The myeloid growth factors G-CSF and GM-CSF have had an impact on the supportive care of cancer patients as well as on the strategies utilized in chemotherapy dose intensification. Therapy with these factors has not been
The myeloid growth factors G-CSF and GM-CSF have had an impact on the supportive care of cancer patients as well as on the strategies utilized in chemotherapy dose intensification. Therapy with these factors has not been associated with improvements in survival, and hence an examination of their effects on economic outcomes is central to rational decision making regarding their use. Erythropoietin therapy has been shown to decrease transfusion requirements in anemic cancer patients undergoing chemotherapy and to improve the quality of life in responding patients. The available data on the economic outcomes associated with the use of these three factors in oncology practice are reviewed.
Escalating health care costs mandate that a rigorous assessment of economic outcomes be included in the overall evaluation of a new technology. At the same time, the effect of a new treatment on the quality of life of cancer patients is increasingly recognized as a legitimate goal of treatment and an important outcome to consider in the overall assessment of the costs and benefits of a new therapy. Because it is likely that, with time, the results of these two less traditional outcome analyses in oncology will increasingly guide clinical decision making, it is important that they accurately reflect the true impact of the technology in community practice and that clinicians fully understand the methodologies involved in these analyses and their findings.
The hematopoietic growth factors are a class of glycoproteins involved in the regulation of hematopoiesis and mature effector cell function. The recombinant human forms of three of these factors-granulocyte colony-stimulating factor (G-CSF, filgrastim, Neupogen), yeast-derived granulocyte-macrophage colony-stimulating factor (GM-CSF, sargramostim, Leukine), and erythropoietin (EPO, epoetin alfa, Epogen, Procrit)-have shown sufficient efficacy in clinical trials to be approved for community use in oncology practice in the United States. While all three factors have shown significant efficacy in the supportive care of oncology patients in various clinical settings, none has been shown to increase survival or tumor response rates; hence, economic and quality of life outcome assessments have become particularly important components of the overall assessment of these new technologies.
Unfortunately, there are relatively few directly measured data available regarding the economic or quality of life impacts of any of these factors, and the available analyses are based largely upon assumptions drawn from the clinical effects measured directly in randomized clinical trials. For G-CSF and GM-CSF, some directly measured data regarding impact on health care charges are available, but few data are available regarding the effects of these factors on the quality of life of the cancer patient. For EPO the situation is reversed: The impacts of this factor on the quality of life of the cancer patient have been directly measured in both phase III and phase IV studies, but there are no published data regarding the economic impact of the use of this factor in oncology practice.
In approaching the study of the economic outcomes of relatively expensive new drugs such as the hematopoietic growth factors, it is important to state clearly the perspective and scope of the analysis. While cost studies focused on a particular component of the health care system, such as the hospital pharmacy, are important for budgeting purposes, the results of these studies cannot be used to estimate the true cost impact of the new technology or to support responsible clinical decision making. For these purposes, the perspective of the cost analysis should be the health care system or society as a whole, and the scope of the analysis should include all the costs, both direct and indirect, associated with the new therapy, including the treatment of any drug-related toxicities and all of the direct and indirect costs that the new therapy eliminates.
Studies such as these, which compare the health care costs associated with two alternative approaches to treatment, are termed cost-minimization studies, and are typically utilized when the two approaches yield equivalent clinical outcomes such as tumor response rates and survival. If a cost-minimization analysis shows that the new therapy is associated with increased overall costs, decisions regarding its use will then depend upon which outcomes, if any, are improved. In these broader scope analyses, termed cost-benefit studies, the increased costs associated with the new treatment are balanced against improvements in nonclinical outcomes such as improved quality of life, improved functional status, and decreased indirect costs, including lost days from work.
The hematopoietic growth factors are unique new technologies in that they are applied to a broad spectrum of patients at varying risks for the complications they may prevent, such as febrile neutropenia, prolonged hospitalization, and red blood cell transfusions. These complications have associated costs, and the initial question about economic impact focuses on to what extent the savings associated with preventing these complications offset the cost of the growth factor. These cost offsets will depend, in part, upon the risk of the prevented complication in that subset of patients.
Figure 1 presents the results of a hypothetical cost-minimization study for a hematopoietic growth factor. In this example, as the risk of a preventable complication, such as severe anemia or febrile neutropenia, increases (as the magnitude of the efficiency variable rises), the efficiency of the use of the factor increases and the cost offsets associated with its use increase. At the threshold value of the efficiency variable, these offsets fully cover the cost of the factor, and at higher values, the use of the factor becomes cost saving.
Cost-minimization studies such as these are an important first step in the economic evaluation of the growth factor. They provide two useful pieces of information: First, these studies suggest that in select subgroups of patients, the use of the growth factor can be justified on the basis of direct cost savings alone. Second, they provide estimates of the magnitude of the increase in direct health care expenditures associated with growth factor therapy in the remaining subgroups of patients. These estimates can be used in future cost-benefit analyses that take into account any additional benefits of therapy with the growth factor, such as improved quality of life or decreased indirect costs to the health care system or society.
This paper will review the published data regarding the economic impacts of the hematopoietic growth factors in oncology practice. It will include a brief description of the seminal randomized clinical trials, with an emphasis on their implications for economic analyses, followed by a synopsis of the available cost-minimization and quality of life studies of the use of these factors in the treatment of cancer. Finally, this paper will summarize our current understanding of the cost-benefit balance for these factors and point out gaps in the data where further studies are needed for a more mature assessment of the impact of these agents.
Any cost assessment of hematopoietic growth factors must begin with a review of the state of knowledge regarding the technical aspects of their use, such as dose and schedule, clinical efficacy, and toxicities. For obvious reasons, the review should focus on randomized, controlled clinical trials, to ensure that any effects attributed to the growth factors are actually associated with their use. As will be seen, randomized clinical trials designed to demonstrate clinical efficacy usually do not define the optimal cost-effective use. Still, these studies are a necessary and important starting point for economic analyses.
The Myeloid Growth Factors in Cancer Treatment
There are three distinct rationales for the use of myeloid growth factors in the care of the cancer patient. The first is as part of supportive care during chemotherapy given at standard doses. In this approach, the goal of growth factor therapy is to decrease the duration or severity of neutropenia and thereby decrease the incidence of serious infections. In studies done to date, the risk of infections has been estimated by measuring the incidence of febrile neutropenia. While much remains to be learned about this supportive care application, it is best characterized from both a clinical and a cost standpoint.
