There are two erythropoietin products currently approved in the U.S. The first approved agent was epoetin alfa, which is manufactured, distributed and marketed by Amgen, Inc. under the proprietary name EPOGEN. The same epoetin alfa product, manufactured by Amgen, Inc., is also marketed and distributed by Ortho Biotech, L.P., a subsidiary of Johnson & Johnson, under the proprietary name PROCRIT. EPOGEN/PROCRIT was licensed in June 1989, with the following indication: “treatment of anemia associated with chronic renal failure, including patients on dialysis (end stage renal disease) and patients not on dialysis.” Under a contractual agreement, Ortho Biotech LP has rights to development and marketing of Procrit for any indication other than for the treatment of anemia associated with chronic renal failure. Epogen and Procrit have identical labeling information for all approved indications based on development programs conducted by Amgen or Ortho Biotech. Labeling was expanded in April 1993 to include a supplemental indication for the treatment of anemia associated with cancer chemotherapy.

The second product was darbepoetin alfa, which is manufactured and distributed by Amgen, Inc., under the proprietary name Aranesp. Aranesp was licensed in September 2001 with the following indication: “for the treatment of anemia associated with chronic renal failure, including patients on dialysis and patients not on dialysis”. Labeling was expanded in July 2002 to include a supplemental indication for the treatment of anemia associated with cancer chemotherapy.

In this briefing document, FDA provides an overview of the pharmacologic effects of erythropoietin, a summary of the data on erythropoietin receptor distribution in normal and malignant tissues, an overview of relevant non-clinical (animal and laboratory) data suggesting a role for erythropoietin in tumor stimulation, and a summary the design and results of studies that supported the approvals of Epogen/Procrit and Aranesp for the indications of treatment of the anemia of renal failure and of anemia associated with cancer chemotherapy, as well as other selected relevant studies provided to the FDA, will be presented. The extent of the information on associations between Epogen/Procrit or Aranesp and the risks of thrombotic events, tumor progression and survival are noted in these summaries. The results of the BEST (INT-76) and Henke studies will also be summarized.

Evidence of an increased risk of thrombotic events associated with use of exogenous erythropoietin was noted in the trials that supported the original approval of Epogen/Procrit. Excessive or poorly-controlled pharmacodynamic effects of erythropoietins have the potential to precipitate cardiovascular adverse events (AEs), some severe or catastrophic. These events are thought to have, as their basis, alterations in rheologic and/or hemodynamic factors related to increasing erythropoiesis, and include accelerated hypertension, congestive heart failure, pulmonary edema, ischemic events (stroke, transient ischemic attack [TIA], acute myocardial infarction [MI], thrombosis of vascular access [TVA], peripheral ischemia/gangrene), and seizures. However, evidence for an increased risk of fatal cardiovascular events and impaired survival associated with the administration of exogenous erythropoietin products according to a specific treatment strategy (i.e., targeting of higher hemoglobin levels than required for avoidance of transfusion) came several years after the original approval of Epogen/Procrit. As a result of the “Normal Hematocrit Study”, the labeling of Epogen/Procrit was modified to include warnings regarding this increased risk and association with this treatment strategy. The labeling for Aranesp carries similar warning statements.

Impaired survival and evidence of possible tumor stimulation associated with erythropoietin products has been observed in the BEST Study (EPO-INT-76) and the study by Henke, et al, published in the Lancet in October 2003. These two studies are large, multicenter, randomized, placebo-controlled studies whose purposes were to assess the impact of supplemental erythropoietin use on survival and tumor outcomes. It is notable that these trials are larger than any conducted

The data from these two studies may not be applicable to the U.S. licensed products. Both the BEST and the Henke studies used erythropoietin products that are not available in the U.S. and both studies used treatment strategies (high target hemoglobin) that are not recommended in labeling for either Epogen/Procrit or Aranesp. However, the biochemical differences between various erythropoietin products are not associated with marked differences in the pharmacodynamic properties of the different products when used at recommended doses, thus effects observed with these non-US-licensed products may be also be associated with the U.S. licensed product. Furthermore, the presence of erythropoietin receptors on tumor and tumor vasculature and the stimulatory effect of erythropoietins on certain tumor lines suggest a plausible reason for concern. In addition, while the treatment strategies used in the BEST and Henke trials are not consistent with current labeling for the U.S. licensed products, the studies used to support labeling for treatment of anemia associated with cancer for Epogen/Procrit or Aranesp are smaller and were not of adequate design to rule out the potential for tumor stimulation or a survival decrement of a specific magnitude.

Data from non-clinical and clinical studies provide a sound basis for FDA’s request for additional clinical studies to assess the safety and optimal manner for administration of erythropoietin to patients with cancer. Erythropoietin products are used an alternative form of supportive care. In clinical studies in anemic cancer patients, treatment with an erythropoietin product can reduce the proportion of patients who receive red blood cell transfusions by approximately 35-50%, beginning about one month after initiation of treatment. It should be noted that claims of improvement in quality of life are not been supported by data submitted to the FDA and there is insufficient evidence to support such a claim from literature reports due to the lack of adequate and well-controlled trials. In discussion with both firms, FDA has requested and both firms have agreed to conduct adequately designed trials that will assess whether, when administered in accordance with current labeling, there is evidence of tumor stimulation or impairment in survival (due to tumor stimulation, thrombotic events, or any cause) with Epogen/Procrit or Aranesp. Amgen, Inc. and Ortho Biotech LP will present their proposed approach for addressing these concerns. The approach consists primarily of randomized, placebo-controlled trials, potentially supplemented by additional preclinical studies. The FDA requests that the advisory committee comment on the adequacy of the proposed approach, in particular with regard to following:

· Study population, e.g., those with primary tumors where erythropoietin receptors (EPOr) are commonly present on tumor, EPOr commonly present and are shown to be functional, and EPOr not commonly present or when present are ordinarily not functional (to assess for effect mediated through angiogenesis rather than direct tumor stimulatory effect)

· Magnitude of decrement in time-to-progression and/or survival that studies should be powered to detect.

· Replication of results/number of different primary tumors that should be evaluated.

I. Erythropoietin Biology and Mechanism of Action

Erythropoietin is a glycoprotein whose main function is to stimulate the proliferation and differentiation of erythroid precursors in the bone marrow.[1]^,[2] Erythropoietin is a 165 amino acid monomeric polypeptide[3] containing two intramolecular disulfide bonds. The primary sequence encodes one consensus O-linked glycosylation site, and three N-linked consensus glycosylation sites. The primary sequence of Darbepoetin alfa, an erythropoietin analog, contains two additional N-linked glycosylation sites resulting from amino acid substitutions in the peptide backbone. These additional oligosaccharide side chains increase the molecular weight of the protein from approximately 30 kDa to 37 kDa. Darbepoetin has a three-fold longer terminal half-life than erythropoietin alfa[4], and a five-fold lower affinity for erythropoietin receptors. In addition to proliferation and differentiation of erythroid precursors, erythropoietin has also been shown to be an erythrocyte survival factor, by modulating pro- and anti-apoptotic mechanisms, and a pro-angiogenic factor[5]. Studies show erythropoietin stimulates the proliferation and migration of endothelial cells in vitro, and stimulates the expression of other angiogenic growth factors, namely Vascular Endothelial Growth Factor (VEGF) and Placental Growth Factor (PlGF)[6].

Erythropoietin is produced mainly in the kidneys, though several other tissues produce lesser amounts of the growth factor^{1, 2}. Erythropoietin transcription and the protein’s release into the bloodstream are both induced by hypoxic conditions^1,
2. The erythropoietin gene contains a hypoxia-responsive element. Hypoxia-inducible factor-1 (HIF-1), a transcription factor, is activated when cells are exposed to hypoxia. HIF-1 then binds the hypoxia-responsive element and up-regulates erythropoietin gene expression[7].

