Human granulocyte-macrophage colony-stimulating factor modulates in vitro growth in only a minority of continuous human tumour cell lines

Granulocyte-macrophage colony stimulating factor (GM-CSF) has potential usefulness in a range of clinical conditions, including the treatment of patients with myelosuppression induced by chemotherapy and/or radiotherapy. Prior to any extensive use of this material, however, assessment of its effects on non-haematopoietic tumour cell growth appeared warranted. Accordingly, five laboratories, all members of the EORTC Clonogenic Assay Screening Study Group, have monitored in vitro responses to GM-CSF, using their own individual assay procedures, in a series of 18 human tumour cell lines, predominantly of non-haematopoietic origin, 25 tumour biopsy specimens and samples from five normal bone marrow aspirates. Significant growth stimulation by GM-CSF addition was rare, being absent in all 25 “fresh” ovarian tumour samples tested, but was consistently observed in four of the 18 continuous tumour cell lines tested (1 breast and 3 ovary) and all five normal bone marrow aspirates.     

Effects of granulocyte-macrophage colony-stimulating factor on in vitro growth of human solid tumors

Solid tumor biopsies from 33 patients were tested in vitro to evaluate the growth modulatory effects of recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF). In 29 of 33 studies (88%), addition of GM-CSF either had no effect on in vitro growth, or induced growth inhibition. While significant growth inhibition was observed in 10 studies, marked inhibition was only observed in three studies. However, all dose-response curves were usually flat, suggesting indirect effects. Moderate growth stimulation was observed in four instances, which may have been due to residual granulocyte-macrophage progenitors within the biopsies. We conclude that GM-CSF has little or no growth-modulatory effect on most nonhematopoietic neoplasms. The primary role of GM-CSF in patients with solid tumors appears to be in prevention or reversal of myelosuppression associated with therapy. Thus, while GM-CSF seems unlikely to have a role in monotherapy of cancer, it is also unlikely to have its utility compromised by enhancement of tumor growth.     

Biosynthetic granulocyte-macrophage colony-stimulating factor enhances neutrophil cytotoxicity toward human leukemia cells

Purified biosynthetic (recombinant) human granulocyte-macrophage colony-stimulating factor (GM-CSF) enhances antibody-dependent cell-mediated cytotoxicity (ADCC) of human neutrophils toward human promyelocytic leukemia cells (HL-60), B-lymphoma cells, and human T-leukemia virus II-infected human B-lymphoblastoid cells. The stimulation of antibody-dependent cell-mediated cytotoxicity is rapid (less than an hour), occurs at picomolar concentrations of GM-CSF, and does not require the presence of GM-CSF during the killing reaction. Therefore, neutrophils may be targeted toward tumor cells by antibody and their tumoricidal activity enhanced by GM-CSF in vitro. These results suggest that GM-CSF may have therapeutic utility in cancer therapy by increasing the number and activity of effector cells directed toward tumors by receptors to the immunoglobulin Fc fragment.     

Human colony-stimulating factor-producing lung cancer tissue releases a differentiation-inducing factor for human leukemic cells

Human lung cancer that induced marked granulocytosis in both the patient and tumor-transplanted nude mice (G2 mice) and from which conditioned medium (G2-T-CM) exhibited human and mouse active colony-stimulating activity (CSA) has been reported (K. Ikeda et al. Cancer Res 1985; 45:4144-4249). Recently, we found differentiation-inducing activity (DIA) in G2-T-CM, which differentiated human promyelocytic leukemic cells (HL-60) to macrophage-like cells. Differentiated HL-60 cells were considered to be mature macrophages as judged by the positivity of butyrate esterase activity, the acquisition of Fc receptor, and the increment in capacity of phagocytosis and nitroblue tetrazolium reduction. The DIA in G2-T-CM was not attributed to interferons known to have DIA, because interferon activity was not found in G2-T-CM by bioassay (less than 4 U/ml) and by radioimmunoassay for gamma-IFN (less than 0.1 U/ml). Molecular weight of DIA was 36,000 Da and separated from CSA of which molecular weight was 22,000 Da by gel filtration on Sephadex G-150. DIA and CSA were also separated on chromatofocusing chromatography, because isoelectric point of DIA was mainly less than 4.0 and that of CSA was 4.3-5.7. This DIA was stable after heat treatment (56 degrees C for 30 min or 100 degrees C for 10 min) and in acidic condition (pH 2.0 for 24 hr). G2-T-CM is a good source of differentiation-inducing factor for further purification and molecular cloning.     