The second rationale is the facilitation of chemotherapy dose intensification, with or without the use of autologous hematopoietic progenitor cells. In this approach, the goal of growth factor therapy is to increase the dose intensity of the administered chemotherapy and thereby improve tumor outcomes. Because the efficacy of dose intensification in improving tumor outcomes is likely to vary with tumor type and stage of disease, and because the demonstration of improvement in survival requires large studies of long duration that have not yet been completed, this is the most difficult application of hematopoietic growth factors to fully analyze from a cost-benefit standpoint.
The third rationale is the treatment of established infection in patients who are neutropenic following myelosuppressive chemotherapy. In this approach, the goal of growth factor therapy is to decrease the duration of febrile neutropenia and, hence, to decrease the cost of caring for these patients. The appropriate clinical trial, and type and scope of cost analysis, depends upon the rationale and goals of myeloid growth factor therapy. Table 1 lists the three rationales for myeloid growth factor therapy and their implications for clinical trials and cost analyses.
G-CSF or GM-CSF as Supportive Care During Chemotherapy
The best characterized application of myeloid growth factors in oncology is their use to prevent febrile neutropenia during myelosuppressive chemotherapy. In the initial phase II studies of G-CSF and GM-CSF, patients received at least two cycles of myelosuppressive chemotherapy and were given the growth factor during one of the cycles; treated cycles were compared with untreated cycles, which served as the control [1,2]. In both studies, growth factor therapy was associated with a decrease in the duration of neutropenia, and G-CSF therapy was associated with a decreased incidence of hospitalization for febrile neutropenia. This observation led to the randomized, placebo-controlled, double-blind study of G-CSF for the prevention of febrile neutropenia in patients with small-cell lung cancer undergoing cyclophosphamide-based chemotherapy . In this study, patients were treated with placebo or G-CSF (approximately 5 mcg/kg), by daily subcutaneous injection beginning 24 hours after the last of three daily doses of chemotherapy and continuing until the absolute neutrophil count had nadired and returned to at least 10,000/mm³. Patients were removed from the double-blind trial and offered open-label G-CSF during subsequent chemotherapy cycles if they developed febrile neutropenia.
During the first chemotherapy cycle, when the study groups were fully comparable to each other and most representative of the universe of cancer patients, the incidence of hospitalization for febrile neutropenia was 50% in the placebo group and 24% for patients receiving G-CSF. During cycles of blinded treatment, the number of days of treatment with intravenous antibiotics, the number of days of hospitalization, and the incidence of confirmed infections were reduced by approximately 50% when G-CSF was given, compared with placebo. These effects were observed during all six chemotherapy cycles. G-CSF therapy was associated with a 20% incidence of bone pain; no toxicities that increased health care costs or limited the drug's use were reported.
These results were subsequently confirmed in a similarly designed European study in which patients remained on study if febrile neutropenia developed, and were managed on subsequent cycles with chemotherapy dose reductions . In this study, the incidence of febrile neutropenia was decreased by 50% in patients given G-CSF, even though these patients received significantly more of the planned chemotherapy dose. From these studies, it is reasonable to conclude that G-CSF therapy is associated with approximately a 50% reduction in the risk of febrile neutropenia, at least in populations of patients for whom this risk is relatively high without G-CSF.
A similar trial was subsequently conducted with Escherichia coli-derived GM-CSF . In this study, 290 patients with small-cell lung cancer undergoing cyclophosphamide-based chemotherapy were randomized to observation or to GM-CSF at a dose of 5, 10 or 20 mcg/kg/day by daily subcutaneous injection commencing 24 hours after the last dose of chemotherapy. GM-CSF therapy was associated with a decreased duration of neutropenia and, for patients receiving the 10 mcg/kg dose, decreased requirements for antibiotic therapy during cycle 1 (11% versus 25% in the observation group). However, fever was observed more frequently in the patients receiving 10 or 20 mcg/kg of GM-CSF than in patients on the observation arm; the incidence of fever in patients receiving the 5 mcg/kg dose was no different from the observation arm. GM-CSF therapy was associated with fever, edema, myalgias, bone pain, and skin rashes. There was no documented decrease in the incidence of febrile neutropenia or culture-confirmed infections associated with GM-CSF therapy.
A second randomized clinical trial has been carried out in 104 patients undergoing ifosfamide (Ifex) chemotherapy for germ-cell tumors . In this study, patients were randomized to receive GM-CSF during cycles 1 and 2 or 3 and 4. During cycle 1, GM-CSF therapy was associated with a significantly lower risk of "clinically relevant" infections (24% versus 45%) and febrile neutropenia (22% versus 43%). No significant effects of GM-CSF on infection or febrile neutropenia rates were observed during subsequent chemotherapy cycles, and GM-CSF therapy was discontinued in 14% of cycles because of toxicity.
It is reasonable to conclude that the impact of GM-CSF on the incidence of febrile neutropenia or significant infection during myelosuppressive chemotherapy remains unclear; therapy with this factor may be associated with a decreased infection risk, but this observation has not been made in later chemotherapy cycles or confirmed in all studies. It is possible that GM-CSF is more toxic than G-CSF in this population of patients, and that therapy-related fevers may be mistaken for serious infection, obscuring the therapeutic effects. If this factor can decrease infection risk in community oncology practice, the optimal schedule and dose remain to be defined, and the magnitude of the impact on febrile neutropenia risk is not clear from studies to date. Moreover, the relative toxicities of the E coli- and yeast-derived GM-CSF preparations require clarification, and, if the yeast-derived preparation is less toxic, it should be tested in the prevention of febrile neutropenia in randomized clinical trials.
G-CSF and GM-CSF in Chemotherapy Dose Intensification
The myeloid growth factors can also be used to facilitate chemotherapy dose intensification. In this approach to their integration into clinical oncology practice, the goal is not to decrease the toxicities or resources utilized during myelosuppressive chemotherapy but rather to improve the clinical outcomes of cancer treatment: cure rate and overall survival. We are still relatively early in the exploration of the potential of this strategy, and there are few or no data available regarding the success of this approach in achieving its goal in the management of common solid tumors. Because the field is evolving rapidly, it is neither possible nor particularly helpful to precisely define the economic outcomes associated with the current approaches to dose-intensive regimens. It is possible to conclude that these economic outcomes will vary depending upon patient characteristics such as tumor type and prior therapy, and that, in general, they will improve (become cheaper) with time, as the field evolves. Still, it is likely that these approaches will always consume more resources than standard-dose chemotherapy regimens, and their integration into general clinical practice and their cost-benefit analyses will depend upon a clearly demonstrated improvement in cancer outcomes.