There are two types of erythropoietin receptors, high affinity receptors, expressed predominantly on hematopoietic cells, with kDs of approximately 95 pM, and low affinity receptors expressed on non-hematopoietic cells, with binding affinities of approximately 16 nM. The binding affinity of an erythropoietin ligand for the erythropoietin receptor is not only influenced by the type of receptor alone, but also by tissue-dependent receptor numbers and accessory proteins[8]. Normal non-hematopoietic cells expressing the erythropoietin receptor include those of the female reproductive tract (placental trophoblasts, cervical squamous epithelium, uterine glandular epithelium and endometrium, ovarian follicles)[9]^{,[10],[11],[12]}, breast (mammary epithelium)[13], prostate (epithelium)[14], vasculature (endothelium)[15], nervous system (neurons, astrocytes, oligodendrocytes, microglia)[16]^,[17],[18], pancreas (islet cells)[19], and kidney (cortex, medulla, papilla)[20]. While the role of erythropoietin in nonhematopoietic tissues is not completely understood, the erythropoietin receptors expressed on these tissues are functional.

Upon ligand binding, the erythropoietin receptor dimerizes and triggers a variety of responses via several signaling pathways[21]. These pathways include proteins with a variety of functions, including transcription activators, protein kinases and phosphatases, nucleotide exchange factors, phospholipid modifying enzymes, and adaptor proteins. Activation of these pathways results in DNA synthesis, cell differentiation, and cell proliferation. The erythropoietin receptor activates signaling molecules common to several other growth factor receptors²¹. These signaling molecules include Jak2-STAT5[22], Ras-MAP kinase[23], and PI3-kinase[24]. Erythropoietin receptor signaling results in the down-regulation of several pro-apoptotic proteins (Fas-ligand[25], TRAIL²⁵, and BAD[26]), and the up-regulation and activation of anti-apoptotic proteins (Bcl-X_L and Bcl-2)^{6, 21}. No functional differences between high and low affinity receptors, in terms of downstream activities, have been elucidated. Ligand binding to either class of receptor results in cell proliferation and/or cell survival under hypoxic conditions.[27]

In addition, the erythropoietin receptor associates with other receptors, namely the GM-CSF/IL-3/IL-5 ß common chain[28] and c-Kit[29]^,
[30], the receptor for Stem Cell Factor. It is possible that erythropoietin influences a variety of cell types through this mechanism.

II. Preclinical Evidence for a Role of Erythropoietins in Tumor Progression

In Vitro Findings

A substantial body of preclinical studies demonstrates that erythropoietin receptors are present on a variety of malignant cell lines^{6,
10, 20, [31],[32],[33]} as well as on primary tumor cells. Primary tumor cells have been shown to respond to erythropoietin administration by proliferating and forming vasculature^{20, [34],[35],[36]}. This information, coupled with the knowledge that erythropoietin elicits anti-apoptotic effects in stem cells, and and together with the results of recent clinical studies mandates the investigation of the potential role of erythropoietin in tumor progression.

Erythropoietin receptors are expressed on some primary tumor cells, but not on all. For example, neuroblastomas⁶, Ewing’s sarcomas⁶, hepatoblastomas⁶, Wilm’s tumors⁶, brain tumors⁶, cervical carcinomas¹⁰, mammary adenocarcinomas³³, renal carcinomas²⁰, acute monoblastic leukemia³¹, gastric carcinomas³⁴, and endometrial³⁵, cervical¹⁰, and ovarian³⁵ adenocarcinomas all have been demonstrated to express erythropoietin receptors. However, not all tumors of the same type from the same tissue of origin express erythropoietin receptors at similar levels. The presence and levels of erythropoietin receptors may vary from patient to patient. Arcasoy, et al.³³ examined 26 mammary tumor biopsies. Although ninety percent of these tissue samples expressed detectable levels of erythropoietin receptors, the remainder did not.

Interestingly and importantly, erythropoietin receptor expression on primary tumors often directly correlates with disease staging^{10, [37]}, which may reflect the level of hypoxic stress in more advanced and aggressive tumor masses. Thus, under these conditions, increased erythropoietin binding and signaling activity may facilitate tumor survival by initiating the increased expression of anti-apoptotic genes and the down-regulation of pro-apoptotic genes, in addition to facilitating angiogenic activities^{6, 13, [38]}.

The expression of erythropoietin receptors on tumor vascular endothelial cells further suggests that erythropoietin may assist in tumor progression by promoting endothelial cell proliferation and vessel formation within the tumor^{34, 35}. In addition, erythropoietin may regulate tumor vasculature development indirectly; upon erythropoietin administration, the levels of other endothelial growth factors, namely VEGF and PlGF are upregulated⁶. Stimulation of erythropoietin receptors in vascular endothelial cells leads to cell proliferation and increased chemotaxis, providing evidence of a role for erythropoietin/erythropoietin receptors in angiogenesis [39]^,[40],[41].

In addition to the expression of erythropoietin receptors on malignant cells, some tumors also secrete erythropoietin^{6,13,20, 33,35,38}. This may allow a tumor to regulate its own growth by an autocrine pathway. It has been postulated that the local intratumor levels of erythropoietin produced through this autocrine loop may exceed the level of erythropoietin the tumor would encounter from therapeutic doses of erythropoietin. Batra, et al.⁶, postulate that, unlike hematopoietic tissues that respond to relatively low levels of erythropoietin, tumors might require high concentrations of erythropoietin to achieve a response. This high concentration of erythropoietin might be achieved only by local intratumoral erythropoietin production. The effect of therapeutic erythropoietin administration on tumor progression must be considered within this context as well.

Among data addressing the capacity of erythropoietin to stimulate tumor growth, the most convincing are studies demonstrating that tumor progression and tumor angiogenesis can be inhibited by the addition of agents that block erythropoietin binding or signaling: e.g. anti-erythropoietin antibodies, Jak2 inhibitors, and soluble erythropoietin receptors^{32, 33}. Yasuda et al. (2002)³⁵ reported that erythropoietin and

erythropoietin receptor are expressed in malignant tumors of the female

reproductive organs, where tumor cells and capillary endothelium showed erythropoietin receptor immunoreactivity, and that the injection of a monoclonal antibody against erythropoietin or the soluble form of erythropoietin receptor into tumors reduced capillaries and caused tumor destruction in a dose-dependent manner. Histopathologic changes including, fragmented cellular DNA, and the absence of phosphorylated Jak2 and STAT5 cells in tumors, relative to controls suggested that the tumor and capillary cell decrease resulted from apoptotic cell death. These studies support the conclusion that erythropoietin signaling contributes to tumor survival and to the promotion of tumor growth and angiogenesis.

It should be noted that the mere presence of erythropoietin receptors on tumor cells does not necessarily confer on such cells the capacity to respond to erythropoietin. These receptors must be functional. Not only must they be able to bind erythropoietin, but they must be able to activate the downstream intracellular signaling pathways through which erythropoietin elicits its biological activities. Some in vitro studies on tumor cell lines have demonstrated the lack of proliferation upon exogenous erythropoietin administration, despite the presence of erythropoietin receptors on the surface of cells. Westphal et al.³¹ evaluated over twenty tumor cell lines, including AML, breast, pancreatic, prostate, and kidney cell lines. This study demonstrated that the proliferation rates of these erythropoietin receptor-positive cell lines were not influenced by the addition of exogenous erythropoietin. Moreover, addition of erythropoietin did not increase the tyrosine kinase activity in these cells. The presence of erythropoietin receptors on these cells was not essential for the growth of these cells in culture. A lack of proliferative response on erythropoietin receptor-positive cells was also observed by Takeshita, et al.[42] when primary AML and melanoma cells were treated with exogenous erythropoietin. These studies, as opposed to the ones indicating a trophic effect of erythropoietin, beg the question as to whether the tumors that respond express the high affinity receptor versus the low, whether the levels differ, or whether the amount of endogenous erythropoietin produced by cells that fail to respond to exogenous erythropoietinin renders such cells resistant to further stimulation by erythropoietin.

In Vivo Findings

While the majority of the published data supporting that erythropoietin can promote tumor growth, survival and angiogenesis was obtained in in vitro systems, in vivo data are also available. First, the direct exposure of uterine and ovarian tumor slices transplanted into nude mice to erythropoietin antagonists resulted in reduction in size. Immunohistochemical staining revealed a decrease in erythropoietin -responsive malignant and capillary endothelial cells through apoptotic cell death[43]. Second, treatment of xenograft models of stomach choriocarcinoma and melanoma with erythropoietin antagonists inhibited angiogenesis and survival of tumor cells. In contrast, treatment of the xenograft models with an erythropoetin-mimetic peptide promoted angiogenesis and tumor survival³². Third, experiments conducted using a rat syngeneic mammary adenocarcinoma cell line implanted into the subcutaneous tissue of rats in a chamber revealed that erythropoietin antagonists delayed tumor growth³³.

