysis).

• The treatment of anemia related to therapy with zidovudine in HIV-infected patients.

• The treatment of anemia in patients with nonmyeloid malignancies where anemia is a result of the effect of concomitantly administered chemotherapy.

• The treatment of anemic patients (Hb >10 to <13 g/dL) scheduled to undergo elective, noncardiac, nonvascular surgery to reduce the need for allogeneic blood transfusions and of patients at high risk for perioperative transfusions with significant, anticipated blood loss (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999).

Amgen, Inc. is licensed to market Epogen® for the treatment of dialysis patients with anemia of ESRD; Ortho Biotech, Inc. is licensed to market Procrit® for the treatment of all other FDA-approved uses, including the treatment of anemia in predialysis patients.

The drug has been used off-label for a variety of other uses, including anemia of prematurity (Faulds and Sorkin, 1989), anemia of chronic disease (e.g., rheumatoid arthritis) (Faulds and Sorkin, 1989; Pincus, Olsen, Russell, et al., 1990; Salvarani, Lasagni, Casali, et al., 1991; Vreugdenhil and Swaak, 1990;), anemia of myelodysplastic syndromes (Adamson, Schuster, Allen, et al., 1992; Ludwig, Fritz, Leitgeib, et al., 1993; Stein, Abels, and Krantz, 1991;), sickle-cell anemia (Faulds and Sorkin, 1989), anemia of multiple myeloma (Ludwig, Fritz, Leitgeib, et al., 1993); and anemia following high-dose chemotherapy with stem-cell support (Henry, 1998).

In patients with chronic renal failure receiving epoetin therapy, the most common adverse effect of the drug is hypertension, occurring in approximately 24 percent of such patients (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). In other patient populations, hypertension occurs rarely (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). Other more serious adverse effects of the drug include hypertensive encephalopathy, seizures, and thrombotic/vascular events (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999); however, these more serious events also appear to occur most often in patients being treated with epoetin for anemia resulting from renal failure (Amgen, Inc., 1999; Markham and Bryson, 1995; Ortho Biotech, Inc., 1999).

Patients receiving epoetin for anemia related to conditions other than chronic renal failure (e.g., zidovudine treatment or other chemotherapy) demonstrate no serious adverse effects directly attributable to drug therapy (Amgen, Inc., 1999; Markham and Bryson, 1995; Ortho Biotech, Inc., 1999). Differences in the severity of adverse effects in different patient populations, or even within populations, may be attributable to differences in the underlying disease states and/or comorbidities, such as the presence of hypertension or cardiac disease (Beguin, 1998b; de Andrade, Frei, and Guilfoyle, 1999). Several large (i.e., study populations of over 2,000 patients), community-based, single-arm studies of epoetin in anemic cancer patients have described no unexpected adverse effects in this population (Demetri, Kris, Wade, et al., 1998; Glaspy, Bukowski, Steinberg, et al., 1997). The most common adverse events noted were fever, decreased white cell count, nausea, vomiting, and asthenia, all occurring in fewer than 10 percent of patients (Demetri, Kris, Wade, et al., 1998; Glaspy, Bukowski, Steinberg, et al., 1997). One study reported disease progression occurring in 23 percent of patients; however, this was not directly attributable to epoetin treatment (Glaspy, Bukowski, Steinberg, et al., 1997). According to one set of investigators, "The incidence and type of adverse events experienced by the patients in this trial do not appear to differ from patients with cancer, appear to be consistent with the underlying disease state and the chemotherapeutic regimens, and are consistent with those observed in the double-blind, placebo-controlled trials" (Demetri, Kris, Wade, et al., 1998).

The FDA-approved product information for epoetin alfa states that the safety and efficacy of the drug have not been established in patients with a known history of a seizure disorder or underlying hematologic condition (e.g., sickle cell anemia, myelodysplastic syndromes, hypercoagulable disorders) (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). Use of epoetin alfa is contraindicated in patients with (1) uncontrolled hypertension, (2) known sensitivity to mammalian cell-derived products, or (3) known hypersensitivity to human albumin (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999).

If patients fail to respond or to maintain a response to epoetin therapy in the recommended dosing range, the following possible causes of delayed or diminished response should be considered: iron deficiency, underlying infectious inflammatory or malignant process, occult blood loss, underlying hematologic diseases, folic acid or vitamin B₁₂ deficiency, hemolysis, aluminum intoxication, or osteitis fibrosa cystica (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). Although no specific serum erythropoietin concentration can be stipulated above which patients would be unlikely to respond, treatment of patients with "grossly elevated" serum concentrations (e.g., >200 mU/mL) is not recommended.

During therapy with epoetin alfa, iron stores are quickly mobilized and utilized as erythrocytes are produced. Absolute or functional iron deficiency may develop (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). Functional iron deficiency is defined as insufficient iron to support erythropoiesis despite normal iron stores and is generally characterized by normal or elevated ferritin levels in the presence of low transferrin saturation (serum iron concentration divided by the iron-binding capacity). Functional iron deficiency is presumably caused by the inability of the body to mobilize iron stores rapidly enough to support increased erythropoiesis (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). In patients receiving epoetin therapy, transferrin saturation should be at least 20 percent and ferritin levels should be at least 100 ng/mL (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). Transferrin saturation and ferritin concentrations should be determined prior to and also during therapy (Amgen, Inc., 1999; Ortho Biotech, Inc., 1999). However, both serum ferritin and transferrin saturation have limitations as diagnostic tests for functional iron deficiency, and other tests are being investigated (Glaspy and Cavill, 1999).

Supplemental iron is recommended to achieve and maintain target hematocrit levels in patients with chronic renal failure, and intravenous administration may be necessary (National Kidney Foundation. 1997). Uniform recommendations regarding iron supplementation for cancer patients with anemia do not exist. Not all trials of epoetin therapy in cancer patients have supplemented patients with iron, and those that have generally administered iron orally. Controlled trials are currently underway to evaluate the effect of iv iron on the response of cancer patients to epoetin therapy (Glaspy and Cavill, 1999).

Cancer-Related Fatigue

Fatigue is the most frequently reported symptom of cancer and cancer treatment in patient surveys (Portenoy, Thaler, Kornblith, et al., 1994; Vogelzang, Breitbart, Cella, et al., 1997) and is often more distressing to patients than is pain (Nail and Winningham, 1995; Vogelzang, Breitbart, Cella, et al., 1997). Although there is no universally accepted definition, fatigue has been described as "diminished energy, or an increased need to rest that is disproportionate to any recent change in activity" that results in a "sustained deterioration in the usual ability to perform either physical or intellectual activities" (Miaskowski and Portenoy, 1998). Cancer-related fatigue was recently accepted as a diagnosis in the International Classification of Diseases, 10^th Revision, Clinical Modification (Portenoy and Itri, 1999).

A recent population-based survey reports that cancer-related fatigue is common and also suggests that oncologists underestimate its impact on patients (Vogelzang, Breitbart, Cella, et al., 1997). Of 419 patients who had undergone either chemotherapy (35 percent), radiotherapy (39 percent), or both (24 percent), 78 percent reported experiencing fatigue during the course of their disease and treatment; and 71 percent reported that fatigue somewhat or significantly (32 percent) had an impact on their daily routine. Although 61 percent of patients reported that their daily lives were more adversely affected by fatigue than by cancer-related pain, only 37 percent of oncologists held this view. Another study (Irvine, Vincent, Graydon, et al., 1994), which compared 101 cancer patients with 53 healthy individuals, reported that the proportion of cancer patients experiencing fatigue increased significantly from 39 percent at baseline to 61 percent with treatment. A recent study comparing cancer inpatients undergoing palliative care with age and sex-matched volunteers reported that 75 percent of the cancer patients experienced fatigue of greater intensity than that experienced by 95 percent of the controls (Stone, Hardy, Broadley, et al., 1999).

In cancer patients, the etiology of fatigue is multifactorial and has both physiologic and psychosocial dimensions (Winningham, Nail, Burke, et al., 1994). Causes include anemia, metabolic disturbances, disease progression, and adverse treatment-related effects. Disease stage and status (remission versus stable versus progressive; localized versus metastatic), type (hematologic versus solid tumor), and particular symptoms can be linked not only to anemia, but also to pain, loss of muscle function, loss of appetite, and fatigue (Langer, 1997; Winningham, Nail, Burke, et al., 1994). In a study of correlates of fatigue in cancer patients receiving chemotherapy or radiotherapy, changes in weight and white blood cell count were significantly correlated with fatigue whereas changes in hematocrit and Hb were not (Irvine, Vincent, Graydon, et al., 1994).

Anemia results in a reduction in the oxygen-carrying capacity of blood. Reduced delivery of oxygen to muscle tissue reduces work capacity and exercise tolerance and may result in perceived weakness. Brain function may also be affected, with possible changes in mood and perception. Activities may be curtailed because of lack of physical and mental energy.

Reduced availability of substrate for energy metabolism, or accumulation of abnormal substances that impair intermediary metabolism, could affect normal physiologic and/or cognitive function, resulting in symptoms of fatigue (Miaskowski and Portenoy, 1998). For example, a hypermetabolic state that can accompany tumor growth may result in increased substrate need for energy metabolism. Nausea and decreased appetite frequently accompany cancer therapy. Inadequate nutrition, as well as poor nutritional uptake, may lead to decreased substrate availability. Cancer cachexia, thought to be related to a change in energy transformation processes, results in weight loss and fatigue.

Abnormally increased production of the cytokine IL-6 has been associated with neoplasms and with tissue injury produced by radiation therapy (Miaskowski and Portenoy, 1998). In clinical trials, exogenous IL-6 has been associated with fatigue (Gordon, Nemunaitis, Hoffman, et al., 1995; Sosman, Aronson, Sznol, et al., 1997; Weber, Yang, Topalian, et al., 1993). Endogenous production of IL-6 can be stimulated by interferon alfa immunotherapy, which is also associated with fatigue (Jones, Wadler, and Hupart, 1998). Evidence suggests that both IL-6 and interferon alpha can affect the function of thyroid cells; thus, endocrine disease may explain some aspects of fatigue experienced by cancer patients (Jones, Wadler, and Hupart, 1998).

Cancer patients undergoing radiation therapy appear to experience a transient decline in neuromuscular efficiency that is unrelated to cardiovascular or psychologic status (Monga, Jaweed, Kerrigan, et al., 1997). Neurologic diseases, such as the peripheral neuropathy that can accompany some cancer chemotherapy (e.g., with vinca alkaloids) and that is sometimes seen in multiple myeloma patients, may also result in decreases in the efficiency of neuromuscular functioning.

Metabolic disturbances and centrally acting drugs may also cause sleep or arousal disorders, which could contribute to fatigue. Affective disorders such as depression can also be associated with difficulty sleeping and with fatigue (Miaskowski and Portenoy, 1998).

Because the etiology of cancer-related fatigue is multifactorial, there are a variety of avenues that may be useful in its management. Correction of anemia, either by RBC transfusion or after stimulation of RBC production, can relieve the portion of a patient's fatigue that is caused by anemia but does not address other causes of fatigue. This suggests that multimodality approaches should be explored and evaluated for effectiveness.

Pharmacologic approaches may be used to alleviate fatigue. First, elimination of nonessential centrally acting drugs, dose reductions, or alteration of dosing regimens may be beneficial. Antidepressants are indicated when there is a clear diagnosis of depression; patients treated successfully for major depression often report reduction in fatigue (Miaskowski and Portenoy, 1998).

Behavioral modifications may improve symptoms of fatigue. Educating patients regarding potential methods of reducing fatigue may relieve anxiety and provide reassurance and an element of control. Exercise has been postulated to reduce fatigue. Studies of exercise interventions during or after chemotherapy programs suggest that aerobic exercise reduces fatigue in study populations compared with that in control populations (Dimeo, Stieglitz, Novelli-Fischer, et al., 1999; Dimeo, Tilmann, Bertz, et al., 1997). Efforts to maintain nutritional status may avoid or alleviate metabolic dysfunction and prevent some cancer-associated fatigue (Kalman and Villani, 1997). Maintenance of regular sleep patterns and reduced activity schedules may also have a positive impact.

Instruments Used to Measure the Effects of Anemia Treatment on Quality of Life

Quality-of-life instruments may be global or disease specific. Global instruments are intended for use across various chronic disease populations and permit comparison of quality-of-life outcomes among interventions and diseases. Today, the Short Form 36-item Health Survey (SF-36) is the most commonly used global measure of health status. But global instruments may fail to address issues that are important to patients and relevant to treatment of specific diseases. In oncology trials, there has been an attempt to balance the tradeoffs between global and disease specific measures by combining a cancer core instrument with a module specific to a particular malignancy or treatment (Beitz, Gnecco, and Justice, 1996). The primary cancer core instruments for use in clinical trials are the European Organization for Research and Treatment of Cancer (EORTC) core Quality of Life Questionnaire (QLQ-C30) and the general fanctional assessment of cancer therapy scale. Each of these core instruments can be supplemented with more specific modules, for example the EORTC QLQ-LC13 for lung cancer or the FACT-Ovarian.

Clinical trials of epoetin initially used unidimensional linear analog self-assessment scales (LASA) to measure changes in the quality of life related to anemia. Several studies of epoetin have used a three-item LASA in which patients rate their perceived energy level, ability to perform activities of daily living, and overall quality of life on a 100 mm-ruled line representing a continuum from lowest to highest possible assessment for each item (Abels, 1993; Demetri, Kris, Wade, et al., 1998; Glaspy, Bukowski, Steinberg, et al., 1997; Littlewood, Bajetta, Cella, et al., 1999). The reference cited by these investigators for the 3-item LASA is an earlier study of a 21-item LASA (Gough, Furnival, Schilder, et al., 1983) that was compared with other methods of quality-of-life assessment in patients with advanced cancer, but was not formally validated for reliability and validity. Nor is it clear that the wording for the three-item LASA was identical to that of items in the earlier study. An unvalidated 10-item LASA that addresses physical, functional, emotional, and social domains has also been used (Kurz, Marth, Windbichler, et al., 1997; Leitgeb, Pecherstorfer, Fritz, et al., 1994; Ludwig, Sundal, Pecherstorfer, et al., 1995). Limitations of these unidimensional instruments are the lack of validation and the inability to assess the relationship between multiple dimensions of quality of life or to directly compare scales using different questions.

More recent trials of epoetin use a multidimensional instrument, the Functional Assessment of Cancer Therapy-Anemia (FACT-An) (Cella, 1997). The core of the FACT-An is the Functional Assessment of Cancer Therapy-General (FACT-G), which contains 29 questions that can be used to generate subscale scores regarding physical, functional, emotional, and social well-being, as well as satisfaction with the patient-physician relationship (Cella, Tulsky, Gray, et al., 1993). Data from 1,172 cancer patients who answered the FACT-G questionnaire indicated that fatigue was the symptom most often reported (73 percent). As a result, two additional subscales were produced to assess fatigue and anemia in cancer patients. The FACT-Fatigue (FACT-F) consists of a fatigue-specific subscale of 13 items that was added to the FACT-G. The FACT-An was produced by adding to the FACT-F seven nonfatigue items relevant to anemia in cancer patients. Because of their recent development, experience with FACT-F and FACT-An is more limited than experience with the EORTC QLQ-30 or FACT-G.

The EORTC QLQ-C30 was developed for use in international clinical trials (Aaronson, Ahmedzai, Bergman, et al., 1993). This questionnaire incorporates five functional scales (physical, daily activity, cognitive, emotional, and social), three symptom scales (fatigue, pain, and nausea and vomiting), and a global health and quality-of-life scale. The EORTC QLQ-C30 has been shown to correlate well with items that examine similar dimensions in the SF-36 (Apolone, Filiberti, Cifani, et al., 1998; Chie, Huang, Chen, et al., 1999).

The FACT-G was validated in a population of 854 cancer patients (Cella, Tulsky, Gray, et al., 1993). FACT-G scores correlated well with other similar measures completed at the same time (e.g., r=0.79 for the Functional Living Index-Cancer; r=−0.52 for the Eastern Cooperative Oncology Group [ECOG] performance score). Test-retest correlation coefficients were all above 0.80. FACT-G physical, functional, and total scores differentiated patients by cancer stage (p<0.01). The FACT-G has been compared with the EORTC QLQ-C30 (Kemmler, Holzner, Kapp, et al., 1999). Correlations between corresponding subscales were best for the physical, functional, and emotional subscales (r=0.48 to 0.66) but lower for the social subscales (r=0.14). Correlations between the FACT-F and FACT-An and more widely used quality-of-life assessment instruments such as the SF-36 have not yet been reported.

The FACT-An and FACT-F questionnaires were validated on a sample of 50 patients with a variety of solid and hematologic malignancies, who had Hb levels ranging from 7 to 15.9 g/dL and were rated at different performance levels (Cella, 1997; Yellen, Cella, Webster, et al., 1997). Test-retest reliability was 0.87 for both, and internal consistency was high for the Fatigue subscale (alpha range, 0.93 to 0.95) but lower for the nonfatigue FACT-An subscale (0.59 to 0.70). All FACT scales were significantly related with other known measures of fatigue: the Profile of Mood States (POMS) Vigor and Fatigue subscales and the Piper Fatigue Scale. The FACT-An has been translated into other languages, and studies are underway to validate the instrument in larger populations and evaluate the effect of interventions to reverse fatigue caused by anemia.

Reporting the Results of Quality-of-Life Assessments

The results of quality-of-life studies should be translated into clinically meaningful terms; that is, what is the smallest change in a quality-of-life score that patients perceive as beneficial?² Statistical significance (p value) of the difference between treatment and control groups with respect to the magnitude of change in raw scores on quality-of-life instruments indicates only the likelihood of the results occurring by chance in the study population. It is more difficult to interpret the clinical significance of the results. Calculation of an effect size expresses quality-of-life change in the context of the variation within the population, and there are conventions for defining an effect size as small, moderate, or large. A clearer picture emerges when the magnitude of quality-of-life changes can be interpreted in comparison to the magnitude of more familiar clinical changes or results.

Effect size, calculated as the mean change divided by the standard deviation of the baseline scores, indicates what proportion of the background variation the change represents. Effect sizes of 0.2 to about 0.5 are considered to be small, 0.5 to 0.8 to be moderate, and greater than 0.8 to be large (Cohen, 1977; Kazis, Anderson, and Meenan, 1989). For example, in a validation study for the FACT-An scale, the authors observed a mean difference of 17.4 points between individuals with Hb levels of 11 to 12.99 g/dL and > 13 g/dL and a mean difference of 8.6 points between individuals with levels <11 g/dL and 11 to 12.99 g/dL (Yellen, Cella, Webster, et al., 1997). These represent moderate (0.75) and small (0.32) effect sizes, respectively. In a controlled intervention study, the effect size in the treatment group can be compared with the effect size in the control group, or the effect size can be calculated from the difference in the mean change scores between the two study arms.

One approach to interpreting the clinical significance of the magnitude of effect is to "anchor" quality-of-life score changes to other clinical changes or results for which the clinical significance of the magnitude of change has been established. For example, one could relate quality-of-life score changes to the differences observed between patients with ECOG performance scores that differ by one unit. Mean FACT-An total and subscale scores were significantly different for patients with an ECOG performance scale rating of 0 versus patients with a performance scale rating of 2 or 3 (combined) and were significantly different for patients with a performance scale rating of 1 versus 2 or 3. This suggests that the FACT-An can differentiate between these performance score categories. In addition, FACT-G (only) was able to differentiate between patients with performance scores of 0 versus 1 (Yellen, Cella, Webster, et al., 1997). The most commonly reported anchoring method involves relating changes in disease-specific quality-of-life scores to the change in a patient-reported global quality-of-life score that is relevant to the study (Lydick and Epstein, 1993). In this way, the more informative properties of the multidimensional quality-of-life instrument are linked to a more interpretable single-item question. Determining the clinical significance of the magnitude of quality-of-life score changes obtained from a recently developed disease-specific instrument such as the FACT-An may benefit from being anchored to a more familiar global measurement.

It is possible that the clinical significance of a given magnitude of change depends on the baseline value. For example, a small change could be more clinically significant for a patient with a low baseline quality-of-life score than for a patient with a high baseline. Additionally, changes in one direction could have more clinical significance than changes of equal magnitude in the opposite direction. However, this type of information is not currently available and may depend upon the quality-of-life scale.