Growth factors alone can be utilized to facilitate chemotherapy dose intensification; the data to date suggest that this approach can support approximately a twofold chemotherapy dose intensification before neutropenia, thrombocytopenia, and nonhematologic toxicities become unacceptable. It is not yet clear that this level of chemotherapy dose intensification will improve clinical outcomes in most settings, although the approach has promise and is under study in the adjuvant therapy of breast cancer. Much greater levels of chemotherapy dose intensification are possible utilizing both growth factors and autologous hematopoietic progenitor cells. While the impact of these approaches to the management of solid tumors on clinical outcomes is still being defined, the strategy, in the form of autologous bone marrow transplantation (ABMT), was in use prior to the introduction of the myeloid growth factors into clinical practice. Hence, it is possible to examine the impact of GM-CSF and G-CSF on clinical and economic outcomes in this setting, using ABMT without growth factors for comparison.
Both GM-CSF, given at a dose of 250 mcg/m²/day [7,8], and G-CSF, given at a dose of 10 mcg/kg/day [9,10], have been shown to accelerate neutrophil recovery and to decrease the duration of infections in patients undergoing ABMT. The magnitude of the benefit varies considerably in different studies, and most studies are historically controlled, adding to the difficulty of drawing firm and precise conclusions regarding the impact of these factors on length of hospital stay or resource consumption. In no study to date has the use of myeloid growth factors been associated with improvements in tumor response rates or survival, and it is appropriate to attempt to examine the economic outcomes of this application in cost-minimization studies. Although the reported impacts on the duration of hospitalization vary, for the initial cost-minimization estimates we will assume that therapy with either factor has the potential to decrease resources consumed in most settings by an amount equivalent to 1 week of hospitalization.
The introduction of the myeloid growth factors into clinical practice has been associated with the development of an alternative to autologous bone marrow as a source of hematopoietic progenitor cells. Therapy with GM-CSF or G-CSF, either alone or following a dose of chemotherapy, is associated with the appearance in the peripheral blood of progenitor cells that can be harvested via leukapheresis and utilized as hematopoietic support following high-dose chemotherapy. Peripheral blood progenitor cells (PBPCs) harvested during GM-CSF [11-14] or G-CSF [15-21] therapy have been used successfully as the sole source of autografting. While no randomized, controlled clinical trials have been published, the historically controlled studies support the conclusion that the use of PBPCs is associated with a shorter duration of thrombocytopenia and decreased platelet transfusion requirements, as well as possibly further shortening of the duration of neutropenia and decreases in the number of hospital days and blood transfusions, compared with ABMT followed by growth factor therapy.
Most protocols have utilized G-CSF and GM-CSF at doses of approximately 5 to 10 mcg/kg/day given for 7 days, with pheresis done for 1 to 3 days. Many of the GM-CSF studies have utilized chemotherapy in addition to growth factor for PBPC mobilization, and in this setting up to 2 weeks of growth factor therapy may be necessary before PBPC collection is complete. There are no randomized comparisons of the two factors, although one comparative study found that PBPC mobilized with G-CSF yielded the shortest durations of hospitalization .
It is an understatement to note that the optimal regimen for PBPC harvesting has not been defined. Combinations of growth factors, such as stem cell factor and G-CSF, and combinations of chemotherapy and growth factors both appear to produce higher yields of PBPCs than does monotherapy with a myeloid growth factor, but it is not clear that these higher numbers of PBPCs produce more rapid engraftment. Ongoing clinical trials are exploring this area, as well as the potential of PIXY 321, erythropoietin, and the combination of interleukin-3 (IL-3) and GM-CSF for the mobilization of PBPCs. It is likely that future studies will help to define optimized mobilization protocols that are less resource intensive , and to develop standards for an adequate PBPC harvest that can then be used to compare the cost efficiency of different mobilization protocols.
It has not yet been conclusively demonstrated that the administration of GM-CSF or G-CSF following high-dose chemotherapy with PBPC grafting is associated with the same clinical benefits observed after autologous marrow transplantation. This issue has obvious economic implications. If the administration of myeloid growth factors not only yields a superior hematopoietic progenitor cell product but also eliminates the need for growth factors following transplantation, their use in this setting is associated with excellent economic outcomes. The published preliminary data to date are conflicting on this important point [24,25].
Finally, it is becoming clear that myeloid growth factors may have a role in supporting engraftment following allogeneic bone marrow transplantation  and that it will be possible to safely substitute PBPCs for bone marrow in allografting . This is likely to be an area of rapid evolution in clinical care, with the potential for improved economic outcomes in the future.
G-CSF and GM-CSF for the Treatment of Febrile Neutropenia
Because the most efficient use of a growth factor will provide the best economic outcomes, it is tempting to focus their use in patients who have already developed febrile neutropenia. While this strategy eliminates unnecessary use in patients who will do well without growth factor therapy, it does not currently permit the prevention of hospitalizations. The goal of growth factor therapy in this setting is to shorten the duration of hospitalization and thereby to decrease the resources consumed in treatment for febrile neutropenia. However, it is not clear that this will be the most efficient approach to myeloid growth factor use for the problem of chemotherapy-associated febrile neutropenia. While endogenous growth factor production may be temporarily impaired by the administration of cytotoxic chemotherapy, by the time patients present with febrile neutropenia, serum G-CSF levels are usually elevated , suggesting that the optimal time for exogenous cytokine administration may have passed.
One randomized trial of a myeloid growth factor for the treatment of established febrile neutropenia during chemotherapy has been published . In this study, 218 patients with febrile neutropenia were randomized to receive antibiotics alone or antibiotics with G-CSF, 12 mcg/kg/day. The study was double blinded and placebo controlled. Patients remained on G-CSF until the absolute neutrophil count had recovered to 5,000/mm³ and they had been afebrile for 4 days. In this study, G-CSF therapy was associated with a significant decrease in the duration of febrile neutropenia (median, 5 vs 6 days, P less than .01) but was not associated with a significant decrease in the duration of hospitalization. The risk of prolonged hospitalization (more than 11 days) was reduced by 50% (relative risk, 2.1; 95% confidence interval, 1.1 to 4.1; P = .02). The investigators concluded that G-CSF may have some benefit in decreasing the duration of hospitalization in patients at risk for prolonged hospitalizations for febrile neutropenia. To date, no hematopoietic growth factor has been shown to reduce the duration of established febrile neutropenia in a prospective randomized trial.