Conclusions

From the regulatory perspective there are sufficient preclinical data to support the hypothesis that erythropoietin can promote tumor growth, survival, and angiogenesis. Although there are, undoubtedly, additional preclinical studies that could be conducted to further elucidate the mechanisms underlying erythropoietin’s apparent effect on tumors, these studies would not be able to directly assess the clinical relevance of this effect. It is recommended that the ability of erythropoietin to promote tumor growth, survival, and angiogenesis be assessed in an appropriately designed clinical trials in patients that have been adequately informed of the potential risk.

III. Clinical Studies of Epogen/Procrit for the Treatment of the Anemia due to Chronic Renal Failure

The results of thirteen clinical studies that included a total of 1,010 patients were used support the approval of Epogen/Procrit for treatment of anemia associated with ESRD. Four other studies were performed in patients with renal failure whose disease was not severe enough to require dialysis. Six studies were conducted in normal male volunteers (157 subjects, 108 of whom received Epogen and 49 of whom received placebo).

Treatment of Anemia Due to Chronic Renal Failure in Patients Undergoing Dialysis

The primary efficacy data were derived from two large multicenter trials. The first study was a multicenter, open-label study in which 412 subjects with end stage renal failure on dialysis received Epogen three times a week. Patients initially received one of three dose levels: one group received 300 U/kg only; one group was dosed at 300 U/kg and subsequent dose reduced to 150 U/kg; and the majority received 150 U/kg only. When a patient achieved a hematocrit of 35%, or completed 12 weeks of therapy at the initial dose, they entered the dose adjustment and long-term maintenance phase of the study.[44] 309 patients were evaluable for efficacy, and 95.5% had an increase in hematocrit of 6 points or reached the target hematocrit of 35% within 12 weeks of the initiation of therapy. Approximately 70% satisfied these criteria within the first four weeks of therapy. Following six more weeks of therapy, 97% of the evaluable patients met these criteria. The percentage of patients who responded to therapy was not significantly different between the three dosage groups: for 300 U/kg, 300 U/kg and 150 U/kg, and 150 U/kg the values were 100%, 94.5% and 95.2%, respectively.[45]

When the enrolled patients achieved a hematocrit of 35%, they were entered into the dose adjustment and long-term maintenance phase, where the dose was individually adjusted to maintain the hematocrit within the target range of 32-38%. Sixty-four percent of the patients required doses of Epogen between 12.5 and 100 U/kg to maintain their hematocrit within the desired target range.

Transfusion requirements decreased within weeks after the initiation of therapy. The pre-study transfusion requirements of 0.52 units per patient per month were reduced to 0.1 units per patient per month after the first four weeks on Epoetin therapy, and to 0.04 units per patient per month or less through 14 months on study.

The second major efficacy study was the U.S. Pivotal Double-blind, Placebo-controlled (DBPC) Multicenter Study in ESRD Patients, which was conducted at three study sites. This study enrolled 100 anemic ESRD patients who were on maintenance hemodialysis. Patients were randomized to receive either placebo or 150 U/kg Epogen t.i.w. for 12 weeks. During the second 12-week study period, all patients received drug on an open-label basis. Once a patient’s hematocrit reached 35%, they were entered into the dose adjustment and long-term maintenance phase of this study.

Of the 62 patients evaluable for efficacy, 95% achieved a hematocrit of 35% or six points over baseline. Ninety-seven percent of the patients (97%) randomized to therapy achieved this efficacy criterion in the blinded phase of the study, and 93% of the patient achieved this criterion after crossover from placebo to Epogen treatment.

Two, double-blind, placebo controlled studies were conducted in Canada in ESRD patients. In a single-center trial, ESRD patients (6 patients per group) were treated IV for nine weeks with either placebo or 50 U/kg, 100 U/kg or 200 U/kg Epogen t.i.w. The rate of rise of hematocrit was dose dependent. In the larger multicenter DBPC trial, patients received placebo (n=40) or Epogen at an initial dose of 100 U/kg (n=78) for 26 weeks. The mean change from baseline for hemoglobin was 0.006 g/dl for placebo-treated patients, and 3.8 g/dl for Epogen -treated patients.

Treatment of Anemia Due to Chronic Renal Failure in Patients not undergoing Dialysis

Four clinical studies were conducted in CRF patients whose disease was not severe enough to require dialysis (non-dialysis CRF patients): two U.S. multicenter double-blind placebo controlled studies, a continuation long-term maintenance study, and a European open-label study. In the first U.S. DBPC study, non-dialysis CRF transplants received placebo (n=31) or Epogen at 50 U/kg (n=28), 100 U/kg (n=28), or 150 U/kg (n=30) intravenously t.i.w. ; patients were treated for 8 weeks, or until their anemia was corrected (hematocrits of 40% for males and 35% for females). Treatment with Epogen increased the hematocrit in a dose-dependent manner; changes of –0.01, 0.13, 0.20 and 0.26 hematocrit points per day were seen for the placebo, Epogen 50 U/kg, 100 U/kg, and 150 U/kg dosage groups, respectively. Upon completion of eight weeks of therapy or correction of anemia, whichever came first, patients in this study were enrolled in a 6-month maintenance study protocol. In the maintenance study, Epogen was administered either intravenously or subcutaneously t.i.w., and the dose was adjusted to maintain a constant elevated hematocrit. Ninety-four percent of all patients in the study corrected their hematocrit, and doses of Epogen 75-150 U/kg per week were shown to maintain hematocrits of 36-38% for up to six months.

In the second U.S. DBPC trial, non-dialysis CRF patients were administered either placebo (n=48) or 100 U/kg Epogen (n=45) t.i.w. subcutaneously for up to 12 weeks or until the hematocrit reached 38-40%, whichever occurred first. Fifty-eight percent of the Epogen-treated patients, versus 4% of the placebo-treated patients, corrected their anemia (hematocrit > 40% for males and > 35% for females) during the study period.

Safety Analyses

Analyses of safety included the data from all studies conducted in patients with chronic renal failure, including those undergoing dialysis and those not undergoing dialysis. Hypertension was the most frequently reported adverse event in both the placebo and Epogen-treated patients. In patients on dialysis, the incidence of reported hypertensive events for all Epogen-treated patients was approximately twice that for placebo-treated patients (0.69 versus 0.33 events per patient-year, respectively). In non-dialysis CRF patients, the rate of hypertension was higher, occurring at a rate of 1.70 and 3.28 events per patient-year in Epogen- and placebo-treated patients, respectively.[46] When patients in the U.S. Phase III multicenter ESRD trial were analyzed for incidence of hypertension as a function of the rate of rise in hematocrit, there was a trend towards more reports of hypertension in the first 90 days of therapy in patients who had increases in hematocrit that were greater than 0.3 points per day.

In placebo-controlled studies enrolling over 300 patients with chronic renal failure, the following adverse events occurred at a higher incidence Epogen-treatment patients as compared to placebo controls: hypertension (24% vs. 18%), headache (16% vs. 12%), arthralgias (11% vs. 6%), diarrhea (8% vs. 6%), vomiting (8 % vs. 5%), and clotted vascular access (6.8% vs. 2.3%). In US and non-US studies, the annual rates (events per patient-year) of clotted vascular access in patients receiving Epogen was 0.25 and 0.27 events/patient-year, respectively.[47]

IV. The Normal Hematocrit Study of Epogen/Procrit in Patients with Chronic Renal Failure and Underlying Cardiovascular Disease

Amgen, Inc. conducted a study in 1,233 patients with chronic renal failure on dialysis[48] and clinical evidence of congestive heart failure or ischemic heart disease. Patients were randomized to an Epogen dose titrated to achieve a target hematocrit of 42% (±3) (the “normal hematocrit group” or an epoetin dose titrated to achieve and maintain a target hematocrit of 30% (±3) (the “low hematocrit group”). The study was designed to test the hypothesis that correction of anemia in chronic renal failure patients on dialysis with clinical evidence of congestive heart failure or ischemic heart disease would have improved survival and better exercise tolerance if treated with Epogen to obtain a higher hematocrit than had been commonly targeted in clinical practice. The primary endpoint was the length of time to death or a first nonfatal myocardial infarction.