Assessment of Quality of Life in Clinical Trials

Several uncontrolled clinical studies of epoetin in cancer patients report statistically significant and favorable changes in measures of quality of life after epoetin treatment compared with baseline. However, factors other than epoetin intervention may affect outcomes; for example, tumor stage and progression, effects of cancer therapy, and changes in cancer therapy regimen. Comparison with historical or prospective but nonrandomized control groups may suffer from selection bias. Randomized controlled trials of adequate power are necessary to determine the true effects of an intervention on quality of life (Leidy, Revicki, and Geneste, 1999). In randomized trials, other factors that can affect cancer patients' quality of life are randomly and evenly distributed among treatment arms.

Quality-of-life results can also be affected by numerous details of the administration of the quality-of-life instrument and by the interaction between the physician and patient. Therefore, trials that assess quality-of-life endpoints should also incorporate specific design features related to the administration of quality-of-life instruments and the analysis and interpretation of results. The FDA recently outlined key features that should be addressed in designing and executing randomized controlled trials to assess quality-of-life endpoints for oncology drugs (Beitz, 1999; Beitz, Gnecco, and Justice, 1996); as summarized below:

• Rigorously validated instruments that permit cross-study comparisons should be used.

• Double-blinding is preferred; if it is not feasible, study personnel involved in the quality-of-life assessment should be blinded to the patients' treatment assignments and responses to treatment.

• The protocol should prospectively identify the quality-of-life outcomes to be measured as evidence of effectiveness, the critical points in time for measurement, and the minimum differences in quality-of-life scores to be considered clinically significant.

• The logistics of questionnaire administration should be handled to minimize the impact on the integrity of the quality-of-life assessment. Feedback from the investigator, treating physician, or staff that affects the patients' sense of well-being is a major source of bias in open-label trials and should be avoided. Ideally, the instrument should be administered prior to discussions with the physician (or other health care provider) as to treatment response, adverse events, or other matters that could affect patients' responses to the quality-of-life questionnaire.

• The study protocol should include a detailed schema and rationale for the time intervals at which the instrument will be administered. The training for staff administering the instrument should be described, with similar information if the instrument is self-administered by the patient.

• There should be a detailed plan for preventing missing data, investigating the pattern of missing data, and addressing missing data in the analysis. If data are missing nonrandomly, serious bias can result. For example, patients with missing questionnaires are frequently the sickest patients or those least responsive to therapy. But failure to respond to specific items in a questionnaire also raises concerns.

The largest published trials of epoetin therapy in oncology patients are two uncontrolled community studies (CSs) of patients undergoing chemotherapy reported by Glaspy, Bukowski, Steinberg, et al. (1997) and Demetri, Kris, Wade, et al., (1998), referred to here as CS-1 and CS-2, respectively.³ Because of their size (over 2,000 patients enrolled in each trial), community-based setting, and measurement of quality-of-life outcomes, the results from these single-arm trials are widely cited. To date, the community studies are the largest source of evidence on quality-of-life outcomes. Cleeland, Demetri, Glaspy, et al. (1999) conducted a multivariate regression analysis of the CS-1 and CS-2 data to determine, for each study, the association between change in Hb level (baseline to final value at study end or last monthly measurement before patient withdrawal) and change in quality-of-life (baseline to final value at study end or patient withdrawal) while adjusting for factors other than epoetin that might affect changes in both Hb levels and quality-of-life scores. Additional data imputation models were developed in an effort to adjust for missing quality-of-life data. This study was supported by Ortho Biotech, and the company supplied additional detail for evaluation (Finkelstein, Berndt, and Crimieux, unpublished manuscript, 1998).

We will summarize the results of the community studies and the finding of the multivariate analysis by Cleeland, Demetri, Glaspy, et al. (1999). Overall, these reports showed an association between increases in Hb and increases in quality-of-life scores. An equally important finding was that tumor response also had an independent effect on changes in quality-of-life scores. Given the biologic mechanisms of anemia in cancer, tumor response would be expected to have an independent effect on Hb, resulting from both the underlying disease and related changes in treatment. These reports, however, did not establish a causal relationship between epoetin treatment and improvement in quality of life. Moreover, a substantial amount of data on tumor response and quality of life are missing in the community studies. In the analysis, the values for missing tumor response data were assumed rather than omit substantial numbers of patients, and imputation methods were applied to the missing quality-of-life data. However, there is potential for significant bias in the results of the analysis. Finally, data collected in uncontrolled community studies is subject to bias resulting from inadequate control over the circumstances under which the quality-of-life assessment instrument is administered. Without controls, the influence of the expectations of physicians and patients may overestimate the magnitude of effect reported.

Summary of Community Studies

Table 4. Comparison of two single-arm trials of epoetin (more...)

Table 4. Comparison of two single-arm trials of epoetin and quality of life

Authors/Year	CS-1 1	CS-2 2
Number of patients enrolled	2,342	2,370
Study period	4 months	2-4 months
Study definition of anemia	None; (8.2% of patients had initial Hb >11)	Hb <11; eligibility restricted to anemic patients
Study definition of response	None	>2 g/dL increase in Hb or Hb >12 or both
Diseases included	Stem-cell therapy and transplant patients not excluded	Stem-cell therapy and transplant patients excluded
Prior radiation treatment	No information	Nearly 90% of patients received no prior radiation
Number of patients who left the trial prior to completing the protocol (percent of enrolled)	983 (42.0%)	1,003 (42.3%)
Number of patients evaluable for hematologic outcomes (percent of enrolled)	2,019 (86.2%)	2,237 (94.4%)
Number of patients evaluable for quality-of-life outcomes (percent of enrolled)	1,498 (64.0%; LASA)	1,761 (74.3%; LASA) 1,579 (66.6%; FACT-An)
Number of patients with pretreatment tumor status data (percent of enrolled)	None	1,368 (57.5%) 3
Number of patients with post-treatment tumor status data (percent of enrolled)	759 (32.4%) collected retrospectively 3	2,117 (89.3%) collected prospectively 3

1

Glaspy, Bukowski, Steinberg, et al., 1997.

2

Demetri, Kris, Wade, et al., 1998.

3

Tumor status/response classified into categories of complete response, partial response, stable disease, progressive disease

More than 2,000 patients were evaluable for hematologic and transfusion outcomes in each trial, 86 percent of enrolled patients in CS-1 and 94 percent in CS-2. When baseline Hb levels were compared with end of study results in CS-1, there was a 1.8 g/dL average increase (p<0.001), and 53 percent of patients achieved an Hb increase of 2.0 g/dL or better. In CS-2, the average increase was 2.0 g/dL (p<0.001), and 61 percent of patients met the study criteria for response. In CS-1, epoetin therapy was associated with an approximately 50 percent decrease both in the proportion of patients who required transfusion and in the number of units of RBCs transfused per patient. In CS-2, the percentage of patients who required transfusion decreased from 29 percent at baseline to 5 percent at month 4, and the mean number of units transfused per patient decreased from 1 to 0.2.

Much smaller percentages of enrolled patients were evaluable for quality-of-life outcomes than for hematologic outcomes. The LASA scales for energy, daily activity, and overall well-being were used in both trials; data were available for analysis for 1,498 (64 percent) patients in CS-1 and 1,761 (74 percent) patients in CS-2. The Functional Assessment of Cancer Therapy-Anemia was used only in CS-2; data on 1,579 (67 percent) of patients were available for analysis. Neither report described protocols to control for systematic bias arising from the timing, circumstances, interactions, or expectations related to the administration of the quality-of-life instruments. Univariate analysis in both trials showed significant, positive associations between increases in Hb levels and increases in all quality-of-life scores. Stratification by end-of-study tumor response status in CS-2 indicated that the association between increases in Hb levels and improvement in quality-of-life scores disappeared when the tumor progressed. LASA scores significantly improved over baseline; analysis of available data found a 30 to 50 percent change in each quality-of-life measure, with a greater than 4 g/dL increase in Hb. By FACT-An, responders to epoetin had a significant improvement in quality-of-life scores; the improvement was much less for nonresponders.

Data on posttreatment tumor response were collected prospectively only in CS-2 and were available for 89 percent of enrolled patients. In CS-1, posttreatment tumor response data were collected retrospectively, but were obtained for only 32 percent of enrolled patients. No pre-treatment tumor status data was collected in CS-1; in CS-2, baseline data were available only for 1,368 (58 percent) of enrolled patients. The missing CS-2 baseline data severely limit the assessment of the effects of change in Hb on changes in quality-of-life scores independent of the effects of baseline tumor status or change in tumor status during treatment.

Multivariate Regression Analysis and Imputation for Missing Data

Cleeland, Demetri, Glaspy, et al. (1999) conducted a multivariate regression analysis of the CS-1 and CS-2 data to determine, for each study, the association between change in Hb level (baseline to final value at study end or last monthly measurement before patient withdrawal) and change in quality-of-life (baseline to final value at study end or patient withdrawal), while attempting to control for factors other than epoetin that might affect both Hb levels and quality-of-life scores. The analysis also attempted to determine the Hb level at which the greatest improvement in quality of life, if any, is achieved. The analysis focused on the relationship between Hb change and change in quality of life score but did not attempt to relate causes of Hb change (i.e., epoetin, transfusion, chemotherapy, or disease) to changes in quality of life.

Variables selected for the analysis included baseline tumor status and change in tumor response, data collected prospectively only in CS-2. The analysis identified change in tumor status to progressive disease and progressive disease at baseline as confounders of the association between change in Hb level and change in quality of life. This is not surprising, since disease status and/or disease treatment are related to Hb level and can independently affect quality of life. In a study of high-dose versus low-dose epoetin treatment of patients with advanced gastrointestinal cancer, Glimelius, Linne, Hoffman, et al. (1998) found that tumor response was as important for improvements in EORTC QLQ-C30 scores as an increase in Hb. Findings reported in the unpublished manuscript by Finkelstein, Berndt, and Crimieux (1998) illustrated a similar effect observed in CS-2. Change in tumor status from stable or in remission at baseline to progressive was associated with a decrease in the LASA overall quality-of-life score of 12 units; for the FACT-An, the corresponding decrease was 15 units. In comparison, an increase in Hb from lower than 7.5 g/dL to 14 g/dL was associated with a 21-unit increase in the LASA score. Interestingly, the authors did not explore the potential effect of a change in tumor status from progressive to stable/remission, which could reasonably be expected to be associated with Hb change and to have a positive, independent effect on quality of life. Moreover, because 47 percent of patients in CS-2 did not report baseline data on tumor status, the authors assumed nonprogressive disease at baseline for the patients with missing data. The analysis reported a significant, independent association between the change in Hb and the change in quality-of-life scores, suggesting that effectors of Hb change might also effect changes in quality of life. The analysis also suggested that the greatest increase in quality of life was associated with an increase in Hb to the range of 11to 13 g/dL. However, modeling the association between change in Hb and change in quality of life in CS-2 is problematic not only because of the difficulty in adequately accounting for all confounders of the association, but also because of two types of missing data.

First, nearly one-half of patients are missing data on baseline tumor status. The rationale for the assumption of nonprogressive disease for all patients with missing data is not clearly explained. Moreover, no sensitivity analyses were presented to show the effects of this assumption on the results reported. It cannot be determined whether this assumption was conservative or might overestimate the relationship between increases in Hb and change in quality-of-life scores. A second problem is that final LASA scores were missing from 25 percent of patients and final FACT-An scores were missing from 33 percent of patients. Missing quality-of-life data are unlikely to be missing at random from the entire population enrolled. In fact, patients who lacked a complete set of data were more likely to be nonresponders to epoetin (Ortho Biotech, Inc., personal communication between Jerome Seidenfeld, Ph.D. and Margaret Piper, Ph.D., M.P.H., 1999).

In an attempt to adjust for nonrandomly missing quality-of-life data, the missing data were imputed based on several known characteristics of the patients with complete data, using the econometric Heckman approach. A sensitivity analysis in which the missing final quality-of-life data were all assumed to show no improvement was performed and reported to confirm the results of the analysis using imputed final quality-of-life data.

In summary, a multiple regression analysis was performed on data from two large uncontrolled community studies of epoetin therapy in cancer patients. This analysis showed an association between increases in Hb and increases in quality-of-life scores. The analysis attempted to control for the independent effects of progressive disease on quality-of-life changes, but did not address the potential effects of change in tumor status from progressive to stable/remission. Systematically missing data on baseline tumor status and final quality-of-life scores prevented drawing clear conclusions regarding the significance and magnitude of the association. The data were derived from uncontrolled studies that did not report rigorous protocols for the conduct of quality-of-life assessment, which could contribute to overestimation of the effects of increased Hb on quality-of-life changes. Finally, this approach can establish an association between change in Hb levels and changes in quality of life, but cannot demonstrate a causal relationship. Nor does this model demonstrate a causal relationship between epoetin treatment and improvement in quality of life in cancer patients undergoing chemotherapy.

Issues Specific to Anemia Primarily Due to Cancer Therapy

The population of greatest interest for this systematic review is patients who are beginning chemotherapy (and/or radiotherapy) and who have mild anemia (Grade 1: Hb 10g/dL to < WNL) or who are at risk of anemia. Four alternatives can be envisioned to manage these patients. The most aggressive option is to treat all patients with epoetin with the intent of achieving or maintaining Hb levels within the normal range. An intermediate approach is to initiate epoetin only for patients in the lower portion of the range for mild anemia (i.e., Hb >10 and <12 g/dL) to prevent the development of moderate (Grade 2: Hb 8 to 10 g/dL) anemia. The third option would delay epoetin treatment until Hb falls to <10 g/dL and patients develop moderate anemia. The fourth alternative would be to manage anemia without epoetin. Patients would receive RBC transfusions when symptomatic or when serious/severe anemia (Grade 3: Hb 6.5 to 7.9 g/dL) developed. Each of these four alternatives could be implemented by setting a specific Hb level at which treatment would be initiated, for example, epoetin treatment at Hb <11 g/dL or transfusion at Hb <8 g/dL.

In the trials included in this systematic review, the percent of patients receiving RBC transfusion was always lower in the epoetin arm than in the control arm. However, transfusion was nonetheless relatively common in patients treated with epoetin. One explanation is that the Hb level for initiating epoetin was too low. But other explanatory factors should also be considered. First, each trial had some proportion of patients who did not demonstrate a hematologic response to epoetin, and therefore epoetin could not protect these patients from transfusion. Second, in trials in which mean baseline was <10 g/dL, a substantial proportion of patients might enter with Hb level far enough below the mean that they were already near the Hb transfusion trigger for the study. This would not occur in a clinical setting in which Hb was monitored on an ongoing basis, with epoetin treatment initiated at a predefined threshold. Third, some cancer treatment regimens might cause an accelerated decline in Hb level relative to other treatment regimens, so that rate of decline as well as absolute Hb threshold may be relevant.

The specific chemotherapy regimen and previous treatment history may influence not only the likelihood and severity of anemia but also its course. The likelihood of symptomatic anemia varies with the specific drug or drug combination given, the dose and dose intensity for each drug, and the number of treatment cycles or the duration of the treatment regimen. For patients given less intense or short-duration regimens and without extensive prior treatment, Hb levels may decline briefly with each chemotherapy cycle and then spontaneously return to normal levels. Such patients are low risk for developing chronic symptomatic anemia or requiring RBC transfusion. In contrast, spontaneous correction of anemia is likely to be substantially delayed in those given more myelotoxic chemotherapy for longer times or in those treated for a second or greater relapse with a history of multiple previous therapies. Several drugs (e.g., cisplatin) have direct toxic effects on the kidney that can reduce endogenous production of erythropoietin (Armitage 1998; Beguin 1996; Moliterno and Spivak, 1996; Wood and Hrushesky 1995).

Radiation therapy can be toxic to bone marrow stem cells. However, the degree to which radiation therapy impairs hematopoiesis depends on the dose and extent of radiation. Total body irradiation has the greatest chance of causing anemia. In contrast, conformal radiation targeted to an internal organ distant from any of the larger marrow-containing bones is unlikely to reduce Hb levels. The role of oxygen in the cytocidal effects of radiation therapy for cancer (Hellman, 1997) raises an additional issue. Investigators have hypothesized that correcting anemia might improve the effectiveness of radiation therapy by increasing tissue oxygenation (e.g., Bush 1986; Dische 1991; Fein, Lee, Hanlon, et al., 1995; Poskitt 1987).

Disease characteristics may also influence the likelihood and severity of anemia. Such characteristics include the tissue or organ site of the tumor, its specific histologic type, the burden of malignant cells present, and the site(s) of their dissemination. Some malignancies also are more likely than others to cause the anemia of chronic disease, possibly because of differences in the production of cytokines that reduce expression of endogenous erythropoetin. Tumor burden may influence the dose, intensity, and duration of anemia-causing therapy as well as the level of cytokine production.

Finally, patients are less likely to have clinical responses to epoetin when malignant cells have replaced a large percentage of the normal bone marrow cells. This can occur in patients with some hematologic malignancies as well as in those with tumors of solid organs or tissues with a propensity to metastasize to the bone marrow (e.g., prostate, breast, or small cell lung cancer). Transfusion may be the only option to manage anemia in such patients. For this reason, patients with extensive malignant involvement of the bone marrow were excluded from many clinical trials of epoetin.

Issues Specific to Anemia Primarily Due to Malignant Disease

Patients with anemia that is primarily due to malignant disease are those who would be anemic whether or not they were receiving concurrent treatment for their malignancy. In principle, it would be useful to know if anemic patients who are not being treated for their malignancy differ from those who are being treated with respect to potential benefits from use of epoetin to manage anemia. However, it is difficult in practice to separate studies on disease-related anemia from studies on treatment-related anemia, since presently in the United States, most patients with a malignancy are treated. Furthermore, those who are not being treated (e.g., those in an apparent complete remission who may have subclinical residual disease) are unlikely to have anemia unless their malignancy is a clonal bone marrow disorder.

For this systematic review, clinical trials of epoetin in which all enrolled patients received concurrent therapy for malignancy are included in the review of evidence for treatment-related anemia (Chapter 3). Trials of epoetin that enrolled some patients not treated for malignancy during the study are included in the review of evidence for disease-related anemia (Chapter 4). To be included among the trials on disease-related anemia, studies also must have restricted enrollment to patients with diseases known to have a high occurrence of the anemia of malignancy. However, results from trials with a preponderance of patients receiving chemotherapy or radiation may not apply to populations with anemia and a malignancy that is not being treated.

Anemias resulting from malignancy are heterogenous. For example, anemia resulting from inhibitory cytokines and/or other humoral factors whose production may be increased in patients with certain malignancies may respond differently to epoetin therapy than does anemia due to defects in the stem-cell population or due to inadequate production of endogenous erythropoietin (Moliterno and Spivak, 1996). Consequently, results from studies on patients with anemia primarily due to hematologic malignancies probably do not speak to the potential benefits of using epoetin in patients with anemia primarily caused by tumors of solid tissues or organs. Furthermore, it is likely that hematologic malignancies with different etiologies of anemia also may differ with respect to hematologic responses to and clinical benefits from epoetin.

Some hematologic malignancies (e.g., MDSs, multiple myeloma) may arise from genetic changes to the bone marrow stem-cell population that may reduce their ability to proliferate and/or differentiate into mature RBCs in response to erythropoietin. Note also that malignant myeloid cells may express receptors for erythropoietin; some have speculated that such cells might proliferate more rapidly after epoetin is administered. Patients with myeloid malignancies (i.e., acute and chronic myeloid leukemias) have been excluded from most (if not all) clinical trials of epoetin because of this possibility.

The MDSs are distinctive among the hematologic malignancies that cause disease-related anemia and that have been treated with epoetin. MDS patients usually are not deficient in endogenous erythropoietin (Stein, Abels, and Krantz, 1991). However, the clonal defect in the hematopoietic stem-cell population may render them resistant to the differentiative and maturational effects of erythropoietin. The hypothesis that pharmacologic concentrations of epoetin might overcome these defects led to a series of single-arm trials.