Erythropoietin Therapy During Cancer Chemotherapy
Recombinant human erythropoietin (EPO, Epogen, Procrit) is now available for the treatment of anemia associated with cancer chemotherapy. The administration of EPO has been associated with a decreased incidence of clinically significant anemia in patients receiving cisplatin-based chemotherapy . In the initial phase I/II studies, EPO given in doses of 50 to 200 U/kg thrice weekly was effective in the treatment of the anemia associated with cisplatin-based chemotherapy [31,32]. This led to a randomized, double-blind, placebo-controlled clinical trial including 132 anemic cancer patients undergoing cyclic cisplatin-based chemotherapy. In this study, the administration of EPO, 150 U/kg thrice weekly for 12 weeks, was associated with an increased hemoglobin concentration, energy level, functional status, and overall quality of life . The data also showed a trend suggesting decreased transfusion requirements in patients receiving EPO (1.2 vs. 2.0 units per patient per month during months 2 and 3, P = .089). When the data from another randomized study in patients receiving non-cisplatin-containing chemotherapy were pooled with these data, patients receiving EPO were found to have received significantly fewer red cell transfusions (1.04 vs 1.81 units per patient per month, P = .009) [33,34].
In another randomized and placebo-controlled trial, the administration of EPO given at a dose of 100 U/kg thrice weekly for 9 weeks to cancer patients with cisplatin-related anemia was associated with increased hemoglobin levels and a sixfold decrease in the mean number of transfusions required (0.3 vs 1.8 units per patient over 9 weeks, P = .01) . The efficacy of EPO therapy in improving hemoglobin levels, decreasing transfusion requirements, and increasing energy level and overall quality of life scores in anemic patients undergoing myelosuppressive chemotherapy in community-based oncology practice has recently been confirmed in a large open-label clinical trial . EPO has not been shown to be effective in decreasing transfusion requirements or accelerating erythropoietic recovery following autologous transplantation . In no study to date has EPO therapy been associated with improved cancer treatment outcomes, such as response rate or survival. The first step toward a cost-benefit analysis is therefore a cost-minimization study.
Much remains to be learned about the impacts of therapy with the myeloid growth factors on economic outcomes in community oncology practice. The first generation of studies has been cost-minimization studies based on hospital charge data. These studies provide a first approximation of cost effects and will need to be updated as new approaches to supportive care in oncology and refined schedules for growth factor use evolve, and as data from more sophisticated approaches to cost studies utilizing resource-based cost estimation techniques become available.
The Prevention of Febrile Neutropenia
A cost-minimization study was done in parallel with the randomized clinical trial of G-CSF in the United States . In this study, hospital and physician charge data were collected for all hospitalizations for febrile neutropenia of patients on the study. These data were used to determine the mean lengths of stay and charges associated with these hospitalizations. The costs of G-CSF were estimated based on the average wholesale price of the drug, the mean duration of G-CSF therapy in the clinical trial, and the weight distribution of cancer patients in the clinical trial. The cost offsets were calculated using charge data, as well as charge data converted to cost data using the cost-to-charge ratios for each participating hospital. These data were used to estimate the impact of G-CSF on the cost outcomes of a single chemotherapy cycle in the average patient.
In this study, the mean duration of hospitalizations for febrile neutropenia in the placebo group was 7.3 days, and mean charges for hospital and physician services were $8,469. These measured data on hospital charges in the placebo group were similar to the results gathered at the time from the Medicare database (Table 2), suggesting that the placebo group was representative, from a resource utilization standpoint, of febrile neutropenia occurring in the community practice of oncology. Hospitalizations for patients receiving G-CSF were significantly shorter and less expensive than those of patients receiving placebo (mean duration, 3.6 days vs 7.3 days; mean charges for physician and hospital services, $4,258 vs $8,469; P less than .01). G-CSF therapy was therefore associated with two potential sources of cost savings: a decreased risk of hospitalization and approximately a 50% decrease in the cost and duration of those hospitalizations that still occurred despite G-CSF therapy. In the context of a clinical trial in which a relatively high rate of febrile neutropenia was measured in the placebo group, the use of G-CSF was estimated to be cost saving when the analysis was based on charges; that is, the cost of acquiring and administering the drug was less than the charges associated with the complications it prevented. The use of G-CSF was close to fully offsetting the cost of administering the drug when the analysis was based on costs (charges × cost-to-charge ratios).
A sensitivity analysis examined the effects of key efficiency variables on the cost offsets associated with G-CSF use (Figure 2). The first efficiency variable examined was the risk of febrile neutropenia without G-CSF use; these analyses suggested that the threshold value for this efficiency variable was approximately 35% in the charge model and 55% in the cost model. As noted, randomized trials designed to document efficacy do not usually define the most cost-effective approach to a drug's use. Because refinements in G-CSF scheduling may decrease the number of doses administered per patient per cycle of chemotherapy by two or three doses per cycle without compromising clinical benefits , the effects of the duration of G-CSF therapy per chemotherapy cycle were examined as a second efficiency variable. These data suggested that decreasing the duration of G-CSF therapy by 2 days would substantially improve the cost offsets.
Finally, because the optimal dosing of G-CSF in the prevention of febrile neutropenia remains to be determined, the effects of patient weight (and G-CSF dose) were examined. Because the drug is supplied in single-dose vials, therapy with G-CSF was found to be associated with increased cost savings in patients weighing 60 kg or less, that is, those patients who are treated with the smaller G-CSF vial. It is possible that the efficacy of G-CSF in the prevention of febrile neutropenia can be achieved with slightly lower doses, and that all adults can be treated with the smaller vial, making the drug more cost saving. The issue of G-CSF dosing is very important from a cost outcomes standpoint and should be addressed in future clinical trials.
Remarkably similar conclusions regarding the risks of febrile neutropenia at which G-CSF therapy becomes cost saving were reached in an economic modeling study that was not based on directly measured charge or resource consumption data ; this interesting approach is further elucidated in Dr. Lyman's article in this supplement.