Effects of granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor on lung cancer: roles of cyclooxygenase-2

We examined the effects of granulocyte colony-stimulating factor (G-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) on the lung cancer cell lines PC-9, LA-1 and A549. In addition, we examined if the effects of the cytokines on the cell lines are mediated by activation of cyclooxygenase (COX)-2. The three cell lines did not constitutively produce either G-CSF or GM-CSF. G-CSF did not influence cell growth in the three cell lines, while GM-CSF increased cell growth in the A549 and LA-1 lines. G-CSF and GM-CSF dose-dependently decreased cell death in the three cell lines. RT-PCR demonstrated GM-CSF receptor expression in the three lung cancer cell lines, whereas the G-CSF receptor exists only in the PC-9 line. We suggest that G-CSF might rescue the tumor cells from cytotoxicity due to serum deprivation through cellular pathways independent of the G-CSF receptor. G-CSF and GM-CSF increased cyclooxygenase-2 (COX-2) expression in PC-9 and LA-1 cells whereas they decreased COX-2 expression in A549 cells. The COX-2 inhibitor NS-398 increased cell death in PC-9 and LA-1 cells, whereas it decreased cell death in A549 cells. PC-9 and LA-1 clones transfected with sense G-CSF- or GM-CSF showed an increase in COX-2 expression, while COX-2 expression was decreased in transfected A549 clones. COX-2 expression was increased in anti-sense G-CSF- and GM-CSF-transfected A549 clones. Thus, although COX-2 activation seems to induce different biological behavior depending on the cell type, we propose that G-CSF and GM-CSF might accelerate tumor progression by directly regulating COX-2 expression, independently of an autocrine mechanism.     

Antiproliferative and differentiative effect of granulocyte-macrophage colony-stimulating factor on a variant human small cell lung cancer cell line

A variant clone was adopted during passages of a small cell lung cancer cell line, GKT3-1.3. The variant clone exhibited distinct characteristics with alterations in morphology, positive staining with nonspecific esterase stain, and an increase in surface specific markers OKM5, HLA-DR, Mo1, and My7, usually found on monocytes or their precursors. However, it exerted a very rapid proliferation just like immature cells. This new clone, GKT3-1.3V, was shown to have specific binding capacity to granulocyte-macrophage colony-stimulating factor (GM-CSF), with a number of binding sites comparable to that of myelomonocytes or monocytic cell lines. Thus its proliferation was inhibited by GM-CSF in clonogenic assay and suspension culture. Increase in the percentage of cells with surface marker Mo1 by the addition of GM-CSF suggested its differentiative effect. Cell cycle analysis showed that the antiproliferative effect of GM-CSF was due to a block in G0 or G1. The antiproliferative effect of GM-CSF was abolished by the addition of anti-GM-CSF antibody.     

Two phase I studies of carboplatin dose escalation in chemotherapy-naive ovarian cancer patients supported with granulocyte-macrophage colony-stimulating factor.

Recombinant human granulocyte-macrophage colony-stimulating factor (rHuGM-CSF) may reduce myelosuppression and, thus, allow dose escalation of certain chemotherapeutic agents. We conducted two sequential phase I trials of escalating doses of carboplatin and a fixed dose and schedule of rHuGM-CSF in ovarian cancer patients who had not previously had chemotherapy, i.e., chemotherapy-naive patients. In the first trial, patients were assigned to regimens of increasing dose levels of carboplatin (starting at 400 mg/m2) and fixed doses and schedules of cyclophosphamide (600 mg/m2) and rHuGM-CSF (10 micrograms/kg given subcutaneously once daily on days 2-11). Chemotherapy was given every 3 weeks (regimen A). In the subsequent trial, the design was the same except that cyclophosphamide was omitted (regimen B). Fifteen patients received regimen A, and seven patients received regimen B. In regimen A, all three patients treated at the first dose level tolerated five cycles at full doses. Hematologic toxicity was dose limiting at the 600-mg/m2 dose level. When 500 mg/m2 carboplatin was given, six of eight patients tolerated three or four cycles at full doses before requiring dose reductions or treatment delays. In regimen B, doses could not be escalated above the first dose level (600 mg/m2) because of severe hematological toxicity. Nonhematological toxicity was tolerable and managed with acetaminophen, antihistamines, and/or nonsteroidal, anti-inflammatory medication. Compliance was excellent. We conclude that (a) rHuGM-CSF can be given safely and reliably to chemotherapy-naive ovarian cancer patients receiving these treatment regimens, (b) early and severe thrombocytopenia was a major problem with or without cyclophosphamide with doses of carboplatin at or above 600 mg/m2, and (c) 500 mg/m2 carboplatin administered every 3 weeks is the highest dose in regimen A that can be given safely in the outpatient setting.     

Granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor promote malignant growth of cells from head and neck squamous cell carcinomas in vivo

Granulocyte colony-stimulating factor (G-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) are used to ameliorate cancer therapy-induced neutropenia and mucositis. Yet, first data in head and neck squamous cell carcinoma (HNSCC) indicate an impaired long-term prognosis on G-CSF treatment, and previous studies showed a contribution of both factors to the progression of human epithelial tumors. Therefore, we investigate the role of G-CSF and GM-CSF in progression of tumor cells from human HNSCC. Both factors stimulated proliferation and migration of tumor cell lines established from patient tumors expressing G-CSF and GM-CSF and/or their receptors. Blockade of G-CSF and GM-CSF inhibited tumor cell invasion in a three-dimensional organotypic culture model. The contribution of both factors to tumor malignancy was further confirmed in nude mouse transplants in vivo. Invasive and malignant growth yielding a similar tumor phenotype as the original patient tumor was exclusively observed in G-CSF- and GM-CSF-expressing tumors and was associated with enhanced and persistent angiogenesis and enhanced inflammatory cell recruitment. Although factor-negative tumors grew somewhat faster, they were characterized by lack of invasion, reduced and transient angiogenesis, and large necrotic areas. These data provide evidence for a progression-promoting effect of G-CSF and GM-CSF in human HNSCC and suggest further detailed evaluation of their use in the therapy of these tumors.     

In vivo effect of granulocyte-macrophage colony-stimulating factor on the kinetics of human acute myeloid leukemia cells

Granulocyte-macrophage colony-stimulating factor, (GM-CSF) was given at 8 micrograms/kg daily by continuous i.v. infusion for 72 h to six patients with acute myeloid leukemia (AML) in expansion and one with chronic myeloid leukemia in blastic crisis to determine whether it was possible to augment the proliferative activity of the neoplastic population. The percentage of marrow blasts in S phase (labeling index, LI) was increased in five patients (1.3-, 1.5-, 1.9-, 2.3- and 3.2-fold change). The increase in LI was similar 24 and 48 h after beginning GM-CSF. The RNA Index also increased in patients who showed an increased LI, suggesting that GM-CSF had recruited quiescent neoplastic cells into the cell cycle. Forty eight hours after beginning GM-CSF, chemotherapy was started. The fate of S phase cells, labeled in vivo with bromodeoxyuridine (BrdU) immediately before cytostatic treatment, was monitored. BrdU positive cells were identified by fluorescent antibody for up to 28 days. A preferential killing of BrdU (S phase) cells was observed in 5/7 patients who obtained a complete remission, whereas this was not apparent in the two patients who achieved only a partial remission. Chemotherapy induced a rapid and profound aplasia; its duration, however, was not significantly different from that observed in historical controls. GM-CSF may have a potential role in the treatment of AML, as this study shows that it recruits leukemic cells into the cell cycle without adversely prolonging aplasia after cycle-specific therapy.     