Hellstrom-Lindberg (1995) conducted a meta-analysis of the single-arm studies of epoetin therapy in patients with MDS. The analysis included 17 single-arm studies with a combined total of 205 patients. Response was defined as an increase in Hb of >1.5 g/dL without transfusion support; the overall response rate was 16 percent. Fewer patients with MDS of FAB class RARS responded than did patients with all other FAB classes (7.5 percent versus 21.1 percent, p=0.015), although the study included few patients with RAEB-t or CMML. Patients who were transfusion independent at study entry were more likely to respond than those who were transfusion dependent (44 percent versus 10 percent, p=0.0001). Although the serum concentration of erythropoietin at study entry was lower among responding patients than among nonresponders, this determination did not appear to be an independent predictor of the likelihood of response. For example, the highest response rates were reported for patients in FAB classes RA or RAEB and without previous transfusion history, regardless of baseline serum erythropoietin.

Since all trials included in the meta-analysis by Helmstrom-Lindberg (1995) lacked control groups managed without epoetin, the results are difficult to interpret. They do provide a solid rationale for conducting randomized comparative trials. However, the unique biologic features of MDS make it unlikely that results from controlled trials on patients with this disease are applicable to patients with other malignancies.

Issues Specific to Anemia Resulting from Bone Marrow Ablation Prior to Stem-Cell Transplantation

Patients with malignancies who undergo myeloablative therapy followed by infusion of hematopoietic stem cells invariably go through a period of severely impaired or completely absent erythropoiesis. Neither can they produce new platelets or white blood cells until engraftment of the reinfused stem cells is complete. As a consequence, virtually all transplanted patients require some transfusions with RBCs and platelets. Thus, it is unlikely that epoetin might reduce the percentage of transplant patients who are transfused, although it might reduce the time to RBC engraftment and the number of RBC units transfused per patient.

Early in the investigation of epoetin for patients undergoing stem-cell transplants, there was concern that epoetin might divert too much of the stem-cell pool into the erythroid differentiation pathway. It was hypothesized that if this happened, recovery of neutrophil and/or platelet counts might be delayed. Consequently, the effects of epoetin on neutrophil engraftment and on platelet counts were outcomes of interest in these trials.

For transplants in the inpatient setting, there was interest in whether epoetin treatment might reduce the length of hospitalization by reducing the time to erythroid engraftment. However, recovery of neutrophil and platelet counts also is necessary before hospitalized transplant patients can be discharged. Thus, some studies have investigated the use of epoetin in combination with myeloid growth factors such as filgrastim or sargramostim.

The source of stem cells may also play a role in the need for or response to epoetin. Allogeneic transplants may be from a related or unrelated donor who may be matched or partially mismatched at the six human leukocyte antigen (HLA) loci. With autologous transplants, the patient's own stem cells are reinfused. The potential for epoetin to accelerate engraftment, reduce transfusion use, or improve other outcomes may differ for allogeneic and autologous transplants and may also depend on the degree of HLA mismatch. In addition, stem cells may be obtained from bone marrow or from peripheral blood, which may differ in either the need for or the response to epoetin.

Peripheral blood stem cells (PBSCs) are usually harvested after mobilization with myeloid growth factors. The infusate includes progenitor cells that are farther along the differentiation and maturation pathway towards end-stage RBCs than are present in bone marrow-derived stem cells. Myeloid engraftment is known to be faster with PBSC than with marrow-derived stem cells. Consequently, epoetin treatment after the stem cells are infused may be unable to further accelerate erythroid engraftment in this setting.

Stem-cell transplant patients experience a high incidence of serious and life-threatening adverse effects of treatment that are unrelated to anemia and its management. Few, if any, studies in the transplant setting administer epoetin for more than 2 months or monitor outcomes for more than 3 months. By that time, engraftment of all three lineages is usually complete in most patients. Allogeneic transplants carry a high risk for GVHD, which may cause serious morbidity or mortality. All transplant patients are at risk for serious, life-threatening infections or bleeding episodes until engraftment has occurred. Adverse outcomes due to these complications may overwhelm any potential benefit from managing anemia with epoetin. In addition, patients who die early from transplant-related complications are lost to follow-up for evaluation of erythroid engraftment.

Selection Criteria

Types of Participants

Types of Interventions

We required that studies be designed as controlled trials comparing the outcomes of managing anemia with and without the use of epoetin. Epoetin is intended to prevent the occurrence of anemia that is so severe that RBC transfusion becomes necessary. All trials that met the study selection criteria for this systematic review compared epoetin plus RBC transfusion as necessary to RBC alone. Red blood cell transfusion was initiated at a predefined Hb threshold (usually 7 to 9 g/dL) or at the discretion of the treating physician.

We identified four characteristics of epoetin administration that might affect treatment outcomes: route of administration, starting dose, class of dosing regimen, and duration of treatment as detailed below. Few studies that met the selection criteria for this systematic review directly compared the effectiveness of various doses, dosing regimens, treatment durations, or routes for administering epoetin. For this reason, we attempted indirect comparison of outcomes of the various characteristics of epoetin administration across the studies included in this systematic review.

• Route of administration was subcutaneous or intravenous.

• To facilitate comparisons among studies, all reported dosages were calculated as units per kilogram per week. All studies were classified within a range of starting doses. Where available, we report both starting dose and ending dose for each study.

• Dosing regimens were classified as fixed dose with continuous treatment, decreasing dose, and increasing dose.

• To permit comparison among studies, treatment duration was classified by ranges (e.g., <10 weeks, 12 to 16 weeks).

The three classes of dosing regimens were defined as follows:

• Fixed and continuous dose. All patients treated with epoetin in a trial with a fixed and continuous regimen received the same dose throughout the study.

• Decreasing dose. Studies that utilized a stop/start regimen temporarily discontinued epoetin when Hb levels rose above a predetermined threshold and then resumed treatment when Hb levels fell below a second threshold. Other trials decreased the epoetin dose only for responding patients, seeking to maintain Hb levels within a range specified in their protocol. The stop/start and decreasing dose regimens are considered together in this evidence report ("decreasing dose"), since either approach reduces the amount of epoetin given to responding patients.

• Increasing dose. Patients who did not respond to the initial dose by a specified time had the dose increased, most often by a factor of two.

Costs and Other Economic Outcomes

Our protocol required that data on financial costs were to be abstracted only from studies that reported on a group treated with epoetin and a control group managed without epoetin and met all study inclusion criteria for this systematic review. In addition, the studies were required to report costs for all patients treated (not just those who develop anemia) and capture (or attempt to capture) the full range of costs for patient care over the entire course of treatment (e.g., not just compare the drug costs for epoetin with the costs of the units of blood transfused).

No trials meeting the inclusion criteria for this systematic review reported cost data. As a result, our review of the literature on epoetin costs is limited to a summary of secondary analyses that are discussed in the introductory section of this systematic review.

Determining Study Eligibility

The following procedure was followed to systematically screen citations, select those to be retrieved, and identify those meeting the study inclusion/exclusion criteria. Initially, one of two independent reviewers evaluated each title and abstract against the study inclusion and exclusion criteria. Each reviewer was responsible for one-half of the retrieved citations. The reviewers sorted citations into three categories: "retrieve," "hold," and "uncertain." Full copies were obtained of all articles sorted into the "retrieve" category by either reviewer. Next, each reviewer evaluated all citations sorted into the "hold" category by the other reviewer. Full copies also were obtained of any reference initially sorted as "hold" which the second reviewer sorted as "retrieve." The two reviewers discussed all references sorted as "uncertain," with a bias toward being inclusive.

Table 5. Reasons for study exclusion -- oncology

Table 5. Reasons for study exclusion -- oncology

Reason for Exclusion	Criterion
Not in English	Full-length journal article was not published in the English language.
No primary data	Review article, editorial, commentary, letter, or position statement that did not report primary outcomes data from a clinical trial.
Not a controlled trial	Trial did not include a control population managed without the use of epoetin (any type of control group, including historical controls, is acceptable for inclusion). Or, a nonrandomized, controlled trial did not establish comparability of study populations.
N <10 per study arm	One or more arms of the trial included fewer than 10 similarly treated patients. If a dose-finding study, there were fewer than 10 patients per dose (include if there was a control group and at least one dose arm with 10 or more patients each). For studies of stem-cell transplant patients, there were not at least 10 patients with malignancy in each of the control and experimental arms.
Preoperative treatment or for collecting autologous blood	Study focused exclusively on the use of short-term preoperative treatment with epoetin either to reverse anemia or to support collection of autologous blood prior to cancer surgery. Or, study focused exclusively on the use of epoetin to mobilize stem cells.
No R/O of other anemia causes	For studies of patients with anemia due to disease, articles do not state that one or more other treatable causes of anemia were ruled out. Not applicable to trials of epoetin for treatment-related anemia or stem-cell transplants.
Not same transplant procedure	For studies of stem-cell transplant patients, not all patients received the same transplant procedure (autologous or allogeneic, bone marrow or peripheral blood stem cells), and reported outcomes were not stratified by transplant procedure.
Epoetin administration incomplete	For studies of stem-cell transplant patients, epoetin was not administered from stem-cell reinfusion through recovery of hematopoiesis (e.g., epoetin to mobilize PBSC).
No Hb, Hct reported	Study does not report Hb or Hct as a function of time after initiation of treatment. (Not applicable for studies of epoetin after stem-cell transplants.)

The resulting bibliography of included studies was circulated to the Technical Advisory Group and to the members of the ASH/ASCO Joint Guidelines Panel on Erythropoietin for review for possible omissions. No additional studies meeting the inclusion criteria were identified.

Data Abstraction

Two reviewers independently abstracted data from each eligible study, recording the information with electronic database software (Microsoft® Access 97). The data elements that were abstracted are listed in the data abstraction forms and defined in the accompanying table (Appendix C).

Data elements were grouped into the following broad categories: trial identifiers, study design and methods (including enrollment and withdrawal numbers), patient characteristics, outcomes, and predictors of response.

If an article did not report exact numerical values for one or more of these elements, the reviewers estimated them from figures if they were available in the published reports. After each reviewer completed data abstraction, databases were compared electronically and disagreements were resolved by consensus of the two reviewers, or by a third reviewer if required.

If p values were not reported, but sufficient information was provided, we calculated p values using chi-square analysis or Fisher's exact test (when one or more cells contained five or fewer data points).

Assessment of Comparability of Study Arms

Comparability of study arms was assessed for two purposes. Nonrandomized studies were excluded if the minimum set of data elements we required to determine comparability was not reported or if obvious selection bias was detected. For randomized trials, most of which were small, we sought to identify differences in study arms that might affect the interpretation of results or that might explain heterogeneity of results among studies. Any such findings are reported in the review of evidence.

Each trial was evaluated by two reviewers for comparability of study arms based on the data reported. We required a minimum set of specific elements to assess comparability of study arms. This set included patient age, tumor type, and baseline Hb value for studies of the anemia resulting from cancer therapy and the anemia caused by malignant disease; and patient age, tumor type, and information on the conditioning regimens used in each study arm for studies of anemia after stem-cell transplant. Other elements evaluated related to the severity of anemia, the severity of malignant disease, previous cancer therapy, or cancer treatment while on study.

Note that uncontrolled hypertension is a contraindication for use of epoetin alfa in the FDA-approved labeling for the drug. In addition, many of the trials that met inclusion criteria for this systematic review excluded patients with other comorbidities (e.g., cardiovascular disease, major organ dysfunction) that might affect either the decision to transfuse patients or the frequency of adverse events. Consequently, it was judged unlikely that imbalances between study arms with respect to comorbidities might affect the interpretation of results.

For studies that did not report statistical comparison among study arms, we compared the data elements reported. No studies lacking statistical comparison reported sufficient data to perform a statistical test; so we simply estimated equivalence from the raw numbers or percentages reported. If insufficient data were provided to assess overall comparability (e.g., insufficient elements reported, patient information not separated by study arm, or study published as an abstract only), this was noted.

The number of studies that reported each specific element also was compiled; we also noted which data elements were omitted from the studies for which there was inadequate information to assess overall balance between arms.

Classification of Alternative Hemoglobin Thresholds for Initiating Epoetin Treatment

Our systematic review protocol called for grouping studies by upper limit of Hb level at study entry, as reported in the patient eligibility criteria of each study. The thresholds of interest were: Hb >12 g/dL; Hb >10 and <12 g/dL; Hb <10 g/dL or requiring blood transfusions; and prevention studies. Prevention studies are those studies that enrolled patients whose Hb was above a minimum threshold but without an upper limit of Hb for enrollment. Upon examining the abstracted data, we found that the mean baseline Hb at study entry was generally lower than our initial classification would suggest. As a result, we revised our protocol and grouped studies by mean baseline Hb at study entry.

Our comparison of the studies selected for Part I of this systematic review (Chapter 3, Anemia Resulting Primarily From Cancer Therapy) illustrates why we found mean Hb level at entry to be more informative for grouping the included studies than Hb cutoffs for enrollment. Among these studies, the mean or median Hb level at entry was <10 g/dL in the preponderance of studies (9 of 10 trials, with 1,134 of 1,164 enrolled patients) that we had initially classified as setting upper limits for eligibility between 10 and 12 g/dL. Similarly, the mean or median Hb level was >10 and <12 g/dL in the majority of studies (5 of 6 trials, with 338 of 386 enrolled patients) that set an upper limit for eligibility >12 g/dL.

This systematic review classifies included studies into three categories defined by mean Hb at enrollment: Hb >12 g/dL; Hb >10 and <12 g/dL; and Hb <10 g/dL. We also abstracted and reported standard deviations, where reported. If the standard deviation was not reported, we abstracted the range, if reported. In the few studies where mean Hb was not reported, we used the median Hb. Where there was a discrepancy in mean Hb between the epoetin and the control arms, we classified the trial by the Hb level of the epoetin arm.

Quality Assessment for Sensitivity Analysis

The objective of quality assessment for this systematic review was to identify a group of "higher quality trials," for purposes of sensitivity analysis. Our meta-analysis includes a quantitative sensitivity analysis; and throughout this systematic review, we have included qualitative sensitivity analyses in our summaries of study conclusions. Our sensitivity analyses compare the results reported and conclusions reached from all included studies to results and conclusions drawn by examining the outcomes of only higher quality studies.

Sensitivity analysis based on study quality is useful because trials of lower quality generally overestimate the effectiveness of an intervention compared with higher quality trials. More than two decades ago, Chalmers showed that randomized studies report smaller treatment effects than nonrandomized studies (Chalmers, Smith, Blackburn, et al., 1981). Subsequently, many methodologists have attempted to identify the characteristics that define the quality of randomized trials and to test whether such characteristics have an effect on study results (Schulz, Chalmers, Hayes, et al., 1995).

Although many quality scales have been used to assess the quality of randomized controlled trials, there is a dearth of empirical evidence to validate such scales. Indeed, Juni and colleagues recently illustrated the hazards of using summary quality scores to select or pool studies for meta-analysis (Juni, Witschi, Bloch, et al., 1999). They identified 25 different quality scales, which they tested for a meta-analysis of 17 trials comparing low molecular weight and standard heparin. No significant association between summary quality scores and treatment effects was found; and the results of different quality scales yielded different conclusions as to which treatment was superior.

Although the use of quality summary scores is problematic, there are three domains of study quality that have been tested in empirical studies. These are concealment of treatment allocation during randomization, double-blinding, and handling of withdrawals and exclusions. Although there is evidence suggesting that these quality domains are associated with more valid estimates of treatment effects, all three domains have not been reported as significant in all studies (Juni, Witschi, Bloch, et al., 1999; Moher, Pham, Jones, et al., 1998; Mulrow and Oxman, 1997; Schulz, Chalmers, Hayes, et al., 1995). Moreover, assessment of study quality generally depends on information reported in journal articles, and the absence of such information may reflect incomplete reporting rather than flawed study design. This point is especially germane to studies published prior to the Consolidated Standards of Reporting Trials (CONSORT) statement, which was published in the Journal of the American Medical Association in 1996, to disseminate a standard for completeness of reporting in journal articles (Begg, Cho, Eastwood, et al., 1996).

In an editorial accompanying the Juni study, Berlin and Rennie (1999) suggest that, to be clinically relevant, quality assessment of trials should focus on key aspects of research design relative to the outcomes of interest. Thus, where an outcome requires subjective judgment, double-blinding may be of paramount importance but may matter less for outcomes where there is little discretion regarding assessment or interpretation. In this systematic review of use of epoetin in cancer-related anemia, we had substantial concerns about the impact of subjective judgments on treatment outcomes. The clinical outcomes of interest (transfusion, quality of life, fatigue and other anemia-related symptoms) can obviously be affected by physician behavior and patient perceptions.

For example, transfusion is the most commonly reported of these clinical outcomes. We anticipated that physicians might be more aggressive in transfusing patients they knew were in the control arm and less aggressive in transfusing those they knew were in the epoetin arm. One study included in this systematic review reported relevant data (Thatcher, De Campos, Bell, et al., 1999). In this unblinded study, there was variation among participating centers in the mean Hb at which first transfusion was initiated, and the range was markedly lower for the higher dose than the lower dose epoetin group (8.7 to 10.8 g/dL versus 8.4 to 12.9 g/dL). Many studies included in this systematic review had a transfusion trigger, i.e., a prespecified Hb level at or below which transfusion was initiated, intended to achieve consistency in transfusion practice. However, these studies rarely reported the mean Hb level at which transfusion actually occurred, so it is impossible to confirm whether transfusion practice was, in fact, consistent across study arms.

In developing our approach to sensitivity analysis, we put highest priority on comparing the outcomes of double-blinded randomized controlled trials to all other trials of weaker design with respect to minimizing the effects of subjective judgements on treatment outcome. While a high proportion of trials included in this systematic review were randomized, the preponderance of these were unblinded. We also assessed study quality with respect to the domains of allocation concealment and handling of exclusions and withdrawals. But in defining higher quality studies, we wanted to avoid criteria that might be overly sensitive to how a trial was reported, thus diminishing the pool of double-blinded randomized controlled trials for sensitivity analysis. For the domain of handling of exclusions and withdrawals, we set a stringent cutoff for the maximum number of patients that could be excluded from the analysis of results. Although we looked for, and often found, an explanation of exclusions and withdrawals, this was not required. As was the case in a prior evidence report for the AHRQ, information on concealment of allocation was reported infrequently (Aronson, Seidenfeld, Samson, et al., 1999) and was not required for our group of higher quality studies.

Patient Populations

Comparability of Study Arms

Table 6. Study arm comparability

Table 6. Study arm comparability

Study Arms Comparable?	How Assessed?	Number of Studies	N Enrolled (controls+treated)	N Evaluable (controls+treated)
Insufficient data	Estimated by reviewers	6	256 (112+144)	249 (109+140)
Yes	Reported statistical tests	13	1,538 (687+851)	1,458 (640+818)
Yes	Estimated by reviewers	3	133 (61+72)	131 (61+70)
Totals		22	1,927 (860+1,067)	1,838 (810+1,028)

We examined included studies for potential imbalances between study arms as described in the methods section. Sixteen studies were judged to have comparable study arms (Table 6); of these, only three actually reported a statistical comparison. Six randomized controlled trials did not report sufficient data for the reviewers to assess comparability of study arms. Note that any nonrandomized trials without sufficient data to assess comparability of study arms were excluded from this systematic review.

Table 7. Reporting on Elements of Study Arm Comparability (more...)

Table 7. Reporting on Elements of Study Arm Comparability

Element	Number	Element	Number
Type of malignancy(ies)	19	No. patients with bone marrow metastases	2
Baseline Hb value	17	No. previous chemotherapy regimens	2
Patient age	16	No. radiotherapy cycles during study	2
Chemotherapy dose intensity (by platelet or neutrophil count)	11	No. patients with previous platinum-based chemotherapy	1
No. transfusion-dependent patients	10	No. previous chemotherapy cycles	1
No. patients with platinum-based regimen(s) during study	7	No. patients with previous radiotherapy	0
No. chemotherapy regimens during study	6	No. patients with previous total body irradiation	0
Performance score	5	No. patients with total body irradiation during study	0
No. chemotherapy cycles during study	4

The specific elements reported to address comparability of study arms varied considerably from study to study. Table 7 lists the elements sought to assess study arm balance, in decreasing order of frequency for the number of studies that reported each element.

Despite incomplete reporting of details, it is unlikely that imbalances in the study arms pose a threat to validity for this analysis. The evidence base consists largely of randomized controlled trials, with 1,690 patients (88 percent of the total patients) randomized into 18 trials.