We still do not have data regarding the impact of G-CSF on the risk of febrile neutropenia in clinical situations in which that risk is relatively low. We also do not have confirmation in a larger data set of the effects of G-CSF on the duration and resource consumption of hospitalizations for febrile neutropenia that still occur despite the drug's use. Moreover, the use of charge data to derive cost impact estimates is complicated by the inconsistent relationships between individual charges and either actual costs or resources consumed. It is hoped that further cost-minimization studies will be done in this area, and that they will involve detailed measurement of the resources consumed in G-CSF treatment and in the management of febrile neutropenia.
Assuming that G-CSF therapy is always associated with a 50% reduction in the risk of febrile neutropenia regardless of the risk without the drug, it is possible to express predicted cost impacts of this drug in a more resource-based fashion. While these analyses cannot substitute for more rigorous studies measuring the resources consumed in the acquisition of G-CSF and the management of febrile neutropenia, they can be instructive. In these analyses, the cost offsets associated with G-CSF use are expressed in terms of the number of doses of G-CSF required by the whole population at risk to prevent 1 day of hospitalization for febrile neutropenia .
Figure 3 shows the results of such an analysis, done assuming that G-CSF therapy is associated with a 50% reduction in both the incidence of hospitalization for febrile neutropenia and the duration of those hospitalizations that still occur despite G-CSF therapy (as observed in the one published cost study). This example explores the effects of variations in the number of doses of G-CSF utilized per patient per cycle. The break even or threshold point will depend upon the dose of G-CSF given to each patient, the relative resource consumption per hospital bed for an individual institution, and the price paid for the G-CSF, but it is likely that in most health care systems this will occur when between 5 and 10 doses of G-CSF are being given to prevent one patient day of hospitalization. These values are represented by lines; the points at which the curves intersect these lines can be mapped to the threshold risks of febrile neutropenia at which the administration of G-CSF becomes cost saving. This analysis supports the importance of optimized G-CSF scheduling to the cost-effective use of this drug; small changes in the number of doses of G-CSF utilized per chemotherapy cycle produce large changes in the threshold risk of febrile neutropenia.
The curves generated in this type of analysis have an interesting feature. As the risk of febrile neutropenia without G-CSF falls below 20%, the number of doses of G-CSF required to offset one hospital day begins to rise rapidly. In this range of the curve, small errors in febrile neutropenia risk assessment have large implications in terms of cost offsets. Contrariwise, at risks of febrile neutropenia greater than 30%, changes in this efficiency variable have much less effect on cost offsets. This serves to emphasize the need for better techniques for accurately estimating the risk of febrile neutropenia in patients undergoing myelosuppressive chemotherapy, especially when that risk may be low.
Because we have fewer data confirming the effects of G-CSF on the duration of hospitalization than we do regarding its effects on the risk of this complication, it is reasonable to explore the effects of this hospitalization duration variable on the cost offsets associated with G-CSF use. Such an analysis is shown in Figure 4. Clearly, if future studies demonstrate that the initiation of G-CSF therapy prior to the onset of neutropenia has no effect on the duration of hospitalization for febrile neutropenia when it does occur, the threshold values of febrile neutropenia risk will be higher. Further studies of this variable will be important to our understanding of the cost impacts associated with myeloid growth factor therapy in this setting.
There are no published data regarding the impact of GM-CSF on economic outcomes when used for the prevention of febrile neutropenia during myelosuppressive chemotherapy. As noted above, randomized trials published to date have yielded conflicting data regarding the impact of GM-CSF therapy on the risk of hospitalization with febrile neutropenia during myelosuppressive chemotherapy. Moreover, the one study that found an effect of GM-CSF therapy on the risk of hospitalization detected it only during the first cycle of chemotherapy. As currently used in community practice, myeloid growth factors are most often administered during the second or later chemotherapy cycles, after an episode of profound or prolonged neutropenia or febrile neutropenia during an earlier cycle. Any meaningful estimation of the economic effects of a factor used in the community setting will depend upon its demonstrated ability to decrease the risk of febrile neutropenia over many cycles of chemotherapy.
Clearly, more clinical data are needed to support meaningful cost-minimization studies with respect to GM-CSF in this setting. These studies should include a better characterization of the relative toxicities of the E coli- and yeast-derived preparations, as well as a clear demonstration of the impact of therapy with this molecule on the risk of febrile neutropenia and any costs associated with managing its toxicities. Utilizing the data from the G-CSF studies regarding the duration and costs of febrile neutropenia, it is possible to draw some preliminary conclusions regarding the predicted cost impacts of myeloid growth factors for which sufficient clinical and economic data are not yet available.
Figure 5 shows the resource-based cost offsets for a hypothetical growth factor, with curves drawn assuming factors with varying degrees of efficacy in the prevention of febrile neutropenia. The data reviewed suggest that the 50% curve represents the predicted offsets for G-CSF. Based on these curves, one would predict that the cost-minimization characteristics and the threshold points of a myeloid growth factor used to prevent febrile neutropenia will depend mainly upon the degree of the factor's efficacy. It is likely that, in addition to the GM-CSF preparations, other myeloid growth factors, such as PIXY 321, and combinations of growth factors, such as IL-6 and G-CSF, will be tested in the prevention of febrile neutropenia. It will therefore be important to carefully characterize the degree of efficacy of each factor in randomized, controlled trials, because it is likely that this variable, more than the unit cost of a dose of the growth factor, will determine its economic outcome profile relative to G-CSF.
The Treatment of Established Febrile Neutropenia
As noted above, there are relatively few data available concerning the impact of therapy with myeloid growth factors initiated at the time of onset of febrile neutropenia on the duration of hospitalization for that complication. There are no published directly measured hospital charge or resource consumption data to support assessments of economic outcomes. The need for more clinical and economic outcomes data is obvious. However, it is possible to draw some preliminary conclusions from the one published G-CSF study . As noted, in that study, the effect of G-CSF on the mean duration of hospitalization for the entire group did not achieve statistical significance (G-CSF, 8.7 days; placebo, 10 days; P = .09). If we assume that the observed decrease in mean length of stay is actually representative of the effects of therapy with this factor in this setting, then 6.7 of the 12 mcg/kg/day doses of G-CSF would be required to prevent one hospital day. When, for the sake of comparison, these doses are converted to the 5 mcg/kg/day doses used in the prevention trials, 16 doses are required to prevent one hospital day. The data to date support the conclusion that the use of myeloid growth factors to treat patients with established febrile neutropenia is not associated with as favorable an economic outcome as is the use of G-CSF to prevent febrile neutropenia in the high-risk patient. Using decision analysis modeling, Dr. Lyman has independently come to very similar conclusions .