This study was halted at the third interim analysis on the recommendation of the Data Safety Monitoring Board. At 29 months, there were 183 deaths and 19 first nonfatal myocardial infarctions in the group with a normal hematocrit and 150 deaths and 14 nonfatal myocardial infarctions in the low hematocrit group. (Figure 2) Even though these differences did not reach the prespecified statistical stopping boundary, the study was halted “because differences in mortality between the groups were recognized as sufficient to make it very unlikely that a continuation of the study would reveal a benefit for the normal-hematocrit group and the results were nearing the statistical boundary of a higher mortality in the normal-hematocrit group.”[49]

Safety Results

The incidences of non-fatal MI were 3.1% and 2.3% in the normal and low hematocrit groups, respectively. The incidences of CVA (39% versus 29%) and all other thrombotic events (22% versus 18%) were also higher in the normal hematocrit group. There was a trend to decreasing mortality with increasing hematocrit values within both groups. (Figure 1)

This study demonstrated that the aggressive use of erythropoietin to correct the hematocrit to normal values is associated with higher risks in subjects with chronic renal failure and pre-existing cardiovascular disease. As a result, a Warning was added to the Epogen and Procrit Package Inserts.

These findings also led FDA to examine the risks associated with the rate of rise of hemoglobin in the studies that were submitted to support the licensure of Aranesp for treatment of anemia in chronic renal failure (see below).

Figure 1: Mean (±SE) Mortality Rate as a Function of the Average Hematocrit Value in the Normal-Hematocrit and Low-Hematocrit Groups.

Figure 2: Kaplan–Meier Estimates of the Probability of Death or a First Nonfatal Myocardial Infarction in the Normal-Hematocrit and Low-Hematocrit Groups.

V. Clinical Studies of Aranesp® (darbepoetin alfa) for the Treatment of the Anemia due to Chronic Renal Failure

The licensure of Aranesp in subjects with chronic renal failure was based on two active controlled open-label studies in EPO naïve patients, and two randomized double-blind non-inferiority studies in patients who had previously been on a stable dose of epoetin alfa. The two studies in EPO naïve patients were Study 211: An Open-Label Randomized Study of ARANESP and Recombinant Human Erythropoietin (r-HuEPO) (EPOGEN) for Treatment of Anemia in Patients With End-Stage Renal Disease Receiving Dialysis (North American Phase 2 Study in EPO-Naïve Subjects), and Study 202: A Randomized Study of ARANESP and Recombinant Human Erythropoietin (r-HuEPO) for Treatment of Anemia in Predialysis Chronic Renal Failure Subjects (European Phase 2 Study in Pre-Dialysis, EPO-Naïve Subjects). The two studies in patients who had previously been on a stable dose of EPO were: Study 117: A Randomized Double-blind, Non-Inferiority Study of IV ARANESP Compared to IV Recombinant Human Erythropoietin (EPO) for Treatment of Anemia in Patients with End-Stage Renal Disease (ESRD) Receiving Hemodialysis, and Study 200: A Randomized, Comparative Study of ARANESP Recombinant Human Erythropoietin for Prevention of Anemia in Subjects With Chronic Renal Failure Receiving Dialysis.

Studies In Erythropoietin-Naïve Patients

In Study 211, 120 subjects with chronic renal failure, on dialysis, were randomized 3:1 to receive 0.45 mg/kg of Aranesp QW or 50 U/kg of EPO administered T.I.W. IV or s.c. for 20 weeks.

In Study 202, 160 subjects with chronic renal failure, but not receiving dialysis, were randomized in a 3:1 ratio to receive Aranesp 0.45 mg/kg QW or EPO 50 U/kg BIW s.c. for ≤ 24 weeks, with both agents to be administered subcutaneously.

For both studies, dose adjustments were to be made, if necessary to achieve a hemoglobin increase of ≥ 1.0 g/dl above baseline, to within a target range of 11-13 g/dl.

The primary efficacy endpoint in both studies was the proportion of subjects achieving a hemoglobin target, defined as a hemoglobin ≥ 1.0 g/dl from baseline and a hemoglobin concentration of ≥ 11.0 g/dl during the study. The time points for assessment of the endpoint were 20 and 24 weeks in Study 211 and 202, respectively.

Efficacy results

In Study 211 the hemoglobin target was achieved by 72% (95% CI: 62%, 81%) of the 90 patients treated with Aranesp and 84% (95% CI: 66%, 95%) of the 31 patients treated with Epoetin alfa. The mean increase in hemoglobin over the initial 4 weeks of Aranesp was 1.10 g/dl (95% CI: 0.82 g/dl, 1.37 g/dl).

In Study 202 the primary efficacy endpoint was achieved by 93% (95% CI: 87%, 97%) of the 129 patients treated with Aranesp and 92% (CI: 78%, 98%) of the 37 patients treated with Epoetin alfa. The mean increase in hemoglobin from baseline through the initial 4 weeks of Aranesp treatment was 1.38 g/dl (95% CI: 1.21 g/dl, 1.55 g/dl).

Studies In Patients Previously Stable on Erythropoietin

The objectives of both of these studies were to show that Aranesp was not inferior to EPO for the treatment of anemia in patients with ESRD receiving dialysis, and to compare the safety of the two agents.

Study 117, the North American pivotal phase 3 study, was a randomized, double-blind, active control study of Aranesp versus EPO for the maintenance of hemoglobin in subjects with ESRD receiving hemodialysis. Study 200, the European/Australian study, was similar in design, except that it was open-label, and subjects could be receiving hemodialysis or peritoneal dialysis.

The primary efficacy endpoint was the change in hemoglobin from baseline through the evaluation period (Week 21-28).

Subjects were to be on a stable regimen of Epoetin, with a baseline Hgb between 9.5-12.5 g/dl at the time of enrollment. In Study 200, subjects could be receiving Epoetin alfa or Epoetin beta at baseline, whereas Study 117 enrolled subjects on Epoetin alfa only. After a 2-week screening and baseline periods, subjects were to be randomized 2:1 to Aranesp or Epoetin alfa or beta. Subjects assigned to Epoetin were to continue on their previous dose of Epoetin. Subjects assigned to Aranesp were to switch to Aranesp, at a total weekly starting dose that was based on the total weekly Epoetin dose at the time of randomization using the proportionality 1 mg Aranesp to 200 units Epoetin. Hemoglobin was to be maintained within a target range of –1.0 to +1.5 g/dl of the baseline hemoglobin and between 9-13 g/dl for up to 28 weeks, with dose adjustments as needed per protocol-specified algorithms.

Efficacy Results

The results of Study 117 showed that the distributions of changes in hemoglobin from baseline through the evaluation period were similar for the two treatment groups. The prospectively defined primary efficacy endpoint, the mean change in hemoglobin, adjusted for center and baseline hemoglobin concentration, was similar in the Aranesp and Epoetin arms: 0.24 ± 0.10 g/dl versus 0.11 ± 0.07 g/dl (mean ± SEM) respectively. The difference between groups was 0.13 g/dl (95% CI: -0.08, 0.33). The lower boundary of the 2-sided 95% CI was above the protocol-specified non-inferiority margin of –1.0 g/dl, providing support that Aranesp was not inferior to Epoetin in maintaining Hgb in this study.

For Study 200, the results showed that the change (mean ± SD) in Hgb from baseline to the evaluation period was similar in the Aranesp and Epoetin arms (0.05 ± 0.80 versus 0.00 ± 0.87 g/dL, respectively). After adjustment for covariates (center, frequency of Epoetin dosing at baseline, modality of dialysis, route of administration, and baseline Hgb concentration), the difference in the mean change in Hgb between the 2 groups was 0.03 g/dL (95% CI: -0.16, 0.21). The lower limit of the 2-sided 95% CI was above the protocol-specified non-inferiority margin of -0.5 g/dL, providing support that Aranesp was not inferior to Epoetin in maintaining the mean Hgb concentration in this study.

Safety Analyses: Relation between adverse events, hemoglobin, and hemoglobin rate of rise

The development program of Aranesp for the anemia of chronic renal failure (CRF) indication evaluated the use of Aranesp in the settings of pre-dialysis, peritoneal dialysis, and hemodialysis. In addition, it assessed two types of Aranesp use: 1) the treatment of anemia in subjects who had not been treated previously with erythropoietins (i.e., anemia “correction” studies); and 2) as maintenance therapy for patients whose anemia had been treated with a stable regimen of Epoetin alfa (EPO) prior to study enrollment (i.e., “conversion” studies). Details of FDA’s safety analyses can be found in the Medical Officer’s clinical review of Aranesp for anemia of CRF.[50]

Overall, the safety database included 1598 Aranesp-treated subjects and 600 EPO-treated subjects, with median lengths of exposure of 24 and 28 weeks, respectively. The substantive investigations were either active-controlled studies using EPO as a comparator, or uncontrolled studies. The vast majority of the clinical experience was unblinded.