Subpopulations of Interest

Table 8. Evidence on subpopulations

Table 8. Evidence on subpopulations

Characteristic	Subpopulations	Number of Studies	N Enrolled (controls+treated)	N Evaluable (controls+treated)
Tumor type (studies counted twice if outcomes stratified)	Hematologic	2	221 (78+143)	216 (76+140)
	Solid organs and tissues	15	1,221 (542+679)	1,166 (510+656)
	Mixed	7	844 (355+489)	815 (339+476)
Therapy regimen	Radiotherapy only	3	132 (55+77)	130 (55+75)
	Nonplatinum chemotherapy, no or unknown radiotherapy	3	586 (224+362)	561 (211+350)
	Nonplatinum chemotherapy and radiotherapy	2	86 (43+43)	82 (41+41)
	Platinum chemotherapy, no or unknown radiotherapy	9	777 (363+414)	721 (329+392)
	Platinum chemotherapy and radiotherapy	3	240 (122+118)	240 (122+118)
	Unknown chemotherapy type, no or unknown radiotherapy	2	106 (53+53)	104 (52+52)
Patient age	Adults 1	18	1,763 (778+985)	1,680 (731+949)
	Pediatric	3	108 (54+54)	104 (52+52)
	Geriatric	0	0	0
	Unspecified	1	56 (28+28)	54 (27+27)
Transfusion history	<20% previously transfused	3	204 (75+129)	200 (74+126)
	>80% previously transfused	1	50 (25+25)	50 (25+25)
	Unknown or 21-79%	18	1,673 (760+913)	1,588 (711+877)
Iron supplementation	Both arms supplemented	9	449 (229+220)	444 (226+218)
	Neither arm supplemented	3	194 (75+119)	193 (75+118)
	EPO arm only 2	3	152 (73+79)	145 (71+74)
	Neither arm specified	7	1,132 (483+649)	1,056 (438+618)

1

Includes 2 assumed to be all adults, based on included malignancies.

2

Controls: 2 trials, not supplemented; 1 trial, not specified.

Table 8 shows the number of studies and patients included in this systematic review relevant to the subpopulations of interest. In this table, analyses of stratified subgroups within a single study are tabulated with studies that were homogeneous for the same characteristic.

Type of malignancy

Two studies (n=221 enrolled) restricted enrollment to patients with hematologic malignancies, and 15 studies (n=1,221 enrolled) restricted enrollment to patients with tumors of solid organs or tissues. In seven of the latter 15 studies (n=513 enrolled), all patients had one type of solid tumor (Evidence Table I-1). This included three trials on ovarian cancer (n=215) and one trial each on breast cancer (n=62), small cell lung cancer (n=130), Ewing's and osteosarcoma (n=30), and cervical cancer (n=76). Seven trials (n=844) enrolled patients with either hematologic or nonhematologic malignancies.

Cancer treatment regimens

Of the 19 studies of patients receiving chemotherapy, 12 (n=1,017 enrolled) used regimens that included either cisplatin or carboplatin and 5 trials (n=672 enrolled) used regimens that did not include platinum. Two studies (n=106 enrolled) did not provide information on the cancer treatment regimens utilized. Seven trials (n=384 enrolled) used chemotherapy alone, three trials (n=132 enrolled) used radiation therapy alone, and five trials (n=326 enrolled) used chemotherapy plus radiation therapy. Seven trials (n=1,085 enrolled) did not provide information on use of radiation therapy.

Age

Three small trials (n=108 enrolled) were limited to pediatric patients. Although many trials included patients older than 65 years, and some even included patients in their eighties and nineties, none of the included trials provided data separately for geriatric populations.

Transfusion history

Most trials did not provide information on the previous transfusion history of the patients enrolled. Furthermore, for most trials with data, the percentage of previously transfused patients was between 21 percent and 79 percent and outcomes were not reported separately based on previous transfusion history. Only three studies (n=204 enrolled) reported outcomes for patient groups in which <20 percent were previously transfused, and only one study (n=50 enrolled) reported on patients of whom >80 percent were previously transfused.

Iron supplementation

Nine trials (n=449 enrolled) supplemented both epoetin-treated and control arms with iron, and three trials (n=194 enrolled) did not supplement either arm.

Predictors of response

Table 33. Results for any potential predictor reported (more...)

Table 33. Results for any potential predictor reported by two or more studies

Predictor	Number of Studies Reporting	Number of Studies with Significant Result	Comment
Baseline serum erythropoietin level	6	0
Observed/predicted serum erythropoietin ratio (O/P)	2	0	ten Bokkel Huinink, de Swart, van Toorn, et al., (1998) reported nonsignificant trend for O/P <0.8, p=0.147 (chi-square)
Type of malignancy	4	0
Presence of metastases	2	0
Disease severity	4	0
Platelet count	2	0
Ferritin	3	~1	Henke, Guttenberger, Barke, et al., (1999) reported a correlation between serum ferritin and the Hb increment at week (r2=0.34)
Serum iron	3	0
Total iron-binding capacity	2	0
Type of chemotherapy	2	0
Patient age	2	0

Ten studies (n=955) reported information on factors thought to predict response to epoetin treatment. Table 33, presented in the Results section of this chapter, reports the results on any factor that was analyzed as a potential predictor of response in at least two papers.

Interventions

All trials compared the outcomes of managing anemia with epoetin treatment (plus transfusion if necessary) to those obtained with RBC transfusion alone in patients undergoing therapy for a malignancy.

Epoetin

The characteristics of epoetin administration of interest to this analysis are route of administration, dose, dosing regimen, and duration of treatment. Patients in the epoetin arms were transfused if necessary, generally under the same conditions as for control patients (see Red Blood Cell Transfusion, following).

Route

Table 9. Two-Arm Studies of Subcutaneous Epoetin

Table 9. Two-Arm Studies of Subcutaneous Epoetin

Range of Weekly Epoetin Dose	Epoetin Treatment Regimen	Treatment Duration	No. of Studies	N Enrolled (controls+treated)	N Evaluable (controls+treated)
300-450 U/kg	Fixed dose and continuous treatment	<10 weeks	1	34 (17+17)	34 (17+17)
		2-16 weeks	2	265 (127+138)	227 (105+122)
		>20 weeks	1	63 (46+17)	56 (40+16)
		Subtotals	4	362 (190+172)	317 (162+155)
300-450 U/kg	Decreasing doses at Hb > specified thresholds	< 10 weeks	1	100 (50+50)	99 (49+50)
		12-16 weeks	3	351 (172+179)	340 (166+174)
		Subtotals	4	451 (222+229)	439 (215+224)
300-450 U/kg	Increasing doses for nonresponders	12-16 weeks	2	91 (40+51)	89 (39+50)
		>20 weeks	2	429 (148+281)	408 (137+271)
		Subtotals	4	520 (188+332)	497 (176+321)
300-450 U/kg	Totals for all low-dose studies		12	1,333 (600+733)	1,253 (553+700)
750-1,000 U/kg	Fixed & continuous	12-16 weeks	1	50 (25+25)	50 (25+25)
	Decreasing doses	< 10 weeks	3	164 (105+59)	162 (105+57)
		>20 weeks	1	30 (15+15)	30 (15+15)
750-1,000 U/kg	Totals for all high-dose studies		5	244 (145+99)	242 (145+97)

Table 10. Three-arm subcutaneous studies -- direct (more...)

Table 10. Three-arm subcutaneous studies -- direct comparison of dosage

Study	Epoetin Treatment Regimen	Epoetin Treatment Duration	Weekly Dose (units per kg per week)	N Enrolled	N Evaluable
Thatcher, De Campos, Bell, et al., 1999	Stop/start or decreasing doses	>20 weeks	0	44	44
			450 sc	42	42
			900 sc	44	44
ten Bokkel Huinink, de Swart, van Toorn, et al., 1998	Stop/start or decreasing doses	>20 weeks	0	34	33
			450 sc	46	45
			900 sc	42	42

Seventeen of the trials in this evidence review are two-arm trials comparing subcutaneous epoetin with transfusion alone. Table 9 summarizes the dosages, dosage regimens, and treatment durations used in these studies. Two three-arm trials (n=252) compared subcutaneous epoetin at two different doses with transfusion alone (ten Bokkel Huinink, de Swart, van Toorn, et al., 1998; Thatcher, De Campos, Bell, et al., 1999); these are summarized in Table 10.

Table 11. Studies of intravenous epoetin

Table 11. Studies of intravenous epoetin

Weekly Dose of Epoetin	Epoetin Treatment Regimen	Treatment Duration	No. of Studies	N Enrolled (controls+treated)	N Evaluable (controls+treated)
450 U/kg	Increasing	12-16 weeks	1	24 (12+12)	20 (10+10)
1,200 U/kg	Decreasing	>20 weeks	1	30 (14+16)	29 (14+15)
450/900 U/kg	Decreasing	< 10 weeks	1	44 (11+19+14)	44 (11+19+14)

Three studies (n=98 enrolled) used intravenous administration (Table 11). Because the biological effectiveness of intravenous epoetin is less, relative to the same dose administered subcutaneously, we did not compare outcomes of intravenous with those of subcutaneous epoetin.

Dose

In studies using the subcutaneous route of administration, the most common initial dose was 150 units/kg administered three times weekly; when a higher dose was used, the most common was 300 units/kg administered three times weekly. A few studies used different dosages and frequencies. To facilitate comparison, dosages were calculated as units/kg per week and classified into one of two categories: 300 to 450 and 700 to 1,000 units/kg per week. We assumed that, for the range of doses and frequencies in these categories, differences in frequency of administration would have a negligible effect on the bioavailability of the dose. Twelve two-arm trials (n=1,333) used the lower dose range and five trials (n=544) used the higher dose range. Two three-arm trials (n=252) compared initial doses of 450 to 900 units/kg per week.

One study (Oberhoff, Neri, Amadori, et al., 1998) treated all patients with the same total dose, 5,000 units daily, regardless of weight or body size. This trial (n=217) was omitted from the comparison of interventions since dosage could not be classified as described.

Dosing regimen

Among studies using subcutaneous epoetin in the lower dose range, four (n=520) increased the initial dose for nonresponders after a fixed period of time; four (n=451) decreased the dose for responders; and four (n=362) used a fixed and continuous dose throughout the treatment period. The two trials (n=252) that compared initial doses of 450 to 900 units/kg per week used decreasing dose regimens.

Duration of treatment

Treatment duration was grouped into three classes: >20 weeks, 12 to 16 weeks, and <10 weeks. In two trials, treatment duration was not reported, but a minimum duration of treatment could be ascertained and was used to group the study into one of the three classes. Of studies using epoetin subcutaneously, treatment duration was >20 weeks in six trials (n=774), 12 to 16 weeks in eight trials (n=757), and <10 weeks in five trials (n=298).

Outcomes of Interest

Efficacy Outcomes

Table 12. Evidence to compare outcomes by baseline (more...)

Table 12. Evidence to compare outcomes by baseline mean or median hemoglobin

Outcome	Hb Level at Baseline	No. of Studies	N Enrolled (controls+treated)	N Evaluated (controls+treated)
Percent of patients responding	Not specified	1	54 (24+30)	49 (22+27)
	< 10 g/dL	6	1,026 (437+589)	960 (399+561)
	>10 and <12 g/dl	2	103 (66+37)	96 (60+36)
	>12 g/dL	2	178 (68+110)	176 (68+108)
	TOTALS	11	1,361 (595+766)	1,281 (549+732)
Change in Hb levels	< 10 g/dL	7	883 (369+514)	855 (353+502)
	>10 and <12 g/dl	4	216 (120+96)	214 (119+95)
	>12 g/dl	5	308 (131+177)	306 (131+175)
	TOTALS	16	1,407 (620+787)	1,375 (603+772)
Percent of patients transfused	< 10 g/dL	9	1,134 (491+643)	1,064 (451+613)
	>10 and <12 g/dl	5	347 (183+164)	335 (175+160)
	>12 g/dl	3	222 (90+132)	222 (90+132)
	TOTALS	17	1,703 (764+939)	1,621 (716+905)
RBC units per patient	< 10 g/dL	7	725 (350+375)	671 (319+352)
	>10 and <12 g/dl	3	208 (76+132)	203 (74+129)
	>12 g/dl	2	160 (59+101)	160 (59+101)
	TOTALS	12	1,093 (485+608)	1,034 (452+582)
Quality of life	< 10 g/dL	5	749 (302+447)	722 (287+435)
	>10 and <12 g/dl	0	0	0
	>12 g/dl	4	232 (97+135)	230 (97+133)
	TOTALS	9	981 (399+582)	952 (384+568)

Table 12 summarizes the evidence available from studies included in this systematic review on the outcomes of interest, grouped by mean or median Hb at study entry.

• Change in Hb levels (or data to calculate Hb change) was reported in 16 trials (n=1,407 enrolled).

• Percentage of patients demonstrating hematologic response was reported in 11 studies (n=1,361 enrolled). Response was defined as an increase in Hb level either by a specified amount (usually 2 g/dL; see Evidence Table I-1) or to a specified target, without use of RBC transfusion.

• Percentage of patients transfused was reported in 17 trials (n=1,703 enrolled).

• The number of RBC units transfused per patient was reported in 12 trials (n=1,093).

• Quality of life was reported in nine trials (n=981).

That all outcomes were not reported by all studies raises the possibility of reporting bias. For example, 9 of 10 studies with baseline Hb <10 g/dL reported percentage of patients transfused, but 4 of 12 studies (33 percent) with baseline Hb >10g/dL did not. Thus, studies of patients at lower baseline risk of transfusion reported percentage of patients transfused less frequently than studies of patients who had higher risk of transfusion.

None of the trials reported on symptoms of anemia (including shortness of breath, dyspnea on exertion, or angina) or number of days in hospital. The only trial to report on changes in performance status used Karnovsky performance status as a quality-of-life surrogate (Leon, Jimenez, Barona, et al., 1998).

Adverse Events

Table 13. Adverse events reported by studies on patients (more...)

Table 13. Adverse events reported by studies on patients being treated for malignancy

Adverse Event	No. of Studies Reporting	N Evaluated (controls+treated)
Any adverse effect (each patient counted once only)	10	1155 (473+682)
Hypertension (highest freq. if systolic/diastolic separated)	9	722 (285+437)
Deep vein thrombosis or thromboembolism	6	580 (238+342)
Hemorrhage and/or thrombocytopenia	2	161 (80+81)
Skin rash, irritation, and/or pruritus	6	372 (121+251)
Seizures	4	408 (200+208)
Injection site pain	3	177 (77+100)
Fatigue (separate from quality-of-life reporting)	4	699 (264+435)
Withdrawals (due to adverse events)	8	846 (387+459)
Mortality (from any cause, while on study)	2	338 (137+201)

Table 13 summarizes reporting on adverse events by the studies included in this review.

Study Quality

Table 14. Assessment of study quality¹

Table 14. Assessment of study quality¹

Citation	Blinding (required)	Percentage of Excluded Subjects Below Specified Threshold? 2 (required)	Accounted for Excluded Patients?	Allocation Concealed?	Transfusion Trigger?	R/O Other Anemia Causes? 3	Fe Status confirmed? 4	Patients blinded to Hb levels? 5
Mean/Median Baseline Hb <10 g/dL; Adult Patients
Silvestris, Romito, Fanelli, et al., 19956	Unblinded	Yes	No/NS	Yes	NA 7	No	Yes
Oberhoff, Neri, Amadori, et al., 1998	Unblinded	No	No/NS	No/NS	No	No	No
Case, Bukowski, Carey, et al., 1993	Double blinded	Yes	No	No/NS	Yes	Yes	No	No/NS
Henry, Brooks, Case, et al., 1995	Double blinded	Yes	No	No/NS	Yes	Yes	No	No/NS
Cascinu, Fedeli, Del Ferro, et al., 1994	Double blinded	Yes	Yes	Yes	Yes	No	Yes
Kurz, Marth, Windbichler, et al., 1997	Double blinded	Yes	No	Yes	Yes	No	Yes	No/NS
Littlewood, Bajetta, Cella, et al., 1999(Abstract/Slides)	Double blinded	Yes	No/NS	No/NS	Yes	No	No	No/NS
Mean/Median Baseline Hb <10 g/dL; Pediatric Patients
Varan, Buyukpamukcu, Kutluk, et al., 1999	Unblinded	Yes	Yes	No/NS	Yes	No	No
Leon, Jimenez, Barona, et al., 1998	Unblinded8	Yes	Yes	No/NS	Yes	Yes	Yes	No/NS
Porter, Leahey, Polise, et al., 1996	Double blinded	No	No/NS	Yes	Yes	No	Yes
Mean/Median Baseline Hb >10 and <12 g/dL; Adult Patients
Markman, Reichman, Hakes, et al., 1993	Unblinded	No	No	No/NS	Yes	N/A9	No
Dusenbery, McGuire, Holt, et al., 1994	Unblinded8	Yes	Yes	No/NS	Yes	No	Yes
Lavey and Dempsey, 1993	Unblinded	Yes	Yes	No/NS	NA7	No	Yes
Wurnig, Windhager, Schwameis, et al., 1996	Double blinded	Yes	Yes	No/NS	Yes	Yes	No
Henke, Guttenberger, Barke, et al., 1999	Unblinded	Yes	Yes	No/NS	NA 7	No	Yes
Quirt, Couture, Pichette, et al., 1996(Abstract)	Single blinded	Yes	No/NS	No/NS	No	No	No	No/NS
ten Bokkel Huinink, de Swart, van Toorn, et al., 1998	Unblinded	Yes	Yes	Yes	Yes	No	No
Mean/Median Baseline Hb >12 g/dL; Adult Patients
Gamucci, Thorel, Frasca, et al., 1993	Unblinded	Yes	No/NS	No/NS	NA 7	N/A 9	Yes
Sweeney, Nicolae, Ignacio, et al., 1998	Unblinded	Yes	Yes	No/NS	NA 7	Yes	Yes	No/NS
Del Mastro, Venturini, Lionetto, et al., 1997	Unblinded	Yes	Yes	Yes	Yes	N/A	Yes	No/NS
Thatcher, De Campos, Bell, et al., 1999	Unblinded	Yes	Yes	No/NS	Yes	Yes	No	No
Welch, James, Wilkinson, et al., 1995	Unblinded	Yes	Yes	No/NS	Yes	N/A	Yes	No/NS

¹1"Higher quality" trials in bold font; nonrandomized studies in italics

2

<5 percent of subjects were excluded in each study arm OR <10 percent of subjects were excluded in each study arm AND the ratio between arms for the percentage of subjects excluded from the analysis was <2:1.

3

Ruled out all of the following: iron, B12, and folate deficiencies, occult bleeding, and hemolytic anemia.

4

Epoetin arm supplemented OR serum Fe, ferritin, and transferrin saturation all monitored and reported in results.

5

Only evaluated for studies reporting quality-of-life outcomes.

6

Mean/median baseline Hb not specified, but excluded patients with baseline Hb >10 g/dL.

7

NA=Not applicable because transfusion outcomes were not reported.

8

Historical controls only; all other nonrandomized studies used concurrent controls.

9

N/A=Not applicable because enrollment limited to nonanemic patients.

NS=not specified.

Six of the 22 trials were identified as higher quality for purposes of sensitivity analysis. Five of these (n=799) were studies of patients with baseline Hb levels <10 g/dL (Cascinu, Fedeli, Del Ferro, et al., 1994; Case, Bukowski, Carey, et al., 1993; Henry, Brooks, Case, et al., 1995; Kurz, Marth, Windbichler, et al., 1997; Littlewood, Bajetta, Cella, et al., 1999). The sixth study (n=30) was in the category with baseline Hb >10 and <12 g/dL (Wurnig, Windhager, Schwameis, et al., 1996). With two exceptions, lesser quality trials were all unblinded. Quirt, Couture, Pichette, et al., (1996) used placebo, but only published an abstract and did not report additional detail on blinding. The trial reported by Porter, Leahey, Polise, et al. (1996) was double blinded but did not meet the criterion for percentage of patients excluded from analysis. Three additional studies also failed to meet the criterion for percentage of patients excluded from analysis (Markman, Reichman, Hakes, et al., 1993; Oberhoff, Neri, Amadori, et al., 1998; Quirt, Couture, Pichette, et al., 1996). However, taken as a whole, the studies included in this systematic review retained a high proportion of enrolled patients for efficacy analysis (95 percent).