Chemotherapy Dose Intensification and Autologous Progenitor Cell Rescue
There are no charge or resource consumption data regarding the economic outcomes associated with the use of myeloid growth factors alone to facilitate chemotherapy dose intensification. It seems obvious that costs will increase, due to higher chemotherapy costs and more serious nonhematopoietic toxicity, without significant cost offsets. If the chemotherapy is of sufficient dose intensity, it is possible that the overall cost of care would be lowered by the use of autologous progenitor cells, even if the therapy is not marrow ablative .
While much remains to be learned, it is possible to comment on the relative economic outcomes of traditional ABMT alone compared with ABMT followed by myeloid growth factors or with the use of PBPCs to support high-dose chemotherapy. A study measuring the total charges associated with autologous transplantation in patients with lymphoid malignancies found a decrease in mean total charges after marrow infusion of $22,700 ($39,800 with growth factor vs $62,500 without; P = .005). Although these data have the limitations of a charge-based study restricted to the post-transplant segment of the hospitalization and did not include the cost of the GM-CSF, the difference in charges is great enough that it is very likely that GM-CSF administered following ABMT is associated with overall cost savings and very favorable economic outcomes. Although the clinical data for G-CSF administered after ABMT are similar to those for GM-CSF, directly measured charge or resource consumption data have not been published.
The relative costs associated with the use of autologous marrow or PBPCs as the source of cellular hematopoietic support have not been studied sufficiently to permit firm conclusions regarding the economic outcomes associated with these two approaches. Two historically controlled studies have measured the costs associated with the harvesting and transplantation of marrow or PBPCs mobilized by G-CSF: in patients with a variety of tumors in the Netherlands  and in patients with breast cancer in the United States . The results of these studies were remarkably similar: PBPCs were associated with approximately a 30% decrease in charges relative to marrow, even when the costs of harvesting were included in the analysis. Other studies have confirmed that G-CSF-mobilized PBPCs appear to be associated with overall cost savings, compared with autologous marrow, although the cost of harvesting, which may or may not be higher with PBPCs, has not always been included in the analysis .
The economic outcomes associated with the substitution of PBPCs for autologous marrow for the support of high-dose chemotherapy are likely to vary depending upon the mobilization regimen employed, patient characteristics such as tumor type and prior therapy, and the specific induction regimen used. If sufficient numbers of PBPCs can be procured and similar cancer outcomes achieved using growth factors alone for PBPC mobilization, there should be cost savings associated with the elimination of the febrile neutropenia that can occur during the mobilization chemotherapy and the more efficient use of resources that should come with increased predictability of pheresis bed utilization. Economic outcomes associated with PBPCs will also be improved in situations in which myelosuppression rather than mucositis is the critical factor in determining resource consumption, clinical recovery, and hospital discharge.
For comparison purposes, it is possible to make rough and very preliminary estimates of the number of doses of growth factor required to offset one day of hospitalization. If G-CSF (10 mcg/kg/day) and GM-CSF (250 mcg/m²/day) decrease the duration of hospitalization following ABMT by 1 week (4 vs 5 weeks), then approximately eight 5 mcg/kg doses of G-CSF and six 5 mcg/kg doses of GM-CSF are required to offset one hospital day. If G-CSF or GM-CSF can be given for 7 days at doses of approximately 10 mcg/kg/day and yield sufficient PBPCs for an allograft, and if the use of these PBPCs and the growth factor after transplant at the same dose is associated with a further decrease in resource consumption equivalent to the elimination of another week of hospitalization (3 vs 5 weeks), then two of these doses, or four of the 5 mcg/kg doses, will prevent one day of hospitalization, compared with autologous marrow without growth factor.
While these estimates are certainly not precise, they probably reflect the magnitude of the efficiency of the use of myeloid growth factors in this setting relative to their economic outcomes in other settings in oncology (Figure 6). If high-dose chemotherapy is to be given, economic outcomes are improved by the use of growth factors and PBPCs. The efficacy of this high-dose chemotherapy in improving survival is still under study for most solid tumors. Obviously, if the approach is shown to be ineffective in a given clinical setting, the best economic outcomes would be realized by discontinuing the dose intensification approach.
Toward Cost-Benefit Analyses
The available cost-minimization data regarding the use of myeloid growth factors in oncology practice have been summarized. Utilizing these data, the clinician will be able to begin to discern the clinical circumstances under which the use of these factors is likely to be cost saving. Little work has been done on the positive impacts of these factors utilized in settings in which they are not fully cost saving, and no data exist to support rational decision making. Cost-minimization studies can only provide estimates of the actual increase in costs associated with factors used in these settings. Some preliminary data regarding the impact of G-CSF on the quality of life in patients undergoing myelosuppressive chemotherapy  and the impact of GM-CSF on the quality of life in patients undergoing bone marrow transplantation  have been published, but there is clearly a need for further and more rigorous studies of the impact of these factors on quality of life and indirect non-health-care costs such as lost productivity of patients and their families.
The true costs associated with red blood cell transfusions in the cancer patient have not been defined, and no studies of hospital charges or resource consumption during randomized, controlled clinical trials of EPO administered during chemotherapy have been published. Hence, it is not possible to draw definitive conclusions regarding cost minimization associated with the use of EPO in this setting. The published wholesale price of the EPO marketed for the treatment of cancer patients is approximately $0.01 per unit. If patients are treated with 10,000 units three times weekly for 12 weeks, the cost of drug acquisition will be approximately $3,600.
The published clinical data suggest that when EPO is used to treat anemic cancer patients undergoing chemotherapy, it will prevent, at most, an average of approximately 2 units of red cell transfusions per patient over the 12 weeks of treatment. If the only costs saved are those related to transfusions, it is unlikely that EPO therapy in this setting is fully cost saving. Appropriate cost-minimization studies have not been done to determine the true costs (acquisition cost minus cost offsets in prevented transfusions) for use of EPO in this setting. These costs, when known, will be balanced against the demonstrated improvements in quality of life  and any additional benefits, such as increased productivity and decreased lost time from work, that are found in future studies.
It is important to note that EPO therapy in cancer patients is early in its evolution, and it is likely that new doses and schedules will be developed that improve economic outcomes. One critical variable is the selection of patients who are likely to respond to a given dose of EPO. In the clinical trials published to date, approximately 50% of treated patients meet criteria for EPO response, and, in general, only responders benefit in terms of improved quality of life and decreased transfusion requirements. If it were possible to select responders prior to or early in EPO treatment, the cost of therapy to the whole population could be halved without changing overall benefits. This would double the cost benefit and might even render the drug fully cost saving. Pretreatment serum erythropoietin levels have not been sufficient to select responders, and work on predictors of response continues.