Given the lack of placebo-controlled studies and the pattern of AEs observed, characterization of the safety of Aranesp during review of the marketing application was not straightforward. Adverse events occur frequently in the CRF patient population, and treatment emergent AEs had to be assessed against this background. Moreover, cardiovascular disease is prevalent in the CRF patient population, and cardiovascular events occur fairly commonly, yet they also constitute a primary manifestation of excessive erythropoiesis, as noted above.

FDA undertook a number of approaches in its assessment of the Aranesp safety database, beyond simple comparisons of AE rates in Aranesp- and Epoetin alfa-treated subjects and subgroups. One of these approaches involved analyses of AEs with putative mechanisms Involving hemodynamic and/or rheologic factors, an approach that involved four analyses:

A) Analysis of AEs by Hgb Concentration. FDA assessed AEs by Hgb concentration as determined on the week of the reported event. This purpose of this analysis was to provide information on potential associations between AEs and specific Hgb levels.

B) Analysis of AEs by Hgb Rate of Rise. FDA assessed AEs by Hgb rate of rise (ROR) during the weeks preceding reported AEs. The objective of this analysis was to assess potential associations between AEs and specific rates of Hgb increase.

C) Analysis of AEs by Hgb Rate of Decline. This complementary analysis assessed rates of AEs by Hgb rate of decline during the weeks preceding the events.

D) Examination of Potential Interaction Between Hgb Concentration, Hgb Rate of Rise and AEs: FDA combined all AEs with putative mechanisms involving hemodynamic and/or rheologic factors, and examined the potential interactions between the rate of these AEs, Hgb concentration, and Hgb ROR.

General Methods:

Weekly Hgb values were classified by both 1-g/dL range and quintile. Ranges were defined as: <10 g/dL, >10 to <11 g/dL, >11 to <12 g/dL, >12 to <13 g/dL, >13 to <14 g/dL, and >14 g/dL. Quintiles were ascertained as: <10.1 g/dL, 10.1 to <10.8 g/dL, ³10.8 to <11.4 g/dL, ³11.4 to <12.2 g/dL, and ³12.2 g/dL.

For each subject-week, the slope of the preceding Hgb-time relation was determined, when possible, using the following approach:

1). The slope of the Hgb-time relation leading up to each date was calculated using Hgb values obtained over a 2-week period (i.e., 3 Hgb values).

2). Missing Hgb values were not interpolated.

3). Hgb values were construed as having been obtained on the week indicated, i.e., the actual date was not used in calculations. Slopes were expressed as weekly change in Hgb concentration.

4). If <2 Hgb values were reported over a 2-week (3-value) period, such that a slope could not be calculated, an attempt was made to calculate slope over a 4-week period.

5). Positive and negative slopes were analyzed separately, with slopes of 0 classified with the positive slopes.

6). Slope (m) was classified by group, as follows:

m > 0.2 and < 0.25 g/dL/week (1 g/dL per <5 to 4 weeks)

m > 0.25 and < 0.333 g/dL/week (1 g/dL per <4 to 3 weeks)

m > 0.333 and < 0.5 g/dL/week (1 g/dL per <3 to 2 weeks)

m > 0.5 and < 1 g/dL/week (1 g/dL per < 2 to 1 week)

m > 1 g/dL/week (1 g/dL per <1 week)

Each AE reported was linked, by reported week of occurrence, to its associated weekly Hgb value range, quintile, and slope. Multiple AEs were linked by pathophysiologic mechanism, e.g., fluid overload included edema, dyspnea, orthopnea and pleural effusion. Congestive heart failure (CHF), abnormal ejection fraction and pulmonary edema were grouped together. Cerebrovascular disorders included cerebral ischemia, intracranial hemorrhage (ICH), and cerebral/subarachnoid hemorrhage. Angina was grouped with coronary artery disease, myocardial ischemia, and chest pain (non-specific chest pain was not included in this category). A category representing thrombosis/ischemia (but omitting TVA) was constructed including the terms arterial occlusion, embolism, arteriosclerosis, carotid stenosis, claudication, peripheral vascular disease, ischemic necrosis, gangrene, superior vena caval syndrome, phlebitis, thrombophlebitis, arterial/venous thrombosis, intestinal ischemia, pulmonary embolism, and TIA.

Absolute Hgb Values (Table 1):

Serum Hgb was analyzed both by quintiles, and by 1-gram/dL categories. For each quintile and category, the denominator used was the number of weekly Hgb values observed that fit that particular category, divided by 1000 (i.e., the number of events per 1000 weekly Hgb observations). The Ns are given at the bottom of Table ~~Table~~ .

For Aranesp -treated subjects, there were trends suggesting possible associations between reported Hgb values >13 and seizures, hypertension (HTN), and arrhythmias, though the latter appeared to be associated with Hgb values >11 g/dL, as well. Importantly, Hgb values of >13 g/dL did not appear to be associated with increased risks of these events. Of note, for ARANESP-treated subjects, Hgb values <10 g/dL appeared to be associated with excess risks of fluid overload, CHF, pulmonary edema, acute MI and TVA, whereas these risks were not apparent at Hgb values ³10 g/dL.

Table 1: FDA Analysis of Relation Between Serum Hgb and AEs With Putative Mechanism Involving Hemodynamic and/or Rheologic Factors: Combined Data (Rates are Events /1000 weekly Hgb Observations)

Hgb Rate of Rise (Table 2):

Adverse events are shown for Hgb rates of increase and decrease in Table 2. Rates were determined by whole fractions of a gram of Hgb (i.e., 1.0, 0.5, 0.33, 0.25, 0.20, and 0.10 g/dL/week). The denominators used for each quintile and category were the number of weekly Hgb slopes that fit that particular category, divided by 1000 (i.e., the number of events per 1000 weekly Hgb observations, table bottom).

For Aranesp-treated subjects, there appeared to be excess risk of HTN, pulmonary edema, cardiac arrest and TVA associated with Hgb ROR >0.5 and particularly 1.0 g/dL/week. There also appeared to be an association between rapid Hgb rise and fluid overload, acute MI and seizures, although the association between Hgb rate of rise and these events was less clear.

FDA also assessed the rates of these AEs in the Aranesp treatment group, with subgroups by history of cardiovascular disease (CVD). Across Hgb and Hgb ROR categories, event rates generally followed similar patterns for subjects with and without a history of CVD, though rates were higher in CVD(+) subjects. There were 2 apparent exceptions:

1) CHF was strongly associated with the extremes of Hgb categories (both <10 and >14 g/dL) in the CVD(+) group, whereas there was only a weak association between CHF and Hgb <10 g/dL in the CVD(-) subgroup. A Hgb ROR exceeding 0.5 was strongly associated with CHF, both in CVD(+) and CVD(-) subgroups.

2) There was no clear association between angina and Hgb ROR in either subgroup; however, a Hgb <10g/dL was strongly associated with angina in the CVD(+) group.

Hgb Rate of Decline:

With respect to falling Hgb in Aranesp-treated subjects (rate of Hgb decrease preceding AEs), there were apparent associations between a >1 g/dL weekly decline in Hgb and worsened anemia, CHF, pulmonary edema, acute MI, cardiac arrest, angina, arterial occlusion, and death, though the numbers of subject-weeks and events in this category were limited.

Interaction Between Hgb Concentration, Hgb Rate of Rise, and AEs:

For the Aranesp treatment group, FDA combined all of the AEs with putative mechanisms involving hemodynamic and/or rheologic factors, and examined the interaction between the rate of these AEs, Hgb concentration and Hgb ROR. These events included: accelerated HTN, fluid overload, edema, dyspnea, orthopnea, pleural effusion, pulmonary edema, CHF, abnormal ejection fraction, angina, coronary artery disease, myocardial ischemia, chest pain (cardiac), arrhythmia, syncope, cardiac arrest, impaired consciousness, encephalopathy, seizure, cerebrovascular disorder, TIA, cerebral ischemia, ICH, subarachnoid hemorrhage, arterial occlusion, arteriosclerosis, carotid stenosis, claudication, gangrene, ischemic necrosis, peripheral ischemia, arterial embolism, phlebitis, superior vena caval syndrome, thrombophlebitis, arterial/venous thrombosis, and intestinal ischemia. Each AE was linked, by week of reported occurrence, to its corresponding Hgb category, as well as to its appropriate Hgb ROR category.