Adequate statistical power was more likely to be a feature of higher quality trials than lesser quality trials. Three of the four studies with more than 100 patients enrolled were higher quality studies (Case, Bukowski, Carey, et al., 1993; Henry, Brooks, Case, et al., 1995; Littlewood, Bajetta, Cella, et al., 1999).

We also assessed studies for additional quality domains and features to control for specific confounders. Of the higher quality studies, three reported adequate concealment of treatment allocation (Case, Bukowski, Carey, et al., 1993; Cascinu, Fedeli, Del Ferro, et al., 1994; Kurz, Marth, Windbichler, et al., 1997). Adequate concealment of allocation was also reported in four lesser quality studies, including the double-blinded trial by Porter (Silvestris, Romito, Fanelli, et al., 1995; Porter, Leahey, Polisse, et al., 1996: ten Bokkel Huinink, de Swart, vanToorn, et al., 1998; Del Mastro, Venturini, Lionetto, et al.. 1997). Five of the six higher quality studies adequately accounted for the reasons subjects were excluded from the analysis of study results. The sixth study by Littlewood, Bajetta, Cella, et al. (1999) is unpublished, and a full report is not yet available. Of the 16 lesser quality studies, 11 were judged to adequately account for patients excluded from analysis; 5 did not.

Of the higher quality studies, all but the one by Littlewood and coworkers (1999), for which a full report is not yet available, reported a transfusion trigger or provided data to demonstrate that Hb at transfusion was comparable for all study arms. Two lesser quality studies reported transfusion outcomes but did not specify a transfusion trigger or show that the mean Hb at transfusion was comparable for epoetin-treated patients and controls. One of these only published an abstract (Quirt, Couture, Pichette, et al., 1996), whereas the second published a more detailed report (Oberhoff, Neri, Amadori, et al., 1998). The question of a transfusion trigger was not applicable to five studies that did not report transfusion outcomes since they could not be included in either the meta-analysis or the sensitivity analysis.

The presence of treatable other causes of anemia might lead to underestimating the effects of epoetin. Three of the higher quality studies reported ruling out anemia caused by iron deficiency, vitamin B₁₂ deficiency, folate deficiency, occult bleeding, and hemolytic anemia prior to randomization (Case, Bukowski, Carey, et al., 1993; Henry, Brooks, Case, et al., 1995; Wurnig, Windhager, Schwameis, et al., 1996). Two lesser quality studies also reported ruling out all these causes of anemia (Leon, Jimenez, Barona, et al., 1998; Sweeney, Nicolae, Ignacio, et al., 1998). Thirteen additional studies reported ruling out at least one of these treatable causes of anemia but did not specify that each of the five potential causes had been ruled out.

In higher quality studies that reported quality-of-life outcomes, the proportion of missing data was much greater for quality-of-life outcomes than for hematologic or transfusion outcomes. Only the study by Kurz, Marth, Windbichler, et al. (1997) reported the same number of patients evaluable for quality-of-life outcomes as were evaluable for transfusion and hematologic outcomes. In all other higher quality studies, data on quality-of-life outcomes were missing for 10 percent or more of enrolled patients from one or both study arms. Thus, with respect to quality-of-life outcomes, double blinding is the only attribute that distinguishes the group of higher quality studies from the lesser quality studies, all of which were unblinded. Moreover, no published studies reported on features known to be important in minimizing bias in assessment of quality-of-life outcomes, including:

• Procedures to minimize the impact of other factors on responses to quality-of-life instruments (e.g., blinding patients to Hb levels or cell counts until after completion of questionnaires);

• Handling of missing data in the analysis; and

• Prospectively defining the minimum differences in quality-of-life scores to be considered clinically significant.

Key Question 1 Results

This section analyzes efficacy outcomes across studies grouped by baseline Hb level. We examine hematologic outcomes, transfusion outcomes, and health-related quality-of-life outcomes. For hematologic and transfusion outcomes, we calculated the difference between the treatment and control arm of each trial, and we report the range of difference for each outcome. For transfusion outcomes, we also conducted a meta-analysis of the odds of transfusion.

Throughout this Results section, "n" refers to the number of patients evaluable. Outcomes tables show higher quality trials in bold font; nonrandomized trials are indicated in italics.

The conclusion to this section addresses two questions:

1. What are the outcomes of managing anemia with epoetin (plus transfusion if necessary) compared with transfusion alone?

2. What are the relative effects of epoetin treatment when different Hb thresholds are used to initiate treatment?

Hematologic Outcomes

Table 15. Hematologic outcomes for studies grouped (more...)

Table 15. Hematologic outcomes for studies grouped by baseline Hb levels¹

Citation	Transf. Trigger or mn Hb @ transf. 2	Baseline Hb	Study Arm	N Enrolled	N Evaluable	EPO Dose (units/kg/week)		% Response	p Value	Difference in % response (epo-control)	Hb Change (± SD)	p Value	Difference in Hb Change (epo-control)
						Start	Final
Mean/Median Baseline Hb <10 g/dL; Adult Patients
Silvestris, Romito, Fanelli, et al., 1995	NA	3	Control	24	22	0		0.0
		3	Epoetin	30	27	450	900	77.8		77.8
Oberhoff, Neri, Amadori, et al., 1998	NA	10.3 4	Control	110	88	0		6.8
		9.6 4	Epoetin	117	101	~450		34.7	0.0001	27.9
Case, Bukowski, Carey, et al., 1993	8.2	9.8	Control	76	74	0		13.5			0.33
	8.2	9.5	Epoetin	81	79	450		58.2		44.7	2.3	0.0001	1.97
Henry, Brooks, Case, et al., 1995	8.5	9.5	Control	65	61	0		6.6			0.45±1.7
	8.2	9.8	Epoetin	67	64	450		48.4	<0.0001	41.8	2.05±2.3	<0.0001	1.60
Cascinu, Fedeli, Del Ferro, et al., 1994	8.0	8.7	Control	50	49	0		2.0			−0.6
		8.6	Epoetin	50	50	300		82.0		80.0	1.9		2.5
Kurz, Marth, Windbichler, et al., 1997	8.0	9.85	Control	12	12	0		0.0			0.22
		9.88	Epoetin	23	23	450	900	56.5	0.001	56.5	3.3		3.08
Littlewood, Bajetta, Cella, et al., 1999(Abstract/slides)	NA	9.7	Control	124	115	0		19.1			0.9
		9.9	Epoetin	251	244	450	900	70.5	0.001	51.4	2.5		1.60
Mean/Median Baseline Hb <10 g/dL; Pediatric Patients
Varan, Buyukpamukcu, Kutluk, et al., 1999	6.0	8.48	Control	17	17	0					−0.07
		8.5	Epoetin	17	17	450					1.71		1.78
Leon, Jimenez, Barona, et al., 1998	6.0	9.5	Control	25	25	0					0.1
		9.8	Epoetin	25	25	750		72.0			2.6	<0.001	2.5
Porter, Leahey, Polise, et al., 1996	8.0	9.4 4	Control	12	10	0
		9.7 4	Epoetin	12	10	450	900
Mean/Median Baseline Hb >10 and <12 g/dL; Adult Patients
Markman, Reichman, Hakes, et al., 1993	8.0	11.14	Control	46	40	0		40.0
		11.54	Epoetin	17	16	350		87.5	<0.005	47.5
Dusenbery, McGuire, Holt, et al., 1994	9.5	11.14	Control	61	61	0					−0.8
		10.34	Epoetin	15	15	1000	500				2.9	0.001	3.70
Lavey and Dempsey, 1993	NA	11.8	Control	20	20	0		5.0			0.0±0.7
		11.9	Epoetin	20	20	900	450	80.0	<0.001	75.0	3.2±1.78	<0.001	3.2
Wurnig, Windhager, Schwameis, et al., 1996	8.5	10.5	Control	14	14	0
		11	Epoetin	16	15	1200						NS
Henke, Guttenberger, Barke, et al., 1999	NA	12.3	Control	11	11	0					0.6±1.4
		10.9	Epoetin 1	19	19	450					3.2±1.6	<0.0001	2.6
		11.4	Epoetin 2	14	14	900					3.5±1.2		2.9
Quirt, Couture, Pichette, et al., 1996(Abstract)	NA	10.7 6	Control	28	27	0					0.6
		10.9 6	Epoetin	28	27	450	900				1.6		1.0
ten Bokkel Huinink, de Swart, van Toorn, et al., 1998	9.7	11.8 4	Control	34	33	0
		12.0 4	Epoetin 1	46	45	450	225
		11.6 4	Epoetin 2	42	42	900	450
Mean/Median Baseline Hb >12; Adult Patients
Gamucci, Thorel, Frasca, et al., 1993	NA	12.7	Control	17	17	0					−1.5±1.67
		12.2	Epoetin	21	21	450					0.9±1.32	<0.005	2.4
Sweeney, Nicolae, Ignacio, et al., 1998	NA	10.7	Control	24	24	0		0.0			0.29
		12.1	Epoetin	24	22	1000	500	45.5		45.5	1.55	0.0012	1.26
Del Mastro, Venturini, Lionetto, et al., 1997	8.0	13.1	Control	31	31	0					−3.1±1
		13	Epoetin	31	31	450					−0.8±1.4	<0.005	2.3
Thatcher, De Campos, Bell, et al., 1999	8.5	13.4 4	Control	44	44	0		34.1			−3.4
	8.6	13.7 4	Epoetin 1	42	42	450	225	52.4	<0.05	18.3	−3.2		0.2
	8.0	13.6 4	Epoetin 2	44	44	900	450	61.4	0.005	27.3	−3.3		0.1
Welch, James, Wilkinson, et al., 1995	8.5	12.8	Control	15	15	0					−2.1
	8.3	13	Epoetin	15	15	900	450				−1.3		0.8

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

2

Single entry = transfusion trigger; multiple entries = mean Hb levels at transfusion.

3

Mean/median Hb level at baseline not specified, but enrollment limited to patients with Hb ≤10 g/dL.

4

The report provided only a median value, not a mean.

5

Change in Hb level calculated as change in hematocrit divided by 3.

6

Did not specify whether reported value is mean or median.

Table 15 summarizes hematologic outcomes: the percentage of patients with a hematologic response and the mean change in Hb level (median if no mean was reported). Our description of results focuses on the percentage of patients responding, which incorporates a minimum level or target Hb change and the proportion of patients achieving it. (See Evidence Table I-1 for each study's definition of response.) Mean change in Hb level is noted for studies where this outcome was statistically significant but percentage of patients responding was not.

Baseline Hb <10 g/dL

Five of seven trials in adults reported a statistically significant hematologic outcome favoring the epoetin arm. All seven trials reported the percentage of patients with a hematologic response. Four trials (n=708), including three of the five higher quality trials (n=519), observed significantly more responses in the epoetin arms than in controls (Henry, Brooks, Case, et al., 1995; Kurz, Marth, Windbichler, et al., 1997; Littlewood, Bajetta, Cella, et al., 1999; Oberhoff, Neri, Amadori, et al., 1998). In addition, a fifth trial, of higher quality, reported a statistically significantly greater change in Hb levels in the epoetin arm than in controls (Case, Bukowski, Carey, et al., 1993). The remaining studies did not report on statistical significance for either hematologic outcome (Cascinu, Fedeli, Del Ferro, et al., 1994; Silvestris, Romito, Fanelli, et al., 1995;). The range of differences between epoetin and control arms for the percentage of patients responding was 28 percent to 80 percent The range of differences was 1.60 to 3.08 g/dL for mean change in Hb levels.

Among the three pediatric trials, Leon, Jiminez, Barona, et al. (1998) reported a statistically significant change in Hb level. However, this trial reported percent of patients responding only in the epoetin arm. Varan, Buyukpamukcu, Kutluk, et al. (1999) reported an increase in Hb level in the epoetin arm but did not report on statistical significance. In the two trials, the differences between arms for change in mean Hb level were 1.78 and 2.5 g/dL. The pediatric trials did not report any other hematologic outcomes.

Baseline Hb >10 and <12 g/dL

Four of seven studies reported a statistically significant hematologic outcome favoring the epoetin arm. Two nonrandomized trials (n=96) reported percentage of patients responding, and the epoetin arm was significantly favored in each (Lavey and Dempsey, 1993; Markman, Reichman, Hakes, et al., 1993). Two additional trials (n=120) reported a significantly greater increase in Hb level (Dusenbery, McGuire, Holt, et al., 1994; Henke, Guttenberger, Barke, et al., 1999). Quirt, Couture, Pichette, et al. (1996) reported a greater increase in the Hb level in the epoetin arm compared with that in the control arm but did not report on significance.

The only higher quality trial (n=29) in this group reported only that there was no significant difference in change in Hb level (Wurnig, Windhager, Schwameis, et al., 1996). The trial by ten Bokkel and colleagues (1998) did not report hematologic outcomes.

In the two studies reporting percentage of patients responding, the differences between epoetin and control arms were 48 percent and 75 percent. For four studies reporting mean Hb change, the range of differences was 1.0 to 3.7 g/dL.

Baseline Hb >12 g/dL

Of the five trials in this group, four (n=276) reported a statistically significant hematologic outcome favoring epoetin. The largest study (n=130; Thatcher, De Campos, Bell, et al., 1999) reported a significantly greater percentage of patients responding in the epoetin-treated arm. Three trials reported a significant difference between study arms for the change in Hb level (Del Mastro, Venturini, Lionetto, et al., 1997; Gamucci, Thorel, Frasca, et al., 1993; Sweeney, Nicolae, Ignacio, et al., 1998). The fifth trial reported a smaller decrease in Hb level in the epoetin arm but did not report on significance (Welch, James, Wilkinson, et al, 1995).

In the two studies reporting percentage of patients responding, the differences between epoetin and control arms were 18 percent and 46 percent. For mean Hb change, the range of differences was 0.1 to 2.4 g/dL.

Summary

We found adequate and consistent evidence that epoetin increases Hb levels and percentage of patients achieving a hematologic response when compared with controls managed by transfusion alone. The percentage of patients achieving a hematologic response and the magnitude of change in Hb levels produced by epoetin does not appear to differ substantially as a function of the baseline Hb level for initiating epoetin treatment. In trials of adult patients where baseline Hb was <10 g/dL, the range of differences between epoetin and control arms for percentage of patients responding was 28 percent to 80 percent. For baseline Hb >10 and <12 g/dL, the differences were 48 percent and 75 percent, and the range of differences was 18 percent to 46 percent for baseline Hb >12 g/dL. For change in mean Hb levels, the range of differences for the three groups of studies was 1.60 to 3.08 g/dL, 1.0 to 3.7 g/dL, and 0.1 to g/dL. Two pediatric studies on patients with baseline Hb <10 g/dL reported change in mean Hb levels; the differences between the epoetin and control arms were similar (1.78 and 2.5 g/dL).

Transfusion Outcomes

Table 16. Transfusion outcomes for studies grouped (more...)

Table 16. Transfusion outcomes for studies grouped by baseline Hb levels¹

Citation	Transf. Trigger or Mn Hb @ Transf. 2	Baseline Hb	Study Arm	N Enrolled	N Evaluable	EPO Dose (units/kg/week)		% Transfused	p Value	Difference in % Transfused (control-Epo)	RBC Units per Patient ± SD	p Value	RBC Units per Patient per 4 Weeks	Difference in RBC Units per Patient per 4 Weeks (control-Epo)
						Start	Final
Mean/Median Baseline Hb <10 g/dL; Adult Patients
Silvestris, Romito, Fanelli, et al., 1995	NA	3	Control	24	22	0
		3	Epoetin	30	27	450	900
Oberhoff, Neri, Amadori, et al., 1998	NA	10.3 4	Control	110	88	0		40.9			0.6		0.6
		9.6 4	Epoetin	117	101	~450		25.7	5	15.2	0.5	0.044	0.5	0.1
Case, Bukowski, Carey, et al., 1993	8.2	9.8	Control	76	74	0		36.86			1.6±0.3		0.8
	8.2	9.5	Epoetin	81	79	450		28.66	NS7	8.56	0.9±0.3	NS	0.5	0.3
Henry, Brooks, Case, et al., 1995	8.5	9.5	Control	65	61	0		68.9			4.0±0.8		2.0
	8.2	9.8	Epoetin	67	64	450		53.1	NS	15.8	4.0±0.9	NS	2.0	0
Cascinu, Fedeli, Del Ferro, et al., 1994	8.0	8.7	Control	50	49	0		57.1			1.8		0.8
		8.6	Epoetin	50	50	300		20.0	0.01	37.1	0.3	0.01	0.1	0.7
Kurz, Marth, Windbichler, et al., 1997	8.0	9.85	Control	12	12	0		66.7			3.6		1.2
		9.88	Epoetin	23	23	450	900	21.7	0.009	45.0	1.4		0.5	0.7
Littlewood, Bajetta, Cella, et al., 1999(Abstract/slides)	NA	9.7	Control	124	115	0		35.76
		9.9	Epoetin	251	244	450	900	236	0.0168	12.76
Mean/Median Baseline Hb <10 g/dL; Pediatric Patients
Varan, Buyukpamukcu, Kutluk, et al., 1999	6.0	8.48	Control	17	17	0		47.1
		8.5	Epoetin	17	17	450		5.9	0.008	41.2
Leon, Jimenez, Barona, et al., 1998	6.0	9.5	Control	25	25	0		96			3.6		1.2
		9.8	Epoetin	25	25	450		16	<0.001	80.0	0.3	<0.001	0.1	1.1
Porter, Leahey, Polise, et al., 1996	8.0	9.4 4	Control	12	10	0		100			13.0 4		3.3
		9.7 4	Epoetin	12	10	450	900	90	NS	10.0	4.5 4	0.01	1.1	2.2
Mean/Median Baseline Hb >10 and <12 g/dL; Adult Patients
Markman, Reichman, Hakes, et al., 1993	8.0	11.14	Control	46	40	0		22.5
		11.54	Epoetin	17	16	350		6.3	NS	16.2
Dusenbery, McGuire, Holt, et al., 1994	9.5	11.14	Control	61	61	0		6.6
		10.34	Epoetin	15	15	1000	500	0.0		6.6
Lavey and Dempsey, 1993	NA	11.8	Control	20	20	0
		11.9	Epoetin	20	20	900	450
Wurnig, Windhager, Schwameis, et al., 1996	8.5	10.5	Control	14	14	0		100			8.4		1.7
		11	Epoetin	16	15	1200		53.3	NS	46.7	2.1	<0.01	0.4	1.3
Henke, Guttenberger, Barke, et al., 1999	NA	12.3	Control	11	11	0
		10.9	Epoetin 1	19	19	450
		11.4	Epoetin 2	14	14	900
Quirt, Couture, Pichette, et al., 1996(Abstract)	NA	10.7 8	Control	28	27	0		29.6			0.7
		10.9 8	Epoetin	28	27	450	900	14.8	NS 7	14.8	0.2
ten Bokkel Huinink, de Swart, van Toorn, et al., 1998	9.7	11.8 4	Control	34	33	0		39.4			1.2		0.2
		12.0 4	Epoetin 1	46	45	450	225	4.4	5	35.0	0.3		0.1	0.1
		11.6 4	Epoetin 2	42	42	900	450	14.3		25.1	0.4		0.1	0.1
Mean/Median Baseline Hb >12; Adult Patients
Gamucci, Thorel, Frasca, et al., 1993	NA	12.7	Control	17	17	0
		12.2	Epoetin	21	21	450
Sweeney, Nicolae, Ignacio, et al., 1998	NA	10.7	Control	24	24	0
		12.1	Epoetin	24	22	1000	500
Del Mastro, Venturini, Lionetto, et al., 1997	8.0	13.1	Control	31	31	0		6.5
		13	Epoetin	31	31	450		0	NS 7	6.5
Thatcher, De Campos, Bell, et al., 1999	8.5	13.4 4	Control	44	44	0		59.1			6.1		0.9
	8.6	13.7 4	Epoetin 1	42	42	450	225	45.2	<0.05	13.9	3.8	<0.01	0.6	0.3
	8.0	13.6 4	Epoetin 2	44	44	900	450	20.5	<0.001	38.6	2.1	<0.001	0.3	0.6
	8.5	12.8	Control	15	15	0		53.3			5.4
Welch, James, Wilkinson, et al., 1995	8.3	13	Epoetin	15	15	900	450	26.7	NS	26.6	4.0	NS

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

2

Single entry = transfusion trigger; multiple entries = mean Hb levels at transfusion.