Having summarized the state of our understanding of the economic outcomes of hematopoietic growth factors in oncology, it is possible to develop a list of 10 important issues that require elucidation in future clinical trials, the results of which will then aid in refining our use of the hematopoietic growth factors in oncology and in optimizing economic outcomes.
1. The impact of myeloid growth factors on the risk of febrile neutropenia in populations of patients at relatively low risk for this complication (risk less than 20%).
2. The minimal effective dose and optimal schedule of G-CSF in the prevention of febrile neutropenia.
3. The impact of the two GM-CSF preparations (yeast and E coli derived) on infection risks during standard-dose myelosuppressive chemotherapy.
4. The clinical efficacy and cost outcomes of lower doses of G-CSF (5 mcg/kg/day) and all doses of GM-CSF in the treatment of patients with established febrile neutropenia.
5. The impact of GM-CSF and G-CSF administration following PBPC transplant on engraftment, length of stay, and resource consumption.
6. The most cost-effective approaches to PBPC harvesting.
7. The relative costs and resources consumed with PBPC and autologous marrow harvesting.
8. The impact of therapy with G-CSF and GM-CSF on the quality of life and indirect economic losses of patients and their families, especially in settings in which these factors are not fully cost saving and add to overall costs.
9. The optimal dose and schedule of EPO administered during chemotherapy, and techniques for selecting patients who are likely to respond to those schedules and doses.
10. The true cost associated with EPO therapy and of red blood cell transfusions in anemic cancer patients undergoing chemotherapy.
1. Antman KS, Griffin JD, Elias A, et al: Effect of recombinant human granulocyte-macrophage colony-stimulating factor on chemotherapy-induced myelosuppression. N Engl J Med 319:593-598, 1988.
2. Gabrilove JL, Jakubowski A, Scher H, et al: Effect of granulocyte colony-stimulating factor on neutropenia and associated morbidity due to chemotherapy for transitional-cell carcinoma of the urothelium. N Engl J Med 318:1414-1422, 1988.
3. Crawford J, Ozer H, Stoller R, et al: Reduction by granulocyte colony-stimulating factor of fever and neutropenia induced by chemotherapy in patients with small-cell lung cancer. N Engl J Med 325:164-170, 1991.
4. Trillet-Lenoir V, Green J, Manegold C, et al: Recombinant granulocyte colony stimulating factor reduces the infectious complications of cytotoxic chemotherapy. Eur J Cancer 29A:319-324, 1993.
5. Hamm J, Schiller JH, Cuffie C, et al: Dose-ranging study of recombinant human granulocyte-macrophage colony-stimulating factor in small-cell lung carcinoma. J Clin Oncol 12:2667-2676, 1994.
6. Bajorin DF, Nichols CR, Schmoll HJ, et al: Recombinant human granulocyte-macrophage colony-stimulating factor as an adjunct to conventional-dose ifosfamide-based chemotherapy for patients with advanced or relapsed germ cell tumors: A randomized trial. J Clin Oncol 13:79-86, 1995.
7. Nemunaitis J, Rabinowe SN, Singer JW, et al: Recombinant granulocyte-macrophage colony-stimulating factor after autologous bone marrow transplantation for lymphoid cancer. N Engl J Med 324:1773-1778, 1991.
8. O'Day SJ, Rabinowe SN, Neuberg D, et al: A phase II study of continuous infusion recombinant human granulocyte-macrophage colony-stimulating factor as an adjunct to autologous bone marrow transplantation for patients with non-Hodgkin's lymphoma in first remission. Blood 83:2707-2714, 1994.
9. Sheridan WP, Morstyn G, Wolf M, et al: Granulocyte colony-stimulating factor and neutrophil recovery after high-dose chemotherapy and autologous bone marrow transplantation. Lancet 2:891-895, 1989.
10. Stahel RA, Jost LM, Cerny T, et al: Randomized study of recombinant human granulocyte colony-stimulating factor after high-dose chemotherapy and autologous bone marrow transplantation for high-risk lymphoid malignancies. J Clin Oncol 12:1931-1938, 1994.
11. Elias AD, Ayash L, Anderson KC, et al: Mobilization of peripheral blood progenitor cells by chemotherapy and granulocyte-macrophage colony-stimulating factor for hematologic support after high-dose intensification for breast cancer. Blood 79:3036-3044, 1992.
12. Tepler I, Cannistra SA, Frei ED, et al: Use of peripheral-blood progenitor cells abrogates the myelotoxicity of repetitive outpatient high-dose carboplatin and cyclophosphamide chemotherapy. J Clin Oncol 11:1583-1591, 1993.
13. Tricot G, Jagannath S, Vesole D, et al: Peripheral blood stem cell transplants for multiple myeloma: Identification of favorable variables for rapid engraftment in 225 patients. Blood 85:588-596, 1995.
14. Bishop MR, Anderson JR, Jackson JD, et al: High-dose therapy and peripheral blood
progenitor cell transplantation: Effects of recombinant human granulocyte-macrophage colony-stimulating factor on the autograft. Blood 83:610-616, 1994.
15. Sheridan WP, Begley CG, Juttner CA, et al: Effect of peripheral-blood progenitor cells mobilised by filgrastim (G-CSF) on platelet recovery after high-dose chemotherapy. Lancet 339:640-644, 1992.
16. Chao NJ, Schriber JR, Grimes K, et al: Granulocyte colony-stimulating factor "mobilized" peripheral blood progenitor cells accelerate granulocyte and platelet recovery after high-dose chemotherapy. Blood 81:2031-2035, 1993.
17. Bensinger WI, Longin K, Appelbaum F, et al: Peripheral blood stem cells (PBSCs) collected after recombinant granulocyte colony stimulating factor (rhG-CSF): An analysis of factors correlating with the tempo of engraftment after transplantation. Br J Haematol 87:825-831, 1994.
18. Nademanee A, Sniecinski I, Schmidt GM, et al: High-dose therapy followed by autologous peripheral-blood stem-cell transplantation for patients with Hodgkin's disease and non-Hodgkin's lymphoma using unprimed and granulocyte colony-stimulating factor-mobilized peripheral-blood stem cells. J Clin Oncol 12:2176-2186, 1994.