Table 2: FDA Analysis of Relations Between Hgb ROR and Rate of Fall and AEs with Putative Mechanism Involving Hemodynamic and/or Rheologic Factors: Combined Events (Events/1000 Weekly Hgb Observations

The results of this analysis are shown in Table 3. The top panel shows the numbers of subject-weeks that fulfill the criteria for Hgb and Hgb ROR categories. These data serve as the denominators (Ns) for these analyses. The middle panel shows the numbers of events for each Hgb/Hgb ROR category, and the bottom panel shows event rates per 1000 subject-weeks. Each row, representing a particular Hgb ROR category, is summed in the right-most column (“All Hgb”). Note that event rates (bottom panel) tend to be similar for all ROR categories  0.50 g/dL/week, whereas event rates for ROR  0.50 g/dL/week tend to be higher. The individual Hgb categories are summed in the bottom row (“All Slopes”). The lowest event rates are in the >11 to 12 g/dL Hgb category. A slight increase in event rate is evident above this range, with a sharper increase in event rate below this range.

Table 3: Interaction Between Hgb Concentration, Hgb Rate of Rise, and AEs – ARANESP Group

Figure 3: Relations Between Hgb, Hgb ROR, and AEs With Hemodynamic/Rheologic Mechanisms – ARANESP Group. Top panel – all subjects; middle panel CVD(+) subjects; bottom panel CVD(-) subjects

Figure 3 (top) displays these event rates graphically. A general trend towards higher event rates for the lowest Hgb categories (<10; >10 to <11) is apparent at the left side of the graph. Event rates are higher in the row corresponding to a Hgb ROR >0.5 g/dL/week. The middle panel shows the event rates for the subset of subjects with a reported history of cardiovascular disease. Rates are generally higher for this subgroup. Of note, higher event rates appear to be associated with the lowest Hgb class (<10 g/dL), and the highest ROR class (>0.5 g/dL/week). For the subset of subjects without a reported history of cardiovascular disease (lower panel), the events rates tend to be lower, but the trends are similar. Importantly, therefore, even subjects without overt cardiovascular disease appear to incur excess risk with Hgb ROR in excess of 0.5 g/dL/week.

In summary, FDA’s exploratory analyses suggested that higher Hgb concentration, per se, is not associated with increased rates of events that involve hemodynamic or rheologic mechanisms. Importantly, however, the hemoglobin rate of rise appears to be particularly relevant with respect to these events. Specifically, a hemoglobin rate of rise in excess of 0.5 g/dL/week appears to be associated with increased event rates, irrespective of the presence or absence of overt cardiovascular disease. This finding provided the basis for the Warning on cardiovascular events in the Aranesp package insert.

VI. Clinical Studies of Epogen/Procrit for the Treatment of the Anemia Associated with Chemotherapy of Cancer.

In April 1993, Epogen/Procrit was approved for the indication of treatment of anemia in patients with non-myeloid malignancies where anemia is due to the effect of concomitantly administered chemotherapy. The data supporting this supplemental indication was based on pooled data from six randomized, placebo-controlled, double-blind, trials, in a total of 131 anemic cancer patients. 72 of these patients were treated with concomitant chemotherapy regimens that did not contain cisplatin, and 59 patients were treated with regimens that contained cisplatin. Patients were randomized to Procrit 150 units/kg or placebo subcutaneously t.i.w. for 12 weeks.[51]

Efficacy Results

Proportion of Patients Transfused During Chemotherapy (Efficacy Population)^a

Chemotherapy Regimen	On Study^b		During Months 2 and 3^c
	EPO	Placebo	EPO	Placebo
Regimens without cisplatin	44% (15/34)	44% (16/36)	21% (6/29)	33% (11/33)
Regimens containing cisplatin	50%(14/28)	63%(19/30)	23% (5/22)	56% (14/25)
Combined	47% (29/62)	53% (35/66)	22% (11/51)^d	43% (25/58)

^aLimited to patients remaining on study at least 15 days.

^b Includes all transfusions from day 1 through the end of the study.

^cLimited to patients remaining on study beyond week 6 and includes only transfusions during weeks 5-12.

^dUnadjusted 2-sided p < 0.05.

Data were not systematically collected on tumor response, tumor progression, or survival. The reviewer noted that “Based in part on the percentage of Epoetin alfa and placebo-treated patients who discontinued therapy due to death, disease progression or adverse experiences (22% and 13% respectively, p= 0.25) the clinical outcome in Epoetin alfa and placebo-treated patients appeared to be similar.

FDA noted that Procrit could potentially serve as a growth factor for malignant tumors, however the pivotal studies that had been used to support supplemented approval were not designed to examine tumor response rate, time to progression or overall survival. Because of this concern, Amgen agreed to the following post-marketing commitment to conduct a study to assess:

“The effect of Epoetin alfa on initial response rate and response rate at the completion of chemotherapy, site of first relapse and overall survival will be investigated in a randomized, double-blind, placebo-controlled phase IV study in patients with limited stage small cell lung cancer.”[52]

VII. Post-Marketing Study to Assess for Tumor Stimulatory effects of Epogen/Procrit: Study N93-004

Protocol, Study N93-004, “The Effect of r-HuEPO in Patients with Small Cell Cancer (SCLC): A Randomized, Double blind, Placebo-controlled Trial” to FDA on January 7, 1993. In July 2001, Amgen and Ortho Biotech LP notified FDA of its intention to prematurely terminate the study after accrual of 225 subjects, (instead of the planned accrual of 400), due to slow accrual rates. FDA stated that the results of than study should be submitted and, based on review of the data, a determination would be made as to whether additional studies would be required.

Study N93-004 was a randomized, double-blind placebo-controlled trial conducted at 35 sites in the United States. The primary objective was to determine the effect of Procrit on tumor response in SCLC in patients receiving treatment with etoposide and cisplatin. The primary endpoint was determination of the objective response rate (defined as partial response plus complete response) after 3 cycles of chemotherapy. The secondary endpoints were: effects of Procrit on survival, hemoglobin, and transfusion rates. Eligible subjects were those with newly diagnosed limited or extensive stage small cell carcinoma of the lung. Subjects were randomized to receive chemotherapy with etoposide and cisplatin every 3 weeks for 3 cycles with either Procrit 150 IU/kg t.i.w. or placebo. Subjects continued to receive study medication until approximately 3 weeks after the final cycle of chemotherapy. No target hemoglobin was specified, but the study medication was to be held if the hemoglobin rose above 16 g/dl and restarted at a 50% dose reduction when the hemoglobin fell to less than 14 g/dl.

The study was designed with a sample size of 400 (200 patients/arm). The sample size was selected in order to be able to exclude, with 90% power, an absolute decrement in overall response rate of 15% in the Procrit-treated arm as compared to placebo, based on the 95% confidence interval around the observed difference in response rate. The assumed response rate in the placebo arm was 60%.

In the intent-to-treat population of 224 subjects, there were 115 in the placebo and 109 in the Procrit arm. There were no differences between the two groups in baseline demographics or other baseline entry variables (including baseline hemoglobin), with the exception of a slightly higher proportion of subjects in the Procrit arm had extensive stage SCLC than in the placebo arm (66% versus 59%). In both arms, 72% received at least 3 cycles of chemotherapy. The median number of cycles was 4 for both arms. The median doses of both etoposide and cisplatin were not significantly different between the groups.

Efficacy Results

For the primary endpoint, the tumor response (CR + PR) after 3 cycles of chemotherapy, the results were as follows (intent-to-treat population):

	Placebo (n=115)	Procrit (n=109)
No. having CR/PR	77	79
Tumor response rate	67%	72%
95% CI around observed response rate	58-67%	64-81%
95% confidence interval around observed difference in ORR	-6% to 18%

The observed difference in tumor response rates between the Procrit and placebo arms was 6% (95% CI: -6%% 18%). The lower bound of the 95% confidence interval around the difference in response rates was –6%, indicating that in this trial, the response rate observed in the Procrit arm would not be more than 6% lower than that of the placebo arm. The methods used to collect data on tumor response and patient follow-up did not permit the determination of time to progression.