A double-blind placebo-controlled study with granulocyte-macrophage colony-stimulating factor during chemotherapy for ovarian carcinoma

In a placebo-controlled double-blind dose-finding trial, 15 patients with ovarian cancer stage III or IV received daily s.c. 1.5, 3, or 6 micrograms/kg recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF). At each dose step three patients received recombinant human GM-CSF, and two received placebo. Chemotherapy comprised 6 cycles of carboplatin, 300 mg/m2, and cyclophosphamide, 750 mg/m2, by i.v. bolus on day 1 every 4 weeks. GM-CSF, given on days 6-12 on an outpatient basis, raised the mean leukocyte count on days 7, 10, and 15 and the mean neutrophil count on days 7 and 10 at all dose levels as compared with the control group. Neutrophil counts of less than 0.5 x 10(9)/liter occurred in 20 of 22 cycles in the control group and in 5 of 17 cycles at the 6-micrograms/kg/day GM-CSF dose level (P less than 0.0005). In comparison with the control group, the mean eosinophil count was higher on days 10 and 15 at all GM-CSF doses, as was the mean monocyte count on day 15. The mean platelet count was raised at the 3- and 6-micrograms GM-CSF doses on days 15 and 22. Chemotherapy dose reduction or postponement due to myelotoxicity occurred in 9 of 28 cycles in the placebo groups versus 5 of 44 cycles in the GM-CSF group (not significant). Local skin infiltrates at the GM-CSF injection sites occurred in 8/9 patients, leading to premature removal of two patients from the study. Capillary leakage of 131I-albumin was increased in all patients 5 days after the first chemotherapy course but was not significantly affected by 4 days of GM-CSF treatment. Tumor necrosis factor alpha and C-reactive protein serum levels increased during GM-CSF administration at the 6-micrograms dose level, but interleukin 6 serum levels were not affected. We conclude that a dose of 3 and 6 micrograms/kg/day GM-CSF reduces the severity of neutropenia and thrombocytopenia after carboplatin-cyclophosphamide. This GM-CSF dose does not induce additional capillary leakage.     

Influence of bestatin on production of granulocyte-macrophage colony-stimulating factor from human peripheral blood mononuclear cells in vitro

Bestatin, (25, 3R)-3-amino-2-hydroxy-4-phenylbutyryl-L-leucine, enhanced proliferation of normal human bone marrow granulocyte-macrophage progenitor cells to form CFU-GM colonies in viscous methylcellulose medium in vitro. To elucidate the mechanisms of this effect, the levels of colony-stimulating factor (CSF) in culture supernatant of 1 x 10(6)/ml peripheral blood mononuclear cells (PBMC) cocultured with various concentrations of bestatin were determined using our newly developed sandwich enzyme-linked immunosorbent assay (ELISA). The level of granulocyte/macrophage-CSF (GM-CSF) in the supernatant in the presence of 0.1 microgram/ml bestatin was 312 pg/ml by this ELISA. The production of GM-CSF increased with increasing amounts of added bestatin.     

In vitro growth of myeloid and erythroid progenitor cells from myelodysplastic patients in response to recombinant human granulocyte-macrophage colony-stimulating factor

Marrow progenitor cells from 14 myelodysplastic (MDS) patients and 17 normal donors were assayed in semisolid cultures supplemented with increasing doses of recombinant human granulocyte-macrophage colony-stimulating factor (rhGM-CSF) or medium conditioned by 5637 bladder carcinoma cells (5637CM). At doses of supplements shown to be optimal for colony formation in cultures of normal marrow, myeloid (day 14) colony numbers were subnormal in 10 of 14 MDS marrows cultured in 5637CM and in 8 of 14 cultures containing rhGM-CSF (2.5 ng/ml). However, a high dose of rhGM-CSF (20 ng/ml) raised myeloid colony numbers in cultures of many MDS marrows, so that 9 of 14 now yielded colonies within the normal range; increased levels of 5637CM failed to do this. Erythroid colony growth was poor in 13 of 14 MDS marrow cultures supplemented with erythropoietin in addition to 5637CM or rhGM-CSF. High concentrations of rhGM-CSF did not increase erythroid growth. These data suggest that myeloid progenitors from the MDS clone may have a decreased responsiveness to hemopoietins which can be overcome at high concentrations of growth factors.     