3

Mean/median Hb level at baseline not specified, but enrollment limited to patients with Hb ≤10 g/dL.

4

The report provided only a median value, not a mean.

5

Calculated odds ratio for transfusion suggests a significant difference, as upper limit of 95 percent confidence interval is <1.0 (see Meta-Analysis).

6

Measured from day 28 to end of study.

7

Calculated odds ratio for transfusion suggests no significant difference, as upper limit of 95 percent confidence interval is <1.0 (see Meta-Analysis).

8

Did not specify whether reported value is mean or median.

Table 16 summarizes transfusion outcomes: the percentage of patients transfused, and the number of RBC units transfused per patient (normalized to 4 weeks on study whenever possible).

Baseline Hb <10 g/dL

Six of seven adult trials, five of which were of higher quality, reported the percentage of patients transfused. All except the trials by Littlewood and colleagues (1999) and Silvestris and colleagues (1995) also reported the number of RBC units transfused per patient, which we normalized to units transfused per patient per 4 weeks for comparative purposes.

In four trials (n=682), including three of higher quality (n=493), significantly fewer patients were transfused in the epoetin arms than in controls (Cascinu, Fedeli, Del Ferro, et al., 1994; Kurz, Marth, Windbichler, et al., 1997; Littlewood, Bajetta, Cella, et al., 1999; Oberhoff, Neri, Amadori, et al., 1998). Oberhoff and colleagues (1998) and Cascinu and colleagues (1994) also reported transfusing significantly fewer RBC units per patient with epoetin treatment. Kurz and colleagues (1997) reported using fewer RBC units per patient in the epoetin arm, but did not report statistical significance. Two trials (n=278), both of higher quality (Case, Bukowski, Carey, et al., 1993; Henry, Brooks, Case, et al., 1995), found no significant difference in percentage of patients or RBC units transfused per patient.

Because the response to epoetin takes several (2 to 4) weeks, and because chemotherapy-related decreases in Hb levels are related to the timing and duration of chemotherapy cycles, the time period selected for reporting can affect whether studies observe significant differences for transfusion outcomes. We believe that outcomes reported from study entry are most useful because reporting in this way facilitates comparison across trials and also gives the best picture of the results of initiating treatment at different baseline Hb levels. Of the 22 trials included in this systematic review of evidence, only three studies reported transfusion outcomes from the 28^th day of epoetin treatment. Henry, Brooks, Case, et al. (1995) reported that the difference in percentage of patients transfused was significant over the 2^nd and 3^rd months of the trial (26 percent versus 56 percent, p<0.005), even though it was not significant over the entire 3-month duration of the trial. Case, Bukowski, Carey, et al. (1993) reported the percentage of patients transfused at month 1 and at months 2 and 3, but the difference between the epoetin and control arms was not significant for either time interval. To date, Littlewood, Bajetta, Cella, et al. (1999) have reported transfusion outcomes only after the 28^th day of epoetin administration; thus, it is not known whether the difference was significant over the entire length of the study.

In the adult trials of higher quality, the range of differences between epoetin and control arms for percentage of patients transfused was 9 percent to 45 percent; the lesser quality study by Oberhoff and colleagues (1998) fell within this range (15 percent). The range of differences for RBC units transfused per patient per 4 weeks was 0 to 0.7.

Two of the three pediatric trials reported that significantly fewer epoetin-treated patients were transfused than were controls (Leon, Jimenez, Barona, et al., 1998; Varan, Buyukpamukcu, Kutluk, et al., 1999). The third trial, by Porter, Leahy, Polise, et al. (1996), had an unusually high percentage of patients transfused in each arm, and there was no significant difference between arms in the percentage of patients transfused. The trials by Leon, Jiminez, Barone, et al. (1998) and Porter and colleagues (1996) each reported a significant reduction in RBC units transfused per patient; Varan, Buyukpamukcu, Kutluk, et al. (1999) did not report on this outcome. The range of differences between epoetin and control arms for percentage of patients transfused was 10 percent to 80 percent. The differences between arms for RBC units transfused per patient per 4 weeks were 1.1 and 2.2 units; note that 2.2 units, from the Porter study, is the difference in median (not mean) units transfused.

Baseline Hb >10 to <12 g/dL

Five of seven trials (n=335), including one of higher quality and two nonrandomized trials, reported the percentage of patients transfused. Three of these studies also reported RBC units transfused per patient.

ten Bokkel Huinink, de Swart, van Toorn, et al. (1998; n=120) observed a significant reduction in the percentage of patients transfused for the epoetin-treated study arm but did not report whether the difference in RBC units transfused per patient was significant. A second trial (n=29), the only one of higher quality, reported that epoetin significantly reduced the number of RBC units transfused per patient (Wurnig, Windhager, Schwameis, et al., 1996). However, the difference between arms in percentage of patients transfused was not significant.

Of the remaining three trials, two (n=110) found nonsignificant differences between arms in percentage of patients transfused (Markman, Reichman, Hakes, et al., 1993; Quirt, Couture, Pichette, et al., 1996). Quirt and colleagues (1996) also reported RBC units transfused per patient but did not report on significance. The third study (n=76) did not report whether the difference in percentage of patients transfused was significant (Dusenbery, McGuire, Holt, et al., 1994).

Among these trials, the range of differences between epoetin and control arms for percentage of patients transfused was 7 percent to 47 percent. The differences for RBC units transfused per patient per 4 weeks were 0.1 and 1.3 units.

Baseline Hb >12 g/dL

Three (n=222) of five studies reported transfusion outcomes.⁵ The trial by Thatcher, De Campos, Bell, et al. (1999) reported significantly fewer patients transfused in the epoetin-treated arm than in controls; the other two trials found no significant differences. The number of RBC units per patient was also significantly lower for the epoetin arm in the Thatcher trial, but differences were not significant in the trial by Welch, James, Wilkinson, et al. (1995). Among these trials, the range of differences between epoetin and control arms for percentage of patients transfused was 7 percent to 39 percent. The difference for RBC units transfused per patient per 4 weeks was 0.3 units and 0.6 units in the lower and higher dose epoetin arms of the Thatcher trial.

The study by Thatcher and coworkers (1999) is noteworthy because patients in all three arms experienced the greatest decrease in Hb levels among all studies included in this systematic review of evidence. Most notably, this is the only study with baseline Hb >10 g/dL in which the epoetin arms approached Hb=10 g/dL during the course of the study. Patients in this study were undergoing platinum-based chemotherapy for small cell lung cancer, a population selected by the authors as being at high risk of transfusion. However, the authors also noted that the tendency of physicians in this unblinded trial to undertransfuse the patients in the arm with the higher epoetin dose may overestimate the magnitude of effect.

Summary

The most robust evidence that epoetin reduces transfusions for patients undergoing therapy for malignancy comes from trials in patient groups with baseline Hb <10 g/dL. Four trials in adults (n=682), including three of higher quality (n=493), reported that significantly fewer patients were transfused in the epoetin arms than in control arms. The three pediatric trials all reported a significant reduction in either percentage of patients transfused or RBC units transfused per patient or both.

Among trials of patients with baseline Hb >10 g/dL, two of eight reported that significantly fewer epoetin-treated patients were transfused, and a third reported a significant reduction in RBC units transfused per patient. The total number of patients in these studies is much smaller than in the Hb <10 g/dL group, individual trials are small in size, and the patients' risk of transfusion may be lower. It is likely that most of these trials lack adequate statistical power to detect a difference. However, of these 12 trials, only 1 was double blinded, and 3 were nonrandomized; these study designs may overestimate the magnitude of effect of epoetin treatment. Thus, bias inherent in the weaker design of these studies may somewhat offset inadequate power because of small sample size.

Relative Effects of the Hemoglobin Threshold for Initiating Treatment on Transfusion Outcomes

No trial directly compared the outcomes of initiating epoetin treatment at alternative Hb thresholds. Thus, only inferences based on indirect comparison are possible as to whether initiating epoetin at one or another Hb threshold results in superior outcomes. Transfusion outcomes do not appear to be superior in trials where epoetin treatment is initiated in groups of patients who have mean Hb >10 g/dL compared with trials where mean Hb is <10 g/dL.

Although it is possible that adequately powered comparative trials might demonstrate the superiority of earlier epoetin intervention, our examination of this evidence base suggests why that may not necessarily prove to be true. First, patients whose baseline Hb level is well below the mean may account for a substantial proportion of transfusions in epoetin-treated patients in trials where baseline Hb is <10 g/dL. Second, in all trials, patients who do not respond to epoetin may account for a substantial proportion of transfusions irrespective of the Hb level at which epoetin treatment is initiated. If these two patient groups account for a substantial proportion of epoetin-treated patients who are transfused, then the greatest yield for reducing the number of patients transfused might come from preventing the Hb level from falling well below 10 g/dL, rather than from setting an Hb threshold well above 10 g/dL for initiating epoetin treatment.

Indirect comparisons

The range of reduction reported in patients transfused and RBC units per patient does not appear to be greater in trials where epoetin treatment was initiated when mean Hb was >10 g/dL compared with trials where mean Hb at epoetin initiation was <10 g/dL. Among trials of adults where baseline Hb was <10 g/dL, the range of differences between epoetin and control arms for percentage of patients transfused was 9 percent to 45 percent. For baseline Hb >10 to <12 g/dL, the range was 7 percent to 47 percent; the range was 7 percent to 39 percent for baseline Hb >12 g/dL. For RBC units per patient per 4 weeks, the differences for the three groups of studies were 0 to 0.7, 0.1 and 1.3, and 0.3 and 0.6.

Distribution of baseline Hb in studies with mean <10 g/dL

Classification of studies by mean Hb at baseline might obscure clinically important variance around that mean within studies in each category. A particular concern is whether, in the group of studies for which baseline Hb was <10 g/dL, the classification might underestimate the benefit of epoetin to patients who had baseline Hb levels >10g/dL. The most useful evidence to address this concern would be data on the likelihood of transfusion as a function of baseline Hb for patients in each study. However no study reported such data; the best available data are from trials that reported SD in addition to mean Hb at study entry.

Table 17. Transfusion rates of epoetin-treated patients (more...)

Table 17. Transfusion rates of epoetin-treated patients and variance in baseline hemoglobin¹

Study (N evaluable, epoetin arm)	Hb Cutoff for Eligibility (g/dL)	Mean Hb (g/dL) at Study Entry (± SD)	Transf. Trigger or Mean Hb at Transf. (g/dL)	Mean Hb at Entry−1 SD (g/dL)	Mean Hb at Entry +1 SD (g/dL)	Final Hb Level (g/dL, ±SD)	% Transfused	Δ % Transfused (control-Epo)
Baseline Hb >10 g/dL; Adult Patients
Cascinu, Fedeli, Del Ferro, et al., 1994 (50)	>11.0 before, <9.0 during chemotherapy	8.6 ± 0.62	<8.0	7.98	9.22	10.5 ± 0.9	20%	38%
Henry, Brooks, Case, et al., 1995 (64)	<10.67	9.8 ± 1.3	8.2	8.5	11.1	11.8 ± 2.3	53.1% (mo 1-3) 26.8% (mo 2-3)	15.8% (mo 1-3) 29.6% (mo 2-3)
Kurz, Marth, Windbichler, et al., 1997 (23)	<11.0	9.88 ± 0.8	<8.0 (or clinical symptoms)	9.08	10.68	13.1	21.7%	45%
Littlewood, Bajetta, Cella, et al., 1999 (244)	<10.5 or (Hb decrease >1.5 and Hb <12)	9.9 ± 1.14	NA (8.2 ± 0.91 before study entry)	8.76	11.04	12.4 ± 0.6	23.0% (after day 28)	12.7% (after day 28)
Baseline Hb <10 g/dL; Pediatric Patients
Varan, Buyukpamukcu, Kutluk, et al., 1999 (17)	<10.0	8.5 ± 0.85	<6.0	7.65	9.35	10.21 ± 2.14	5.9%	41.2%
Leon, Jimenez, Barona, et al., 1998 (25)	<10.5	9.8 ± 0.6	<9.0	9.20	10.4	12.4 ± 1.7	16.0%	80%
Baseline Hb >10 g/dL; Adult Patients
Wurnig, Windhager, Schwameis, et al., 1996 (15)	<11.0	11.0 ± 1.5	<8.5	9.5	12.5	??	53.3%	46.7%
Del Mastro, Venturini, Lionetto, et al., 1997 (31)	>12.0	13 ± 0.7	<8.0 (or clinical symptoms)	12.3	13.7	12.2 ± 1.2	0%	6.5%

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

These trials are summarized in Table 17. Of the eight trials in this evidence base that reported SD in addition to mean Hb at study entry, there are six trials in adults and two trials in pediatric patients. Of the trials in adults, four are in the baseline Hb <10 g/dL category, and there is one each of the higher baseline Hb classifications. Five of the trials in adults are of higher quality. All trials in pediatric patients had baseline Hb <10 g/dL.

In two of the four adult studies with mean Hb at study entry <10 g/dL, patients had baseline Hb one SD or more below the mean border on the Hb level that triggers transfusion. Henry and colleagues (Henry, Brooks, Case, et al., 1995) reported that patients in their trial were transfused at a mean Hb of 8.2 g/dL, while 1 SD below the mean Hb at entry was 8.5 g/dL. In the trial by Cascinu and colleagues (1994), 1 SD below the mean Hb at entry was less than the transfusion trigger (7.98 vs. <8.0 g/dL). The largest difference between the transfusion trigger and 1 SD below the mean Hb at entry is <8.0 versus 9.08 g/dL in the study by Henry and coworkers (1995). Littlewood and colleagues (1999) only report mean Hb at transfusion for those given prior to study entry (8.2 +/− 0.91 g/dL); this can be compared with 1 SD below the mean Hb at entry of 8.76 g/dL. The percentage of epoetin-treated patients transfused in these studies ranged from 20 percent to 53 percent. This suggests, but does not demonstrate, that patients entering these trials with Hb levels well below the mean could largely or entirely account for transfusions in the epoetin-treated arm.

Nonresponders and transfusion

Table 18. Comparison of nonresponse rates and transfusion (more...)

Table 18. Comparison of nonresponse rates and transfusion rates of epoetin-treated patients¹

Citation	Transf. Trigger or Mean Hb @ Transf.	Baseline Hb	N Enrolled	N Evaluable	Percent Not Responding	Percent Transfused	Difference (% not responding - % transfused)
Baseline Hb <10 g/dL
Oberhoff, Neri, Amadori, et al., 1998	NA	9.6	117	101	65.3	25.7	39.6
Case, Bukowski, Carey, et al., 1993	8.2	9.5	81	79	41.8	25.3	16.5
Henry, Brooks, Case, et al., 1995	8.2	9.8	67	64	51.6	53.1	−1.5
Cascinu, Fedeli, Del Ferro, et al., 1994	8.0	8.6	50	50	18.0	20.0	−2.0
Kurz, Marth, Windbichler, et al., 1997	8.0	9.88	23	23	43.5	21.7	21.8
Littlewood, Bajetta, Cella, et al., 1999(Abstract/slides)	NA	9.9	251	244	29.5	23.0	6.5
Leon, Jimenez, Barona, et al., 1998 (pediatric patients)	6.0	9.8	25	25	28.0	16.0	12.0
Baseline Hb >10 g/dL
Markman, Reichman, Hakes, et al., 1993	8.0	11.5	17	16	12.5	6.3	6.2
Thatcher, De Campos, Bell, et al., 1999 (3-arm study)	8.6	13.7	42	42	47.6	45.2	2.4
	8.0	13.6	44	44	38.6	20.5	18.1

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

Table 18 compares the percentage of epoetin-treated patients who failed to achieve a hematologic response with the percentage of epoetin-treated patients transfused. Nine trials reported data on both outcomes. Definitions of hematologic response were consistent among these studies. Six of the nine trials defined response as an increase in Hb level >2 g/dL. Two studies defined response as achievement of Hb levels >10 g/dL without transfusion (Cascinu, Fedeli, Del Ferro, et al., 1994; Thatcher, De Campos, Bell, et al., 1999) and one gave no definition (Markman, Reichman, Hakes, et al., 1993).

The data displayed in Table 18 suggest, but do not demonstrate, that patients who are unresponsive to epoetin may account for a substantial proportion of epoetin-treated patients transfused irrespective of the Hb level at which epoetin treatment is initiated. Only two of the nine trials reported that the percentage of patients transfused was greater than the percentage of patients failing to achieve a hematologic response, and the differences are quite small (1.5 percent and 2 percent) (Cascinu, Fedeli, Del Ferro, et al., 1994; Henry, Brooks, Case, et al., 1995;). Seven of the nine trials reported that the percentage of nonresponding patients was greater than the percentage of patients transfused in the epoetin-treated arm; the range of the differences was 2.4 percent to 39.6 percent.

One limitation of this comparison is that seven of the nine trials that reported on both the outcomes of interest were studies of patients with baseline Hb <10 g/dL. It is possible that Hb levels were too low at study entry, or that the study duration was too short, to adequately demonstrate hematologic response to epoetin. If this were the case, the trials would overestimate the percentage of patients failing to achieve hematologic response. However, the two trials that reported that the percentage of epoetin-treated patients transfused was greater than the percentage of nonresponders also were among the baseline Hb <10 g/dL group of studies. Thus, we believe our point is germane despite the limitations of the available data.

Summary

While it is possible that adequately powered comparative trials might demonstrate the superiority of epoetin intervention at the higher Hb levels, our examination of this evidence base suggests why that may not prove to be true. First, patients whose baseline Hb is below the mean may account for a substantial proportion of transfusions in epoetin-treated patients in trials where baseline Hb is <10 g/dL. Thus the greatest yield for reducing the number of patients transfused in this population might come from initiating epoetin before the Hb level falls substantially below 10 g/dL, rather than by initiating epoetin treatment at a level substantially above 10 g/dL. Second, in all trials, patients who are unresponsive to epoetin may account for a substantial proportion of patients transfused. Initiating epoetin treatment at a higher Hb level is not expected to reduce transfusions in this subgroup of patients.

Meta-Analysis of Odds of Transfusion

Odds of transfusion

Table 19. Studies selected for meta-analysis of the (more...)

Table 19. Studies selected for meta-analysis of the risk of transfusion¹

Citation	Baseline Hb (Control/Treated)	Epoetin Weekly Dose (U/kg)	Odds Ratio 2
Studies of sc epoetin administration
Quirt, Couture, Pichette, et al., 1996	10.7/10.9	450	0.414
Oberhoff, Neri, Amadori, et al., 1998	10.3/9.6	~450	0.501
Kurz, Marth, Windbichler, et al., 19973	9.8/9.9	450	0.139
Del Mastro, Venturini, Lionetto, et al., 1997	13.1/13	450	0.0
Thatcher, De Campos, Bell, et al., 1999	13.4/13.7/13.6	450/900	0.334
Case, Bukowski, Carey, et al., 19933	9.8/9.5	450	0.664
Henry, Brooks, Case, et al., 19953	9.5/9.8	450	0.513
Welch, James, Wilkinson, et al., 1995	12.8/13	900	0.319
Varan, Buyukpamukcu, Kutluk, et al., 1999	8.5/8.5	450	0.073
Littlewood, Bajetta, Cella, et al., 19993	9.7/9.9	450	0.538
Cascinu, Fedeli, Del Ferro, et al., 19943	8.7/8.6	300	0.188
ten Bokkel Huinink, de Swart, van Toorn, et al., 1998	11.8/12/11.6	450/900	0.156
Studies of iv epoetin administration
Porter, Leahey, Polise, et al., 19964	9.4/9.7	450	0.122
Wurnig, Windhager, Schwameis et al., 19964	10.5/11	1200	0.011

1

"Higher quality" trials in bold font; randomized trials only.

2

Odds of transfusion for epoetin-treated patients relative to the odds of transfusion for control patients.

3

Met the criteria for higher quality studies (see Methodology, Assessing Study Quality, and following discussion).

4

Used intravenous epoetin delivery; all other studies used subcutaneous epoetin delivery.

The effect of epoetin on the risk of RBC transfusion in patients with anemia resulting from cancer therapy was estimated using the meta-analysis methods described in Chapter 2, Methodology. To obtain the most accurate estimate of the odds of transfusion, we selected from the 22 studies included in this systematic review only randomized, controlled studies that reported numbers of patients transfused (n=1,439; Table 19).