Safety Results

Thrombotic Vascular Events:

Twenty-two percent of the subjects in the Procrit arm and 23% of the subjects in the placebo arm expired at least one thrombotic vascular event. The incidences of specific subtype of thrombotic vascular event were generally similar between the two treatment arms with the exceptions of chest pain, which was reported by 14% of placebo treated subjects as compared to 7% of Procrit-treated subjects, and vascular (extracardiac) disorders, which was reported in 4% of placebo patients and 10% of Procrit-treated patients.

Incidence of Thrombotic Vascular Events

In Study N93-004

Preferred Term	Placebo N=115	Procrit N=109
Chest pain	16 (14%)	8 (7%)
Chest pain substernal	0 (0)	1 (1%)

Vascular (extracardiac)	5 (4%)	11 (10%)
Cerebrovascular disorder	2 (2%)	3 (3%)
Phlebitis	1 (1%)	3 (3%)
Thrombophlebitis, deep	1 (1%)	2 (2%)
Thrombophlebitis	1(1%)	1 (1%)
Peripheral ischemia	0 (0)	1 (1%)
Phlebitis, superficial	0 (0)	1 (1%)

Platelet, bleeding, and clotting	7 (6%)	5 (5%)
Thrombosis	3 (3%)	2 (2%)
Pulmonary embolism	2 (2%)	2 (2%)
Thrombosis venous arm	2 (2%)	1 (1%)
Thromboembolism	0 (0)	1 (1%)

Myo-, endo-, pericardial- and valve	1 (1%)	2 (2%)
Myocardial infarction	0 (2)	2 (2%)
Angina	1 (1%)	0 (0)

Heart rate and rhythm	0 (0)	1 (1%)
Cardiac arrest	0 (0)	1(1%)

Survival[53]

A total of 201 of the 224 subjects enrolled in this study died at some time during the study treatment period or 3 year follow-up. The overall mortality rate in the Procrit arm (100 of 109 subjects, 92%) was similar to that in the placebo arm (101 of 115 subjects, 88%). The median duration of survival (based on Kaplan Meier estimates) was 10.5 months among Procrit-treated subjects compared with 20.4 months among placebo-treated subjects.

Summary of Survival Over Time in N93-004 (Intent-to-treat Population)

The FDA biostatistical reviewer confirmed the sponsor’s survival analysis, presented in the table below (Sponsor’s Table 17)[54]:

Sponsor’s ITT Survival Analysis in Months^*

(Table 17)

	PLACEBO			PROCRIT
	Estimate	95% CI		Estimate	95% CI
Quartile		Lower	Upper		Lower	Upper
75%	5.9	3.5	7.7	6.6	4.3	7.6
Median	10.4	8.3	12.9	10.5	9.2	12.9
25%	23.3	15.3	27.3	17.1	14.0	20.1

^* Note: To convert days to months, the sponsor used a divisor of 28 days rather than the more usual 30.437 days, which takes into account leap year.

A total of 201 out of 224 subjects enrolled in this study died at some time during the study treatment period or during the 3-year follow-up. A somewhat higher proportion of subjects assigned to the Procrit arm had extensive stage disease at diagnosis (66%) compared to the placebo arm (59%). Since stage of disease (limited vs. extensive) was a stratification factor, the FDA statistical reviewer examined the descriptive stratified Kaplan-Meier analysis provided by the sponsor. Kaplan-Meier survival plots were comparable in the Procrit and placebo arms through Months 17 to 18 after study start. As for the overall ITT population, the variability after study completion along with the small number of subjects in the two extent of disease subgroups does not permit any conclusive statement to be made.

Tumor Outcomes

Response duration was not calculated. No data were provided for date of tumor response.

Time to disease progression (TTP) could not be accurately calculated. In Protocol N93-004 subjects were allowed to withdraw early for progressive disease. In these cases, the date of discontinuation was captured and not the actual date of progression. For the 25 subjects who were discontinued from the study for disease progression, the progression date was on or before the date of discontinuation.”

VIII. Clinical Studies of Weekly Dosage Schedules of Epogen/Procrit for Treatment of Anemia Associated with Cancer Chemotherapy

The recommended starting dose of Epogen/Procrit in cancer patients receiving chemotherapy in the package insert is 150 Units/kg t.i.w. However, many community oncologists administer Epogen/Procrit at a dose of 40,000 Units once a week. FDA is currently reviewing the results of a study[55], in which 344 patients with cancer receiving chemotherapy were randomized to receive either placebo or Epogen/Procrit 40,000 U/week. The patient population consists of anemic patients with a variety of malignancies, who were receiving standard therapy. Given the heterogeneity of the population and cancer treatments, no comparative assessments are possible regarding tumor outcomes (response rates, time-to-progression) or survival. Analyses of adverse events, including incidence of thrombotic vascular events is in progress.

IX. Clinical Studies of Aranesp in the Treatment of the Anemia of Cancer Chemotherapy

Aranesp (darbepoetin alfa) received supplemental approval “for the treatment of anemia in patients with non-myeloid malignancies where anemia is due to the effect of concomitantly administered chemotherapy” in July 2002.[56]

The data supporting this supplemental indication was based on the results of Protocol 980297: A Double-blind, Placebo-Controlled, Randomized, Study of NESP for the Treatment of Anemia in Lung Cancer Receiving Multicycle Platinum Containing Chemotherapy. This was a multicenter, multinational study in which 320 patients were enrolled and randomized 1:1 to receive either Aranesp 2.25 mg/kg QW (treatment arm) or placebo. Eligibility criteria included lung cancer (either small cell carcinoma or non-small cell carcinoma) a cancer treatment plan of at least 12 additional weeks of platinum-containing chemotherapy, and anemia (hemoglobin £ 11 g/dl). The primary endpoint was the estimated Kaplan-Meier proportion of subjects who received RBC transfusions between week 5 and the end of the treatment phase (EOTP). Week 5 was specified since hematologic responses to Aranesp are not observed until 3-6 weeks after the initiation of therapy.

Efficacy Results

The primary efficacy analysis was conducted in patients who had completed the first 4 weeks of study. In this analysis, patients who withdrew or discontinued from the study after week 4 for death or disease progression were censored, while those who withdrew for any other reason were imputed to be transfused (treatment failures for primary endpoint).[57] The Kaplan-Meier proportion of patients transfused was 51% for the patients in the placebo arm versus 21% in the Aranesp arm. The same analysis was also performed in the intent-to-treat dataset, which included patient information across all 12 weeks on. The comparable figures were 60% for the Kaplan-Meier proportion of transfused patients in the placebo arm versus 26% in the Aranesp arm.

Proportion of Patients Transfused in Weeks 5 through End-of-Treatment

In Protocol 980297

	Placebo arm N=149	Aranesp arm N=148
Number of patients transfused	74	39
Kaplan-Meier estimated proportion	51%	21%
95% CI	43, 60	15, 28

Secondary analysis

Proportion of Patients Transfused Between Entry and End-of-Treatment

In Protocol 980297

	Placebo arm N=158	Aranesp arm N=156
Number of patients transfused	89	53
Kaplan-Meier proportion	60%	26%
95% CI	52, 68	20, 33

Safety analyses

There were 314 protocol 980297 patients who received study drug; 156 were randomized to Aranesp and 158 to placebo. A single patient randomized to the Aranesp arm failed to receive the study drug. In the efficacy analyses, he remained in the Aranesp arm to which he had been randomized but for Safety studies he was switched to the placebo arm. Thus safety data are provided for 155 Aranesp-treated patients and 159 placebo-treated patients.

Time to Progression

In the long-term follow-up to the pivotal, phase 3 study (NESP 980297), the median observation period was 12 months for disease progression and 11 months for survival. Long-term follow-up analyses were based on the subjects included in the safety analysis set of Aranesp disease progression. All 314 subjects in the safety analysis set are included in the analysis of disease progression, death, and disease progression or death.