No growth-stimulation of heterotransplanted human testicular cancer cell-lines by recombinant human granulocyte-macrophage colony-stimulating factor (gm-csf)

Hematopoietic growth factors such as granulocyte (G)- and granulocyte-macrophage (GM) colony-stimulating factor (CSF) are currently used as adjunct treatment for cancer chemotherapy in order to ameliorate neutropenia. The current investigation presented here focused on the in vivo effects of GM-CSF on two heterotransplanted human testicular cancer cell lines in a nude mouse model. GM-CSF was applied to tumor bearing mice in a dose range of 3 x 10(5) to 3 x 10(7) U/kg per day i.v. for 14 consecutive days. For all concentrations of GM-CSF a reduction of the tumor volume compared to untreated tumor bearing control mice was noted. The largest retardation of tumor growth was seen at the end of the experiments at day 30 with 3 x 10(7) U/kg of GM-CSF with a relative tumor volume of 72% +/- 8% for cell line H 12.1 and 73.5 +/- 9% for cell line H23.1 in comparison to untreated tumor bearing control mice. No side effects of GM-CSF application such as weight loss or fever were observed. Serum tumor necrosis factor (TNF) levels in mice treated with GM-CSF were considerably lower compared to untreated tumor carrying control animals. No growth stimulation of heterotransplanted human testicular germ cell tumors by GM-CSF has been observed in this nude mice tumor model. Further studies will have to demonstrate, whether GM-CSF is even possibly able to exhibit an antitumor effect. With respect to tumor cell kinetics the clinical use of GM-CSF in the treatment of patients with testicular cancer can be regarded as safe.     

Opposite effect of tumor necrosis factor alpha on granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor-dependent growth of normal and leukemic hemopoietic progenitors.

The effect of recombinant human tumor necrosis factor alpha (TNF-alpha) on normal and chronic myeloid leukemia granulocyte-macrophage progenitors (CFU-GM) growing in semisolid agar cultures in the presence of recombinant granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor was studied. Granulocyte-macrophage colony-stimulating factor-dependent growth of normal and chronic myeloid leukemia bone marrow CFU-GM was greatly enhanced by TNF-alpha at doses of 0.1 to 100 units/ml. Growth enhancement included neutrophil, eosinophil, and monocyte-macrophage colonies and clusters at 7 and 14 days of culture. Since similar results were achieved with highly enriched progenitor cell populations, devoid of accessory cells, an indirect effect on CFU-GM growth through the release by accessory cells of other cytokines upon TNF-alpha stimulation was thus ruled out. By contrast, the same doses of TNF-alpha inhibited the growth of normal CFU-GM in granulocyte colony-stimulating factor-dependent cultures. Taken together, our findings indicate that the final effect of TNF-alpha on normal bone marrow granulocyte-macrophage progenitor growth is dependent on the specific growth factor interacting with it, and that both normal and chronic myeloid leukemia CFU-GM are equally responsive to the combined effects of TNF-alpha and a given colony-stimulating factor.     

Effect of granulocyte-macrophage colony-stimulating factor on chemotherapy-induced oral mucositis in non-neutro-penic cancer patients

The aim of this study was to assess prospectively the efficacy of granulocyte-macrophage colony-stimulating factor (GM-CSF) in the management of chemotherapy-induced oral mucositis in non-neutropenic cancer patients. In a prospective open study, 30 cancer patients with chemotherapy-induced, neutropenia-independent oral mucositis were treated with GM-CSF (Schering Plough Corp, Kenilworth, NJ) prepared as a mouthwash solution (5-10 μg ml^-1). GM-CSF was administered within 24 hours of occurrence of oral mucositis x 4 to 6 times daily. Systemic GM-CSF was not permissible. Oral mucositis was graded according to the modified Radiation Therapy Oncology Group criteria. Six patients were subsequently excluded as they experienced neutropenia during GM-CSF therapy. The remaining 24 patients were all evaluable. Most patients had either Grade 3 or 4 gross (76%) or functional (54%) mucositis. The mean ± SEM gross oral mucositis scores for all 24 patients combined decreased from 3.08 ± 0.18 at baseline to 2.04 ± 0.19(p < 0.0001) after 2 days, 0.92 ± 0.16(p < 0.0001) after 5 days, and 0.25 ± 0.09(p < 0.0001) after 10 days of therapy. Likewise, the mean ± SEM functional oral mucositis scores decreased from 2.71 ± 0.18 at baseline to 1.58 ± 0.19(p < 0.0001) after 2 days, 0.75 ± 0.16(p < 0.0001) after 5 days, and 0.17 ± 0.08(p < 0.0001) after 10 days of therapy. The duration of severe oral mucositis was also shortened as Grade 0 or 1 (gross mucositis score) was evident in seven (29%), 20 (83%), and 24 (100%) patients by the 2nd, 5th, and 10th day of therapy, respectively. Similarly, Grade 0 or 1 (functional mucositis score) reported in 13 (54%), 19 (79%), and 24 (100%) by the 2nd, 5th, and 10th day of therapy respectively. It was found that GM-CSF mouthwash as used in this study has a significant recuperative efficacy on the severity, morbidity, and duration of chemotherapy-induced oral mucositis. A large randomized, placebo-controlled study is warranted to ascertain that benefit and determine the optimal dosages and schedule.     