The results from each study were summarized as ratios of the odds of transfusion for epoetin-treated patients relative to the odds of transfusion for control patients (Table 19). The study odds ratios were combined into a summary estimate using an empirical Bayesian random effects model. The combined odds ratio was 0.370 (95 percent CI, 0.273 to 0.502) in favor of epoetin treatment. The odds ratio is an expression of the relative likelihood that epoetin-treated patients will be transfused compared with the likelihood for control patients. Because most of the included randomized, controlled trials (12 of 14) used the subcutaneous route of delivery and only two randomized, controlled trials used intravenous delivery, data were insufficient to evaluate potential differences as a function of route. The meta-analysis was restricted to studies that delivered epoetin subcutaneously (n=1,390).

Figure 1. Effect of Epoetin on the Odds of Transfusion (more...)

   Figure 1. Effect of Epoetin on the Odds of Transfusion in Patients with Anemia due to Cancer Therapy

A test for homogeneity (DerSimonian and Laird, 1986) indicated that the studies are somewhat heterogeneous (chi-square of 17.306 for 11 degrees of freedom, p=0.0991). Meta-analysis using the same empirical Bayesian random effects model resulted in a combined odds ratio of 0.380 (95 percent CI, 0.282 to 0.513), indicating that the intravenous studies had little effect on the summary estimate, and the results are shown in Figure 1.

Sensitivity analysis by study quality

Table 20. Effect of study quality on the risk of transfusion: (more...)

Table 20. Effect of study quality on the risk of transfusion: 450 Weekly Dose Category

Factor	Coefficient	Standard Error	Odds Ratio 1	95% Confidence Interval
Lower quality: 300-450 dose	−1.985	0.420	0.137	0.060, 0.313
Higher quality: 300-450 dose	−0.793	0.161	0.453	0.330, 0.621

1

Odds of transfusion for epoetin-treated patients relative to the odds of transfusion for control patients.

Figure 2. Effect of Study Quality on the Odds Ratio (more...)

   Figure 2. Effect of Study Quality on the Odds Ratio of Transfusion

A sensitivity analysis was performed using only the studies that met our criteria for higher quality (see Chapter 2, Methodology). All five higher quality studies (Table 19; n=771) that administered epoetin subcutaneously and reported the number of patients transfused used a weekly epoetin dose between 300 and 450 U/kg. The estimate of epoetin effect on transfusion risk in studies using subcutaneous administration and a weekly epoetin dose of 300 to 450 U/kg (n=503) was calculated for higher quality studies and separately for the remaining studies. The results of this analysis are shown in Table 20 and Figure 2. Higher quality studies show a significantly smaller effect of epoetin on the risk of transfusion than do lower quality studies; 95 percent confidence intervals for the odds ratios do not overlap.

Effects of baseline Hb

We also wished to determine if the effect of epoetin on the odds of transfusion depended on the patients' mean baseline Hb. However, evaluating the effect of baseline Hb on the odds ratio for transfusion could be confounded by differences in study quality. Analyzing the effect of baseline Hb level within higher quality studies was not possible because all higher quality studies that gave epoetin subcutaneously enrolled patient groups with mean baseline Hb <10 g/dL. Therefore, the evidence does not allow us to test for an effect of baseline Hb level on the odds of transfusion or to determine by meta-analysis whether there is greater benefit from initiating epoetin treatment at higher Hb levels.

Summary

Table 21. Summary: Meta-Analysis of the effect of epoetin (more...)

Table 21. Summary: Meta-Analysis of the effect of epoetin on transfusion

Analysis of:	Odds Ratio 1	95% Confidence Interval	Number Needed to Treat	95% Confidence Interval
All randomized studies, subcutaneous epoetin delivery	0.380	0.282, 0.513	4.4	3.6, 6.1
All randomized studies, subcutaneous epoetin delivery, higher quality (300-450 weekly dose)	0.453	0.330, 0.621	5.2	3.8, 8.4
All randomized studies, subcutaneous epoetin delivery, lower quality (300-450 weekly dose)	0.137	0.060, 0.313	2.6	2.1, 3.8

1

Odds of transfusion for epoetin-treated patients relative to the odds of transfusion for control patients.

Estimates of the odds of transfusion for epoetin-treated patients relative to the odds of transfusion for control patients for different groups of studies are summarized in Table 21. For all randomized trials that delivered epoetin subcutaneously, the odds of transfusion for epoetin-treated patients was reduced by a factor of 0.380 compared with patients supported with transfusion alone.

The estimated magnitude of effect was smaller when we analyzed only the higher quality studies that used subcutaneous delivery. Since higher quality studies all delivered weekly epoetin doses of 300 to 450 U/kg, the estimated odds ratio was compared to those lower quality studies that used epoetin doses within the same range. The odds of transfusion for epoetin-treated patients was reduced by a factor of 0.453 for higher quality studies compared with a reduction by a factor of 0.137 for lower quality studies. Lack of double-blinding in the lesser quality trials could bias results toward a larger intervention effect.

To express the absolute effectiveness of epoetin, for each estimate, we also calculated the number of patients that would need to be treated to prevent one patient from being transfused (number needed to treat, NNT). The overall NNT calculated for this group of studies is 4.4 (95 percent CI, 3.6 to 6.1; Table 20).⁶ Therefore, four to five patients would need to receive epoetin for one patient to avoid transfusion. For higher quality studies, the calculated NNT is 5.2 (95 percent CI, 3.8-8.4); and for lower quality studies, the calculated NNT is 2.6 (95 percent CI, 2.1-3.8). Thus, the higher quality studies predict one patient would avoid transfusion for every five to six patients treated with epoetin, whereas the lesser quality studies predict one for every two to three treated.

The effect of baseline Hb level may also be confounded by study quality, and there is insufficient evidence to separately determine whether or not there is greater benefit from epoetin therapy depending on the Hb level at the time epoetin treatment is initiated.

Quality of Life

Table 22. Quality of life outcomes for studies grouped (more...)

Table 22. Quality of life outcomes for studies grouped by baseline Hb levels: Comparisons between control and epoetin-treated study arms

Study	Treatment Arm	N Evaluable for Transfusion	N Evaluable for Quality of Life	% Change: Overall QoL	p Value: Overall QoL	% Change: Energy Level	p Value: Energy Level	% Change: Daily Activities	p Value: Daily Activities	Other QoL Measure 2	% Change: Other QoL	p Value: Other QoL
Mean/Median Baseline Hb < 10g dL
Kurz, Marth, Windbichler, et al., 1997	Control	12	12					−14.5		Well-being	4.0
	Epoetin	23	23					−6.5	NS	Well-being	−0.1	NS
	Control	12	12							Physical ability	8.0
	Epoetin	23	23							Physical ability	8.3	NS
	Control	12	12							Social activities	12.8
	Epoetin	23	23							Social activities	1.0	NS
Henry, Brooks, Case, et al., 1995	Control	61	40	0.2		6.2		0.7
	Epoetin	64	46	11.0	0.013	8.8	NS	8.2	NS
Littlewood, Bajetta, Cella, et al., 1999	Control	115	108	NA		−5.8		−6.0
	Epoetin	244	227	NA	<0.01	7.8	<0.001	7.3	<0.01
	Control	115	90							Fact-An: Anemia	−9.4
	Epoetin	244	200							Fact-An: Anemia	14.4	<0.01
	Control	115	90							Fact-An: Fatigue	−4.2
	Epoetin	244	200							Fact-An: Fatigue	5.7	<0.01
	Control	115	?							SF-36	NA
	Epoetin	244	?							SF-36	NA	NS
Leon, Jimenez, Barona, et al., 1998	Control	25	25							Karnovsky PS	1.4
	Epoetin	25	25							Karnovsky PS	8.6	<0.05
Mean/Median Baseline Hb >12 g/dL
Sweeney, Nicolae, Ignacio, et al., 1998	Control	24	24	6.3
	Epoetin	22	22	19.1	0.15, NS
Welch, James, Wilkinson, et al., 1995	Control	15	?15	NA		NA		NA
	Epoetin	15	?15	NA	NS	NA	NS	NA	NS
Del Mastro, Venturini, Lionetto, et al., 1997	Control	31	26							PDI score	2.3
	Epoetin	31	27							PDI score	6.0	NS

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

2

In order to accommodate several "Other" QoL instruments or different statistical testing results, study control and treatment arms may be listed more than once.

Table 23. Quality of life outcomes for studies grouped (more...)

Table 23. Quality of life outcomes for studies grouped by baseline Hb levels: Within-arm comparisons ofbaseline to final values¹

Study	Treatment Arm	N Evaluable for Transfusion	N Evaluable for Quality of Life	% Change: Overall QoL	p Value: Overall QoL	% Change: Energy Level	p Value: Energy Level	% Change: Daily Activities	p Value: Daily Activities
Mean/Median Baseline Hb < 10g dL
Case, Bukowski, Carey, et al., 1993	Control	74	61	−1.0	NS	3.0	NS	1.6	NS
	Epoetin	79	63	5.6	0.09, NS	10.0	<0.05	8.0	<0.05
Henry, Brooks, Case, et al., 1995	Control	61	40	0.2	NS	6.2	< 0.05	0.7	NS
	Epoetin	64	46	11.0	< 0.05	8.8	< 0.05	8.2	< 0.05
Mean/Median Baseline Hb >12 g/dL
Thatcher, De Campos, Bell, et al., 1999	Control	44	27	7.5	NS	1.6	NS	10.8	NS
	Epoetin	42	33	11.7	<0.05	−2.3	NS	3.0	NS
	Epoetin	44	32	6.0	NS	3.2	NS	4.9	NS

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

Nine trials reported on quality of life. Table 22 summarizes the seven studies (three of four higher quality studies) that compared differences in pre- and posttreatment quality-of-life scores between treatment and control study arms. In addition, two studies (Case, Bukowski, Carey, et al, 1993; Thatcher, De Campos, Bell, et al., 1999) only compared pre- and posttreatment results within each study arm, so that the results of epoetin versus no epoetin treatment are not directly compared. One study (Henry, Brooks, Case, et al., 1995) reported both within and between arm comparisons. Table 23 summarizes the three studies that compare pre- and posttreatment results within study arms.

No published studies reported on features known to be important in minimizing bias in assessment of quality-of-life outcomes, including: procedures to minimize the impact of other factors on responses to quality-of-life instruments, handling of missing data in the analysis, and prospectively defining the minimum differences in quality of life scores to be considered clinically significant.

LASA scales are the most frequently used quality-of-life instruments, reported in seven trials. The trial by Littlewood, Bajetta, Cella, et al. (1999) used the FACT-An instrument in addition to LASA scales. Of the two trials that did not use LASA scales, one used Karnovsky performance scales (Leon, Jimenez, Barona, et al., 1998) and one used the Psychological Distress Inventory (Del Mastro, Venturini, Lionetto, et al., 1999). The LASA scale results most frequently reported in these studies are: overall quality of life;, energy level, and daily activities. Kurz, Marth, Windbichler, et al. (1997) used a 10-question LASA scale, which included a question on daily activities. Three other LASA results, considered most important by the authors, were also reported for this study in Table 22. Because the LASA scales are unidimensional instruments, the results of various scales cannot be combined; thus the overall quality-of-life score is not a summary score and is only one of three or more separate dimensions.

Five of the nine studies (n=630 patients evaluable for quality-of-life outcomes) enrolled patient groups with a mean or median Hb level <10 g/dL at baseline; four studies (n=221) enrolled patient groups with a mean or median Hb level >12 g/dL at baseline. There were no studies reporting quality-of-life outcomes that enrolled patient groups with mean or median baseline Hb levels >10 and <12 g/dL.

Baseline Hb >12 g/dL

The three studies that compared the epoetin-treated and control arms with respect to change in quality of life from baseline to final evaluation found no significant differences. Neither Sweeney, Nicolae, Ignacio, et al. and coworkers (1998) nor Welch, James, Wilkinson, et al. (1995) detected significant differences between arms in the magnitude of change in overall quality of life measured on a visual analog scale. Nor were changes in energy level or daily activities significant in the Welch and colleagues (1995) study. Del Mastro, Venturini, Lionetto, et al. (1997) found no significant differences using the Psychological Distress Index.

Thatcher, De Campos, Bell, et al. (1999) used visual analog scales to evaluate the changes between baseline and final evaluation within study arms but did not compare changes between study arms. The only significant difference reported was for overall quality of life in the study arm treated with epoetin at 450 U/kg per week; however, the change in the epoetin arm treated at 900 U/kg per week was not significant. Changes from baseline within each study arm were not significant for energy level and daily activities. By visual inspection of results, it appears unlikely that there are significant differences for any quality-of-life measures between control and treatment arms.

The four studies that reported on quality-of-life outcomes in patient groups with an average baseline Hb of >12 were relatively small, mainly used visual analog scales to assess quality of life, and reported no significant differences in quality-of-life change between control and epoetin-treated study arms. Thus, for studies of patient populations with baseline Hb >12, no conclusions can be drawn with respect to the effect of epoetin treatment on quality of life.

Baseline Hb <10 g/dL

Four of five studies in this group that reported quality-of-life outcomes were considered higher quality studies according to criteria described in Chapter 2, Methodology. However, of the four higher quality studies, three reported fewer evaluable patients for quality-of-life outcomes than for hematologic or transfusion outcomes (Case, Bukowski, Carey, et al., 1993; Henry, Brooks, Case, et al., 1995; Littlewood, Bajetta, Cella, et al., 1999). Ten to 40 percent of enrolled patients were not evaluable for quality of life; therefore, these studies would fail our study quality criterion of <10 percent nonevaluable patients per study arm. Particularly with quality-of-life outcomes, missing data may not be distributed randomly and may be related to the quality-of-life outcome. For example, patients who are most severely ill or who have the worst quality of life may be most likely to have missing quality-of-life data, potentially biasing the analysis of study results.

One of these studies, which enrolled 375 patients with a variety of hematologic and solid tumors who were receiving nonplatinum-based chemotherapy, is the largest study included in this systematic review and is likely to offer the most robust data on quality of life (Littlewood, Bajetta, Cella, et al., 1999). However, as of this writing, the Littlewood study is unpublished and information is available only in an abstract and in copies of a slide presentation from the May 1999 ASCO meeting (generously supplied by the authors).

Littlewood and coworkers (1999) report consistent and statistically significant differences between study arms, favoring the epoetin-treated arm, on each of several instruments, including the visual analog scales on overall quality of life, energy level, and daily activity employed in other studies. In addition, the overall FACT-An scale and FACT-F subscale, which specifically target issues related to anemia and cancer, showed significant differences between study arms. The SF-36 scale, a widely used tool for functional health status, however, did not reveal significant differences between study arms according to the study abstract.

As of this writing, a complete report of the Littlewood trial is not available, so a meaningful interpretation of the results is not possible. Although quality-of-life data using the LASA instrument are missing for relatively few patients (10 percent of the epoetin arm and 13 percent of controls), data are missing for twice as many patients for the FACT-An instrument (20 percent of the epoetin arm and 27 percent of control). No sensitivity analysis or discussion of the implications of the missing data is available. Nor is the complete study protocol available to allow evaluation of the methods used to minimize bias in the collection of quality-of-life data. Interpretation of the clinical significance of changes in quality-of-life scores in the various instruments in this trial also awaits study publication. For example, the information to calculate effect sizes is not given in the published meeting abstract or presentation slides. Further analysis of the FACT-An results in relation to the SF-36 results would also be useful.

Henry, Brooks, Case, et al. (1995), reporting on a subset of patients with a variety of hematologic and solid tumors receiving platinum-based chemotherapy, compared study arms using responses to three questions on independent 100-point visual analog scales. In this study, quality-of-life data were missing for 43 percent and 40 percent of patients in the epoetin-treated and control arms, respectively. Possible bias as a result of missing data was not explored in the publication. From the available data, the control and epoetin-treated arms differed significantly in the magnitude of change for response scores only for the question on overall quality of life but not for the questions on energy level or daily activity. Interestingly, for the questions on overall quality of life and daily activities, baseline and final value within a study arm differed significantly for the epoetin-treated arm but not for the control arm. Thus, significance for comparison between study arms cannot be inferred from within-arm comparisons.

Case, Bukowski, Carey, et al. (1993) reported on the same study for those patients receiving nonplatinum-based chemotherapy, but compared only baseline and final evaluations within study arms and did not compare changes between study arms. Of enrolled patients, data were missing for 22 percent and 20 percent of epoetin-treated and control patients, respectively. The study publication did not include an evaluation of the possibility that nonrandom distribution of the missing data might bias the analysis of results on quality-of-life outcomes. For the available data, the epoetin-treated arm showed a significant difference in baseline to final value for questions on energy level and daily activities. None of the changes in the control arm was statistically significant.

Kurz, Marth, Windbichler, et al. (1997) compared the change from baseline to final quality-of-life evaluation between study arms using patient responses to a five-point visual analog scale for several different questions; those considered most important by the authors are reported in Table 22. All enrolled patients were evaluated; none of the differences was significant. The last study (Leon, Jimenez, Barona, et al., 1998), a nonrandomized trial of 50 pediatric patients, reported only Karnovsky performance scores comparing changes between baseline and final evaluations across study arms for all enrolled patients and found a significant difference.

Summary

The strongest evidence to date on quality-of-life outcomes with epoetin treatment is an unpublished study by Littlewood, Bajetta, Cella, et al (1999), which enrolled patient populations with baseline Hb levels <10 g/dL. This large, randomized controlled trial compared the change in quality-of-life scores between control and epoetin-treated study arms from visual analog scales and from the FACT-An and found positive, significant differences. Results from the SF-36 were in the same direction but not significant. Interpretation of the clinical significance of these findings awaits publication of a complete report. As of this writing, information is not available assessing the key features for evaluation of a quality-of-life study, including adequacy of methods used to minimize bias in the collection of quality-of-life data, the impact of missing quality-of-life data, and the clinical significance of the reported changes in quality-of-life scores.

Of the three other published studies in patient populations with baseline Hb levels <10 g/dL, only Henry, Brooks, Case, et al. (1995) reported a significant between-arm difference favoring epoetin. This study reported significantly improved results on overall quality-of-life visual analog scale but not on energy level or daily activities scales.

Studies enrolling patient populations with average baseline Hb levels >12 are small and unblinded and do not present consistent evidence of improved quality-of-life scores. There were no studies reporting quality-of-life outcomes that enrolled patient groups with mean or median Hb levels >10 and <12 g/dL at baseline.

Key Question 1 Conclusions

What are the outcomes of managing anemia with epoetin compared with transfusion alone?

We found adequate and consistent evidence that epoetin increases Hb levels and the percentage of patients achieving a hematologic response when compared with controls managed by transfusion alone. This was true for pediatric patients as well as adults. For all trials, the range of differences between epoetin and control arms for percent of patients responding was 18 percent to 80 percent. The range of differences for mean Hb change was 0.1 to 3.7 g/dL.

Based on data from the 12 randomized studies that administered epoetin subcutaneously, the odds of transfusion for epoetin-treated patients is reduced by a factor of 0.380 relative to the odds of transfusion for control patients. The calculated NNT to prevent one patient from being transfused for all studies is 4.4 (95 percent CI, 3.6 to 6.1), which suggests four to five patients must be treated to prevent one patient from being transfused. The estimated magnitude of effect is smaller when the analysis is limited to only higher quality studies. The odds of transfusion for epoetin-treated patients is reduced by a factor of 0.453 for higher quality studies compared with reduction by a factor of 0.137 for lower quality studies. For higher quality studies, the calculated NNT is 5.2 (95 percent CI, 3.8 to 8.4), or five to six patients treated for each who avoids transfusion. For lower quality studies, the calculated NNT is 2.6 (95 percent CI, 2.1 to 3.8), or two to three patients treated for each who avoids transfusion.

The strongest evidence to date for an effect of epoetin on quality-of-life outcomes is from an unpublished randomized double-blinded trial by Littlewood, Bajetta, Cella, et al. (1999) in a patient population with mean baseline Hb level <10 g/dL. There were statistically significant differences in score changes on visual analog scales (n=335 evaluable) for three questions and on the FACT-An (n=290 evaluable) that favored the epoetin-treated arm. Results from the SF-36 were not significant. As of this writing, no information is available to assess the study protocol for methods used to minimize bias in the collection of quality-of-life data, the impact of missing quality-of-life data, or the clinical significance of the reported changes in quality-of-life scores. Eight other published studies, which included a total of 516 patients evaluable for quality-of-life outcomes, do not provide consistent evidence that epoetin improves quality-of-life outcomes.