Ninety-four subjects (61% of the safety analysis set) in the Aranesp group and

110 subjects (69%) in the placebo group had disease progression either during NESP

980297 or during the long-term follow-up period evaluated. The hazards ratio for disease progression comparing the Aranesp group to the placebo group was 0.70 (95% CI: 0.53, 0.92) based on the Cox-proportional hazards model that includes only treatment group as an independent variable and was 0.71 (95% CI: 0.54, 0.94) after adjusting for tumor type and region. Figure 1 shows the Kaplan-Meier curve for time to disease progression by treatment group for subjects in the safety analysis set. The median time to disease progression in the NESP group was 29 weeks compared with 22 weeks in the placebo group.”[58]

Survival

“Sixty-six subjects (43%) in the NESP group and 78 subjects (49%) in the placebo group

died within 30 days of the last dose of study drug or during the long-term follow-up period evaluated. The hazards ratio was calculated as 0.80 (95% CI: 0.58, 1.11) based on the Cox-proportional hazards model that includes only treatment group as an independent variable and after adjusting for the effects of tumor type and region. Figure 2 shows the Kaplan-Meier curve for time to death by treatment group for subjects in the safety analysis set. The median time to death in the NESP group was 43 weeks compared with 35 weeks in the placebo group.”[59]

Progression-free Survival

“One hundred five subjects (68%) in the NESP group and 120 subjects (75%) in the placebo group had disease progression or died within 30 days of the last dose of study drug or during the long-term follow-up period evaluated. The hazards ratio was calculated as 0.73 (95% CI: 0.56, 0.95) based on the Cox-proportional hazards model that includes only treatment group as an independent variable and was 0.74

(95% CI: 0.57, 0.97) when adjusted for tumor type and region. This indicates that, on average, subjects in the placebo group had a 35% higher odds of either disease progression or death (i.e., 1 / 0.74) as compared with subjects in the NESP group.”

“The Kaplan-Meier plot of time to either disease progression or death for all subjects in the safety analysis set by treatment group is shown in Figure 3. The Kaplan-Meier plot is very similar to that for time to disease progression, with an apparent prolonged time to either disease progression or death observed in the NESP group compared with the

placebo group. The median time to disease progression or death was 26 weeks for the

NESP group and 20 weeks for the placebo group.”[60]

Exploratory Analyses of Relationships between Rate of Rise of Hemoglobin and Cardiovascular/Thrombotic Events in the Aranesp Integrated Summary of Safety (ISS)

The ISS consisted of 873 patients from 6 major studies who received Aranesp, 115 patients who received Epogen/Procrit in the active-control arms, and 221 patients who received placebo. All patients also received concomitant chemotherapy. The mean observation period for the majority of subjects was 12 weeks on treatment plus 4 succeeding weeks off treatment.

INCIDENCE OF AEs ASSOCIATED WITH EPOETIN ALFA

	ISS			Protocol 980297
	Aranesp	Placebo	Procrit	Aranesp	Placebo
	n=862	n=217	n=113	n=155	159
Hypertension+	3.7%	3.2%	2%	9 (6%)	6 (4%)
Convulsions**	0.6%	0.5%	1%	0	1 (1%)
Thrombotic events$	6.2%	4.1%	1%	7 (5%)	5(3%)

+ Hypertension included preferred terms Hypertension and Hypertension aggravated

**Convulsions included the Preferred terms Convulsions, Convulsions Grand Mal and Convulsions Local

$ Thrombosis includes Preferred terms Thromboembolism, Thrombophlebitis Deep, Thrombosis, Thrombosis Venous, Thrombosis Venous Deep, and Pulmonary Embolism

When the incidence for all of the Preferred Terms are combined there was no major difference in the incidence of thrombotic events between study arms. This was attributable to the large variety of types of venous adverse events. Notably, pulmonary emboli and related thromboembolism were restricted entirely to the Aranesp subjects. There was also a trend toward higher incidences of thrombophlebitis thrombophlebitis deep, and thrombosis in Aranesp-treated patients. The incidences of adverse event by specific preferred terms across the treatment groups in the ISS are displayed in the following table.

Incidence of Thrombotic Events By Preferred Term

Preferred term	rHuEPO	Aranesp	Placebo
	N= 115	N=873	N=221
Embolism, pulmonary	0	11	0
Thromboembolism	0	1	0
Thrombophlebitis	0	4	0
Thrombophlebitis , deep	0	2	1
Thrombosis, venous	1	8	2
Thrombosis, venous deep	3	27	6
Thrombosis	0	5	0
Total # AEs	4 (4.7%)	58 (7.4%)	9 (4.1%)
Total pts with AEs	85	781	221

The FDA requested that Amgen provide an analysis exploring the correlations between rate of rise of hemoglobin (ROR) and the following: death, hypertension, cerebrovascular events (seizures, TIAs, “strokes”), cardiovascular events (myocardial infarction, CAD, arrhythmia, angina, CHF, cardiac failure, cardiac arrest), thrombotic events, ischemic and peripheral vascular disease and pulmonary edema. For comparison, the results of a similar analysis using Aranesp-treated patients enrolled in Amgen-sponsored studies for treatment of chronic renal failure were also provided. .

Adverse event/week per 1000 events[61] as a function of Rate of Rise of Hgb

Adverse Event Term	Patients with CRF N=1598			Patients with Cancer N=873
	Rate of rise in hgb/week			Rate of rise in hgb/week
	< 0.1	0.5 to 1.0	1.0 or >	< 0.1	0.5 to 0.67	0.67 or >
Hypertension	7.5	19.9	13.1	2.1	1.8	5.2
Seizures	0.4	1.0	1.9	1.1	0	1.3
Vascular*	1.7	3.8	7.5	1.1	7.1	3.9
Fluid overload	15.4	18.9	35.4	25.6	26.5	29.5
Cardiac arrest	0.0	1.0	1.9	0	0	0
Myocardial infarction, acute	0.4	0.3	1.9	0	0	0
Pulmonary edema	1.5	3.1	7.5	0	0	0

*thrombosis, ischemia, infarction

The analysis provided suggests that the incidences of hypertension and vascular events are increased in patients with higher ROR compared to those who had less rapid ROR. [62]

The ISS database was also used to examine for a possible relationship between the hemoglobin value at the time of an event and the incidence of specific adverse events, possibly or probably associated with Aranesp. The FDA reviewer concluded that: “The major events of interest were death, hypertension, cardiac events, cerebrovascular events, and thrombotic events. There did not appear to be a relationship between increasing hemoglobin concentration and increased risk of adverse event across any the adverse event groups evaluated. Death (n=18) was the only event group in which the highest event rates for Aranesp and Procrit were in the category representing the highest hemoglobin increase (> 0.667 g/dl/week hemoglobin slope). Of these 18 subjects, 12 received Aranesp, 4 received placebo, and 2 received Procrit. The death rate was 13.6 per 1000 subject weeks and 7.8 per 1000 subject weeks for the Procrit and Aranesp groups, respectively. Of the 18 subjects how died in this slope category, 13 died of disease progression, 3 died of sepsis, and 2 died of clinically suspected but not objectively confirmed pulmonary embolisms.”[63]

Incidence of Selected Adverse Events Occurring

Within 14 days of Rapid Rise of Hgb[64]

	Placebo n=221	Procrit N=115	Aranesp N=873
Number of patients with hypertension (%)	5 (2.2%)	0	12 (1.3%)
Number of patients with Convulsions (%)	0	0	2 (0.2%)
Number of Patients with Thrombosis (%)	5 (2.2%)	2 (1.7%)	14 (1.6%)

Because the combined term “thrombosis” included a small number of specific thrombotic events that were clearly increased in the Aranesp-treated group (notably pulmonary emboli), an assessment was made of the clinical course of these subjects. The displayed in the following table summarizes the dose, schedule, and clinical treatment course of patients who suffered pulmonary emboli.

Assessment of ROR and Maximum Hemoglobin Levels in Relationship to the Development and Time of Onset of Pulmonary Emboli in Aranesp-Treated Patients

Pulm.emboli Subject	Dose schedule Aranesp	Dose ug/kg	Week of AE	Rapid ROR	Week rapid ROR	Max. incr. Hgb ug/dL	Max. hgb ug/dL
10101144	QW	4.5	4	Yes	4	3.1	13.7
10116003	QW	1.0	3	No	-	2.8	13.9
12003012	Q3W	12.0	3	No	-	6.5	15.5
12502021	Q3W	9.0	1	Yes	11	4.4	14.2
12506009	Q3W	9.0	1	Yes	1	0.7	11.3
12609001	Q3W	4.5	11	Yes	17	4.5