A phase I/II trial of recombinant human granulocyte-macrophage colony-stimulating factor in the intensification of cisplatin and cyclophosphamide chemotherapy for advanced ovarian cancer.

A pilot study was undertaken in eight patients to assess the feasibility of recombinant human granulocyte-macrophage colony-stimulating factor (rH GM-CSF) support to intensify standard chemotherapy for advanced ovarian cancer using a shortened 15 day treatment interval. Only four patients completed the course of six cycles of cisplatin 75 mg m-2 and cyclophosphamide 750 mg m-2 with rH GM-CSF, 3-5 micrograms kg-1 day-1, days 3-14, but one of these suffered a toxic death on study. Another died of disease progression. There were two episodes of life-threatening infection (WHO grade 4), and three patients were withdrawn because of various rH GM-CSF-related problems. Although potentially affording some patients the hypothetical benefits of dose intensification, as well as the possible attraction of a shorter duration of chemotherapy, this regimen is not without problems.     

Phase I trial of thiotepa in combination with recombinant human granulocyte-macrophage colony-stimulating factor.

PURPOSE: The ability of growth factors to stimulate marrow recovery suggests their potential for use in dose intensification of cytotoxic drugs. We performed a phase I study of the alkylating agent thiotepa in combination with granulocyte-macrophage colony-stimulating factor (GM-CSF), with the goal of dose-escalation of thiotepa. Thiotepa was selected based on its capacity for dose escalation to more than 1 g/m2 in the marrow transplantation setting. PATIENTS AND METHODS: The starting dose of thiotepa (75 mg/m2) was the highest dose evaluated in our previous phase I trial. Thirteen patients received 22 courses of thiotepa and GM-CSF. The dose of GM-CSF was 10 micrograms/kg subcutaneously daily in six patients and 5 micrograms/kg in seven patients. RESULTS: Three patients (23%) developed grade 3 to 4 neutropenia on the first course, with a recovery to more than 1000/mm3 in 4.7 days (mean). Recovery was as rapid with the 5 micrograms/kg as it was with the 10 micrograms/kg GM-CSF dose. Thrombocytopenia grade 3 to 4 affected seven of 13 (54%) patients in the first course; counts recovered to more than 50,000/mm3 in a median of 15 days. GM-CSF at either dose did not influence markedly the severity or duration of thrombocytopenia, and did not permit dose escalation of thiotepa. Among the seven patients who received a second cycle of treatment, six of seven experienced grade 3 or 4 thrombocytopenia that lasted a median of 15.5 days. Five had thrombocytopenia that lasted more than 35 days after one to three cycles of treatment. Plasma concentrations of thiotepa and tepa were measured by gas chromatography in eight patients. The plasma elimination of thiotepa fit a two-compartment open model with a harmonic mean terminal half-life of 2.44 hours. The mean total body clearance was 217.9 mL/min/m2, and the mean steady-state volume of distribution (Vdss) was 36.8 L/m2. The half-life of tepa was 7.98 hours, and the ratio of the area under the plasma concentration versus time curve (AUC) of tepa to that of thiotepa was 0.79. CONCLUSIONS: These data were consistent with our previous observations at this dose, and indicated that the severity of toxicity in these patients was not explained by aberrant pharmacokinetic indices. We conclude that, independent of effects on neutropenia, severe and cumulative platelet toxicity precludes further escalation of thiotepa dose despite the use of GM-CSF.