What are the relative effects of epoetin treatment when different Hb thresholds are used to initiate treatment?

No trial directly compared the outcomes of initiating epoetin treatment at alternative Hb thresholds. Only inferences based on indirect comparison are possible as to whether initiating epoetin at one or another Hb threshold results in superior outcomes. The most robust evidence that epoetin improves transfusion outcomes for patients undergoing therapy for malignancy compared with transfusion alone comes from trials in patient groups with baseline Hb <10 g/dL. Four trials in adults (n=682), including three of higher quality (n=493), reported significantly fewer patients transfused in the epoetin arms than in control arms.

Transfusion outcomes do not appear to be superior in trials where epoetin treatment is initiated in groups of patients with mean Hb >10 g/dL compared with the outcomes in trials where mean Hb is <10 g/dL. Among trials on adults with mean baseline Hb <10 g/dL, the range of differences in the percentage of patients transfused between epoetin and control arms was 9 percent to 45 percent. For mean baseline Hb >10 but <12 g/dL, the range was 7 percent to 47 percent; and for baseline Hb >12 g/dL, 7 percent to 39 percent. For RBC units per patient per 4 weeks, the differences for the three groups of studies were 0 to 0.7, 0.1 and 1.3, and 0.3 and 0.6. Because of insufficient data, we were unable to determine by meta-analysis whether or not initiating epoetin therapy at higher baseline Hb levels reduces the odds of transfusion.

Although it is possible that adequately powered comparative trials might demonstrate the superiority of epoetin intervention at the higher Hb levels, our examination of this evidence base suggests why that may not prove to be true. First, patients whose entry level Hb is well below the mean may account for a substantial proportion of transfusions in epoetin-treated patients in trials where baseline Hb is <10 g/dL. Thus the greatest yield for reducing the number of patients transfused in this population might come from initiating epoetin before the Hb level falls substantially below 10, rather than by initiating epoetin treatment at a level substantially above 10. Second, in all trials, patients who are unresponsive to epoetin may account for a substantial proportion of transfusions irrespective of the Hb level at which epoetin treatment is initiated. Initiating epoetin treatment at a higher Hb level is not expected to reduce transfusions in this subgroup of patients.

In conclusion, the available evidence is not adequate to determine whether outcomes of epoetin treatment are superior when treatment is initiated in groups of patients who have mean Hb> 10 g/dL compared with treatment in patient groups where mean Hb is <10 g/dL. Randomized controlled trials, double blinded and adequately powered, are needed to answer this question. Inferences from indirect comparison of the results of the available trials do not provide compelling evidence to resolve this issue.

Characteristics of Epoetin Administration on Outcomes

The characteristics of epoetin administration of interest to this analysis are dose, dosing regimen, and duration of treatment. The evidence base for this systematic review did not have adequate comparative studies to assess whether any dosage, dosing regimen, or treatment duration used was superior to another. Thus, the main focus of this section is to determine whether these characteristics of epoetin administration might confound the results of our meta-analysis on the effect of epoetin on the odds of transfusion. A second concern is whether these characteristics of epoetin administration might confound our interpretation of the evidence on the relative effects of epoetin treatment when treatment is initiated at different Hb thresholds.

The meta-analysis conducted on the effect of epoetin on transfusion outcomes also examined whether the characteristics of epoetin administration (dosing regimen, treatment duration, and dose range) have an effect on the estimate of the point estimate for the odds ratio for transfusion. Only epoetin dose appeared to have an independent effect on transfusion outcomes. However, this relationship potentially is confounded by study quality. Moreover, the two trials that directly compared lower and higher doses of epoetin do not provide strong support for the hypothesis that starting doses in the higher range are more effective in preventing transfusions.

Meta-Analysis Findings

The meta-analysis compares the ratio of the odds of transfusion for the epoetin-treated and control arms of each study and is therefore independent of followup duration. However, the components of epoetin treatment (dose, route of delivery, dosing regimen, duration of treatment) are not constant across studies and could confound the summary estimate of effect. Dose and route of delivery are related, since the intravenous and subcutaneous routes are likely to deliver different absolute doses, even when the administered doses are the same. Because most of the included RCTs used the subcutaneous route of delivery (12 of 14) and only 2 intravenous delivery RCTs were available to determine the effect, further analysis was restricted to studies that delivered epoetin subcutaneously (n=1,390).

Table 24. Effect of duration of epoetin treatment, (more...)

Table 24. Effect of duration of epoetin treatment, dosing regimen, and dose category on the risk of transfusion

Factor	Odds Ratio 1	95% Confidence Interval
Analysis by duration of treatment
Duration of <10 weeks	0.291	0.144, 0.587
Duration of 12-16 weeks	0.488	0.309, 0.769
Duration of >20 weeks	0.365	0.229, 0.580
Analysis by epoetin dosing regimen
Fixed dose	0.423	0.227, 0.789
Decreasing dose	0.343	0.222, 0.529
Increasing dose	0.461	0.292, 0.728
Analysis by epoetin weekly dose category
Dose of 300-450 U/kg	0.413	0.304, 0.561
Dose of 700-1000 U/kg	0.178	0.069, 0.459

1

Odds of transfusion for epoetin-treated patients relative to the odds of transfusion for control patients.

We examined the effects of the single categorical variables dosing regimen, treatment duration, and dose range on the estimate of the summary odds ratio for transfusion. Results are summarized in Table 24.

Duration of treatment was grouped into three categories: <10 weeks, 12 to 16 weeks, and >20 weeks. The estimated odds ratios for the likelihood of transfusion at each treatment duration are somewhat different. However, 95 percent confidence intervals for the odds ratios are widely overlapping, and therefore the differences are not significant. Thus, it seems unlikely that duration of treatment has a significant effect on number of patients transfused.

Dosing regimen was grouped into three categories: fixed dose, decreasing dose, and increasing dose. The odds ratios for the likelihood of transfusion for different dosing regimens are similar, and 95 percent confidence intervals are again overlapping, indicating no effect of this parameter on the summary estimate.

Epoetin weekly dose was grouped into two categories: 300 to 450 U/kg and 700 to 1,000 U/kg. The odds ratios for the likelihood of transfusion are significantly different, with confidence intervals that do not overlap. Thus, dose is the only component of epoetin intervention that requires further consideration as a potential confounder. Although it appears that increasing the dose of epoetin decreases the likelihood of transfusion, further analysis casts doubt on this interpretation.

Analysis of the effect of study quality within a single dose range of 300 to 450 U/kg per week indicated a significant effect of study quality (see Figure 2). However, it is not possible to additionally evaluate the effect of dose within higher quality studies, since only one higher quality study administered a dose within the 700 to 1,000 U/kg weekly dose category, and this study used an intravenous route of delivery (Wurnig, Windhager, Schwameis, et al., 1996). Therefore, the available evidence cannot distinguish between the effect of study quality and epoetin dose, and it is possible that higher epoetin doses have no greater effect than do lower epoetin doses.

Direct Comparisons of Dose

Table 25. Direct comparison of epoetin doses in three-arm (more...)

Table 25. Direct comparison of epoetin doses in three-arm studies, subcutaneous administration

Citation	Dose (U/kg per week)	Epo Regimen Class	Epo Tx Duration	N Enrolled	N Evaluable	% Response	p Value	Hb Change ±S.D.	p Value	% Transfused	p Value	RBC Units per Patient	p Value	RBC Units per Patient per 4 Weeks
Thatcher, De Campos, Bell, et al., 1999	0	2	3	44	44	34.1		−3.4		59.1		6.1		0.9
	450 sc	2	3	42	42	52.4	<0.05	−3.2		45.2	<0.05	3.8	<0.01	0.6
	900 sc	2	3	44	44	61.4	0.005	−3.3		20.5	<0.001	2.1	<0.001	0.3
ten Bokkel Huinink, de Swart, van Toorn, et al., 1998	0	2	3	34	33					39.4		1.2		0.2
	450 sc	2	3	46	45					4.4	1	0.3		0.1
	900 sc	2	3	42	42					14.3		0.4		0.1

1

Calculated odds ratio for transfusion suggests a significant difference, as upper limit of 95 percent confidence interval is <1.0 (see Meta-Analysis).

As shown in Table 25, two three-arm studies (n=250) directly compared patients administered subcutaneous epoetin at either 450 U/kg per week (n=87) or 900 U/kg per week (n=86) with controls managed by transfusion (n=77) (ten Bokkel Huinink, de Swart, van Toorn, et al., 1998; Thatcher, De Campos, Bell, et al., 1999). In each study, only the epoetin dosage differed between arms. Both used a decreasing dose regimen and treatment duration of >20 weeks. Mean baseline Hb was >10 but <12 g/dL in the ten Bokkel study, whereas mean baseline Hb in the Thatcher study was >12 g/dL. The results of these trials do not demonstrate better outcomes from the higher dose.

Thatcher and colleagues (1999) reported a significant difference between the two epoetin doses with respect to the percentage of patients transfused and the number of units transfused per patient (Thatcher, De Campos, Bell, et al., 1999). There appeared to be no significant difference between the epoetin dosages with respect to percentage of patients responding or the magnitude of change in Hb level. However, in this unblinded study, the participating centers varied in the mean Hb at which they initiated transfusion. Of particular concern, the range was markedly lower for the higher dose group than the lower dose group (8.7 to 10.8 g/dL versus 8.4 to 12.9 g/dL). Mean Hb for the 7 days prior to transfusion was also lower for the higher dose than for either the lower dose or control arms (8.0 g/dL versus 8.6 and 8.5 g/dL, respectively). The authors note that the higher dose group may have been somewhat undertransfused, possibly exaggerating the effect of the 900 compared with the 450 units/kg per week dose.

In contrast, the trial by ten Bokkel Huinink, de Swart, van Toorn, et al (1998) found no significant difference in percentage of patients transfused and number of RBC units transfused per patient. This unblinded study, which had a transfusion trigger of <9.7 g/dL, reported that the average Hb level at which transfusion actually was initiated was lower and consistent for all three arms (6.8 g/dL). The percentage of patients transfused was slightly higher in the 900 than in the 450 units/kg per week arm (14.3 percent vs. 4.4 percent, respectively), demonstrating no advantage for the higher dose arm. In this decreasing dose regimen, by the third cycle of chemotherapy the median dose of epoetin was reduced by approximately 50 percent in the higher dose arm (900 U/kg to 447 U/kg), compared with about 4 percent in the lower dose arm (450 U/kg to 432 U/kg). However, no hematologic outcomes were reported. The study authors concluded that the higher starting dose was not more effective than the lower starting dose.

Dosage and Dosing Regimen in Two-Arm Studies

Table 26. Hematologic outcomes by weekly dose, two-arm (more...)

Table 26. Hematologic outcomes by weekly dose, two-arm studies of subcutaneous epoetin¹

Citation	Initial EPO Dose (U/kg per week)	EPO Regimen Class 2	EPO Tx Duration 3	N Enrolled	N Evaluable	% Response	p Value	Difference in % Response (Epo-control)	Hb Change ±SD	p Value	Difference in Hb Change (Epo-control)
Weekly doses from 300 to 450 U/kg (lower doses)
Varan, Buyukpamukcu, Kutluk, et al., 1999	0	1	1	17	17				−0.07
	450	1	1	17	17				1.71		1.78
Gamucci, Thorel, Frasca, et al., 1993	0	1	2	17	17				−1.5±1.67
	450	1	2	21	21				0.9±1.32	<0.005	2.40
Oberhoff, Neri, Amadori, et al., 1998	0	1	2	110	88	6.8
	~450	1	2	117	101	34.7	0.0001	27.9
Markman, Reichman, Hakes, et al., 1993	0	1	3	46	40	40.0
	350	1	3	17	16	87.5	<0.005	47.5
Cascinu, Fedeli, Del Ferro, et al., 1994	0	2	1	50	49	2.0			−0.6
	300	2	1	50	50	82.0		80.0	1.9		2.50
Del Mastro, Venturini, Lionetto, et al., 1997	0	2	2	31	31				−3.1±1
	450	2	2	31	31				−0.8±1.4	<0.005	2.30
Case, Bukowski, Carey, et al., 1993	0	2	2	76	74	13.5			0.33
	450	2	2	81	79	58.2		44.7	2.3	0.0001	1.97
Henry, Brooks, Case, et al., 1995	0	2	2	65	61	6.6			0.4*±1.7
	450	2	2	67	64	48.4	<0.0001	41.8	2.0*±2.3	<0.0001	1.60
Kurz, Marth, Windbichler, et al., 1997	0	3	2	12	12	0.0			0.22
	450	3	2	23	23	56.5	0.001	56.5	3.3		3.08
Quirt, Couture, Pichette, et al., 1996	0	3	2	28	27				0.6
	450	3	2	28	27				1.6		1.00
Silvestris, Romito, Fanelli, et al., 1995	0	3	3	24	22	0.0
	450	3	3	30	27	77.8		77.8
Littlewood, Bajetta, Cella, et al., 1999	0	3	3	124	115	19.1			0.9
	450	3	3	251	244	70.5	0.001	51.4	2.5		1.60
Weekly doses from 750 to 1,000 U/kg (higher doses)
Leon, Jimenez, Barona, et al., 1998	0	1	2	25	25				0.1
	750	1	2	25	25	72.0			2.6	<0.001	2.50
Sweeney, Nicolae, Ignacio, et al., 1998	0	2	1	24	24	0.0			0.29
	1000	2	1	24	22	45.5		45.5	1.55	0.0012	1.26
Dusenbery, McGuire, Holt, et al., 1994	0	2	1	61	61				−0.8
	1000	2	1	15	15				2.9	0.001	3.70
Lavey and Dempsey, 1993	0	2	1	20	20	5.0			0.0±0.7
	900	2	1	20	20	80.0	<0.001	75.0	3.2±1.78	<0.001	3.20
Welch, James, Wilkinson, et al., 1995	0 900	2 2	3 3	15 15	15 15				−2.1 −1.3		0.80

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

2

Treatment regimen: 1 = fixed dose; 2 = decreasing dose; 3 = increasing dose.

3

Treatment duration: 1 =≤10 weeks; 2 = 12 to 16 weeks; 3 =≥20 weeks.

Table 27. Transfusion outcomes by weekly dose, two-arm (more...)

Table 27. Transfusion outcomes by weekly dose, two-arm studies of subcutaneous epoetin¹

Citation	Dose (U/kg per week)	EPO Regimen Class 2	EPO Tx Duration 3	N Enrolled	N Evaluable	% Transfused	p Value	Difference in % Transfused (control-EPO)	RBC Units per Patient	p Value	RBC Units per Patient per 4 Weeks	Difference in RBC units per Patient per 4 Weeks (control-EPO)
Weekly doses from 300 to 450 U/kg (lower doses)
Varan, Buyukpamukcu, Kutluk, et al., 1999	0	1	1	17	17	47.1
	450	1	1	17	17	5.9	0.008	41.2
Gamucci, Thorel, Frasca, et al., 1993	0	1	2	17	17
	450	1	2	21	21
Oberhoff, Neri, Amadori, et al., 1998	0	1	2	110	88	40.9			0.6		0.6
	~450	1	2	117	101	25.7	4	15.2	0.5	0.044	0.5	0.1
Markman, Reichman, Hakes, et al., 1993	0	1	3	46	40	22.5
	350	1	3	17	16	6.3	NS	16.2
Cascinu, Fedeli, Del Ferro, et al., 1994	0	2	1	50	49	57.1			1.8		0.8
	300	2	1	50	50	20	0.01	37.1	0.3	0.01	0.1	0.7
Del Mastro, Venturini, Lionetto, et al., 1997	0	2	2	31	31	6.5
	450	2	2	31	31	0.0	NS 5	6.5
Case, Bukowski, Carey, et al., 1993	0	2	2	76	74	33.86			1.6± 0.3		0.8
	450	2	2	81	79	25.36	NS5	8.56	0.9± 0.3	NS	0.5	0.3
Henry, Brooks, Case, et al., 1995	0	2	2	65	61	68.9			4.0+ 0.8		2.0
	450	2	2	67	64	53.1	NS	15.8	4.0+ 0.9	NS	2.0	0.0
Kurz, Marth, Windbichler, et al., 1997	0	3	2	12	12	66.7			3.6		1.2
	450	3	2	23	23	21.7	0.009	45.0	1.4		0.5	0.7
Quirt, Couture, Pichette, et al., 1996	0	3	2	28	27	29.6			0.7
	450	3	2	28	27	14.8	NS 5	14.8	0.2
Silvestris, Romito, Fanelli, et al., 1995	0	3	3	24	22
	450	3	3	30	27
Littlewood, Bajetta, Cella, et al., 1999	0	3	3	124	115	35.76
	450	3	3	251	244	236	0.0168	12.76
Weekly doses from 750 to 1000 U/kg (higher doses)
Leon, Jimenez, Barona, et al., 1998	0	1	2	25	25	96			3.6		1.2
	750	1	2	25	25	16	<0.001	80.0	0.3	<0.001	0.1	1.1
Sweeney, Nicolae, Ignacio, et al., 1998	0	2	1	24	24
	1,000	2	1	24	22
Dusenbery, McGuire, Holt, et al., 1994	0	2	1	61	61	6.6
	1,000	2	1	15	15	0.0		6.60
Lavey and Dempsey, 1993	0	2	1	20	20
	900	2	1	20	20
Welch, James, Wilkinson, et al., 1995	0	2	3	15	15	53.3			5.4
	900	2	3	15	15	26.7	NS	26.6	4.0	NS

1

"Higher quality" trials in bold font; nonrandomized studies in italics.

2

Treatment regimen: 1 = fixed dose; 2 = decreasing dose; 3 = increasing dose.

3

Treatment duration: 1 =≤10 weeks; 2 = 12 to 16 weeks; 3 =≥20 weeks.

4

Calculated odds ratio for transfusion suggests a significant difference, as upper limit of 95 percent confidence interval is <1.0 (see Meta-Analysis).

5

Calculated odds ratio for transfusion suggests no significant difference, as upper limit of 95 percent confidence interval is <1.0 (see Meta-Analysis).

6

Measured from day 28 to end of study.

Tables 26 and 27 summarize data from patients in 17 two-arm trials of subcutaneously administered epoetin compared with data from a control group managed without epoetin; Table 26 reports on hematologic outcomes, and Table 27 reports on transfusion outcomes. Twelve trials (n=1,253) used epoetin doses from 300 to 450 U/kg per week and five trials (n=242) used doses from 750 to 1,000 units/kg per week. Of 1,134 patients enrolled in studies with baseline Hb <10 g/dL, the patient group for which the most robust evidence of transfusion prevention exists, all but 50 were treated with epoetin doses in the 300 to 450 U/kg per week range. This body of evidence clearly demonstrates that doses in the 300 to 450 U/kg per week range are adequate to increase Hb and reduce percentage of patients transfused.

All but two of the trials in the lower dose range used the 450 U/kg per week doses. The results of these two trials show that epoetin can be effective at the lower doses in this range. Cascinu, Fedeli, Del Ferro, et al. (1994; n=100) used a 300 U/kg per week dose and reported a significant reduction in percentage of patients transfused and RBC units per patient. Markman, Reichman, Hakes, et al. (1993; n=63) used a 350 U/kg per week dose and reported a significant difference in percentage of patients demonstrating a hematologic response.

Within the 300 to 450 U/kg per week range, the effect of the dosing regimen is of interest. Using an increasing dose regimen, in which the dose increases for patients who fail to respond to the initial dose of epoetin, could increase the percentage of patients who achieve a hematologic response. Increasing the number of patients who achieve a hematologic response could in turn improve transfusion outcomes, resulting in fewer patients transfused and fewer RBC units per patient. However, no studies included in this review directly compared hematologic response rates according to dosing regimen. The results show no obvious advantages for an increasing dose regimen within the group of studies using an initial dose of 300 to 450 U/kg per week. The range of differences between the epoetin and control arms in the percentage of patients achieving hematologic response is 42 percent to 80 percent for the decreasing dose regimen and 51 percent to 78 percent for the increasing dose regimen. However, such indirect comparison is of little value unless an effect of very large magnitude exists; based on the available data, no conclusions can be drawn.