19. Glaspy J, Chap L, Waisman J, et al: High dose chemotherapy with thiotepa, mitoxantrone and cyclophosphamide (TMC) with autologous progenitor cell support in the treatment of breast cancer. Adv Oncol (in press), 1995.
20. Haas R, Moos M, Karcher A, et al: Sequential high-dose therapy with peripheral-blood progenitor-cell support in low-grade NHL. J Clin Oncol 12:1685-1692, 1994.
21. Kanold J, Rapatel C, Berger M, et al: Use of G-CSF alone to mobilize peripheral blood stem cells for collection from children. Br J Haematol 88:633-635, 1994.
22. Peters WP, Rosner G, Ross M, et al: Comparative effects of granulocyte-macrophage colony-stimulating factor (GM-CSF) and granulocyte colony-stimulating factor (G-CSF) on priming peripheral blood progenitor cells for use with autologous bone marrow after high-dose chemotherapy. Blood 81:1709-1719, 1993.
23. Jones HM, Jones SA, Watts MJ, et al: Development of a simplified single-apheresis approach for peripheral-blood progenitor-cell transplantation in previously treated patients with lymphoma. J Clin Oncol 12:1693-1702, 1994.
24. Dunlop DJ, Fitzsimons EJ, McMurray A, et al: Filgrastim fails to improve haemopoietic reconstitution following myeloablative chemotherapy and peripheral blood stem cell rescue. Br J Cancer 70:943-945, 1994.
25. Spitzer G, Adkins DR, Spencer V, et al: Randomized study of growth factors post-peripheral-blood stem-cell transplant: Neutrophil recovery is improved with modest clinical benefit. J Clin Oncol 12:661-670, 1994.
26. Schriber JR, Chao NJ, Long GD, et al: Granulocyte colony-stimulating factor after allogeneic bone marrow transplantation. Blood 84:1680-1684, 1994.
27. Dreger P, Haferlach T, Eckstein V, et al: G-CSF-mobilized peripheral blood progenitor cells for allogeneic transplantation: Safety, kinetics of mobilization, and composition of the graft. Br J Haematol 87:609-613, 1994.
28. Cebon J, Layton JE, Maher D, et al: Endogenous haemopoietic growth factors in neutropenia and infection. Br J Haematol 86:265-274, 1994.
29. Maher DW, Lieschke GJ, Green M, et al: Filgrastim in patients with chemotherapy-induced febrile neutropenia: A double-blind, placebo-controlled trial. Ann Intern Med 121:492-501, 1994.
30. Gamucci T, Thorel MF, Frasca AM, et al: Erythropoietin for the prevention of anaemia in neoplastic patients treated with cisplatin. Eur J Cancer 29A(suppl 2):S13-S14, 1993.
31. Cascinu S, Fedeli A, Fedeli SL, et al: Cisplatin-associated anaemia treated with subcutaneous erythropoietin: A pilot study. Br J Cancer 67:156-158, 1993.
32. Miller CB, Platanias LC, Mills SR, et al: Phase I-II trial of erythropoietin in the treatment of cisplatin-associated anemia. J Natl Cancer Inst 84:98-103, 1992.
33. Henry DH, Abels RI: Recombinant human erythropoietin in the treatment of cancer and chemotherapy-induced anemia: Results of double-blind and open-label follow-up studies. Semin Oncol 21:21-28, 1994.
34. Case DC Jr., Bukowski RM, Carey RW, et al: Recombinant human erythropoietin therapy for anemic cancer patients on combination chemotherapy. J Natl Cancer Inst 85:801-806, 1993.
35. Cascinu S, Fedeli A, Del Ferro E, et al: Recombinant human erythropoietin treatment in cisplatin-associated anemia: A randomized, double-blind trial with placebo. J Clin Oncol 12:1058-1062, 1994.
36. Bukowski R, Glaspy J, Steinberg D, et al: Phase IV clinical evaluation of recombinant human erythropoietin (r- HuEPO) in anemic cancer patients receiving chemotherapy (abstract). Proc Am Soc Clin Oncol 13:A1594, 1994.
37. Chao NJ, Schriber JR, Long GD, et al: A randomized study of erythropoietin and granulocyte colony-stimulating factor (G-CSF) versus placebo and G-CSF for patients with Hodgkin's and non-Hodgkin's lymphoma undergoing autologous bone marrow transplantation. Blood 83:2823-2828, 1994.
38. Glaspy JA, Bleecker G, Crawford J, et al: The impact of therapy with filgrastim (recombinant granulocyte colony-stimulating factor) on the health care costs associated with cancer chemotherapy. Eur J Cancer 29A(suppl 7): S23-S30, 1993.
39. Crawford J, Kreisman H, Garewal H, et al: A pharmacodynamic investigation of recombinant human granulocyte colony-stimulating factor (R-MetHuG-CSF) schedule variation in patients with small-cell lung cancer (SCLC) given CAE chemotherapy (abstract). Proc Am Soc Clin Oncol 11:A1005, 1992.
40. Lyman GH, Lyman CG, Sanderson RA, et al: Decision analysis of hematopoietic growth factor use in patients receiving cancer chemotherapy. J Natl Cancer Inst 85:488-493, 1993.
41. Glaspy J: Economic effect of myeloid growth factors on cancer treatment. Clin Immunother 2:192-205, 1994
42. Uyl-de Groot CA, Ossenkoppele GJ, van Riet AA, et al: The costs of peripheral blood progenitor cell reinfusion mobilised by granulocyte colony-stimulating factor following high dose melphalan as compared with conventional therapy in multiple myeloma. Eur J Cancer 30A:457-459, 1994.
43. Uyl-de Groot CA, Richel DJ, Rutten FF: Peripheral blood progenitor cell transplantation mobilised by r-metHuG-CSF (filgrastim): A less costly alternative to autologous bone marrow transplantation. Eur J Cancer 30A:1631-1635, 1994.
44. Crawford J, Glaspy J, Vincent M, et al: Effect of filgrastim (r-metHuG-CSF) on oral mucositis in patients with small cell lung cancer (SCLC) receiving chemotherapy (cyclophosphamide, doxorubicin and etoposide, CAE) (abstract). Proc Am Soc Clin Oncol 13:A1523, 1994.
45. Gordon B, Spadinger A, Hodges E, et al: Effect of granulocyte-macrophage colony-stimulating factor on oral mucositis after hematopoietic stem-cell transplantation. J Clin Oncol 12:1917-1922, 1994.