Filgrastim (previously known as recombinant methionyl human granulocyte colony-stimulating factor [r-metHuG-CSF]) is a hematopoietic growth factor that promotes the survival, proliferation, differentiation, and function of progenitor and mature cells of the neutrophil granulocyte lineage [11]. It has potent granulopoietic effects in vivo [12-14], and its clinical pharmacology has been defined [15]. In two phase III studies [16, 17], filgrastim therapy that was started 1 day after chemotherapy reduced the severity and duration of neutropenia and decreased by nearly half the incidence of febrile neutropenic episodes and the average duration of hospitalization and antibiotic use per patient. The use of filgrastim to prevent febrile neutropenic episodes in patients with cancer who receive chemotherapy is now an approved indication in many countries.

Not all patients with cancer who have had chemotherapy and who have been treated with filgrastim receive it to reduce the risk for neutropenic sepsis. Approximately 14% of filgrastim in the United States has been used to treat episodes of postchemotherapy febrile neutropenia after they occur rather than to prevent them (unpublished market research data, Amgen, Inc.). To determine if filgrastim administered in addition to standard empiric inpatient antibiotic therapy was also beneficial in patients presenting with established fever and neutropenia, we conducted a multicenter, randomized, double-blind, placebo-controlled study. This is the first report of a study assessing therapeutic intervention with filgrastim in this clinical setting.

Methods

Top Methods Results Discussion Author & Article Info References

We present the analysis of findings from 218 patients enrolled at four centers between October 1988 and February 1992. The study drug was filgrastim, a recombinant human protein produced in Escherichia coli [18] and supplied by Amgen, Inc., Thousand Oaks, California.

Patients

Eligible patients with febrile neutropenia were adults (age 16 years) receiving chemotherapy for histologically confirmed cancer other than myeloid leukemia who had an oral temperature of greater than 38.2 °C and a neutrophil count less than 1.0 x 10⁹/L. The neutrophil count included the number of bands and polymorphonuclear forms counted in manual differentials from blood smears. Patients were ineligible if they were already receiving antibiotic therapy, if they had septic shock (a systolic blood pressure of <90 mm Hg, poor peripheral perfusion, and abnormal mentation), if they were receiving myeloablative chemotherapy with marrow transplantation support, if they had severe renal impairment (serum creatinine level >0.7 mM), abnormal liver function (bilirubin >5 times the upper limit of normal), were pregnant or breast-feeding, or had a history of allergic reactions to the study antibiotics. Patients could enter the randomized trial only once. If subsequent episodes of fever and neutropenia developed, patients were offered open-label filgrastim therapy according to the same procedures described in the protocol. Data from these episodes are not included in this report. The study was approved by the relevant institutional review boards of each treatment center, and all patients gave signed informed consent.

Study Design

Our study was a randomized, double-blind, placebo-controlled trial. Patients were randomly assigned to receive either filgrastim or placebo from identically labeled vials. Stratification was not done before randomization. The randomization code was maintained by the Biostatistical Department of Amgen, Inc., and was not broken until all collected data had been verified by Amgen and its designee, Biomedicus (Boronia, Australia), and had been audited by a separate department at Amgen. Enrolled patients were admitted to the hospital, and a full medical history was taken and physical examination done. Baseline investigations included full blood counts with differential leukocyte counts and chest radiographs. Cultures of blood (in duplicate), urine, and other clinically suspicious sites were done. Empiric treatment was started with piperacillin (4 g intravenously every 6 hours) and tobramycin (2 mg/kg of body weight intravenously for the loading dose and 80 mg intravenously every 8 hours thereafter). Antibiotic dose was adjusted for patient size, renal and hepatic impairment, and drug serum levels according to standard criteria. Therapy with filgrastim or placebo was initiated within 12 hours of the start of antibiotic therapy at a dose of 12 µg/kg per day by continuous subcutaneous infusion with the use of a portable pump. Filgrastim (supplied in 5- or 10-mL vials as a clear, colorless solution of 300 µg/mL) or placebo (supplied in matching vials for double-blinding) was diluted in a sterile solution of 5% glucose in water to a volume of 50 mL. Patients remained in the study until 4 consecutive afebrile days (peak daily temperature, <37.5 °C) had passed and the blood neutrophil count was greater than 0.5 x 10⁹/L or until 28 days had elapsed. Filgrastim or placebo doses were reduced or the drugs were withdrawn before study completion if the blood neutrophil count exceeded 10.0 x 10⁹/L (dose reduced to 6 µg/kg per day), 15.0 x 10⁹/L (dose reduced to 3 µg/kg per day), or 20.0 x 10⁹/L (drug withdrawn). Intravenous antibiotics were administered until study completion, although the actual drugs or doses could be altered at the investigators' discretion within broad protocol guidelines.

Study End Points

Our study evaluated the effect of filgrastim on the number of days of fever and on the number of days of neutropenia associated with presumed or documented infection that was treated with antibiotics. Blood counts were done daily, and oral temperatures were recorded at least four times per day. Two end points were defined for days of neutropenia and are specified here as the number of days on which the neutrophil count was less than 0.5 x 10⁹/L and less than 1.0 x 10⁹/L. The days of fever end point was defined as the number of consecutive days (after and including the day of presentation) on which the maximum oral temperature was 37.5 °C or greater, plus any subsequent periods beginning with a day of temperature peak of greater than 38.2 °C and ensuing days with peaks of 37.5 °C or greater.

We also analyzed several related end points. The time to resolution of fever was defined as the number of days elapsed before the first complete day with a temperature less than 37.5 °C (subsequent fevers were not included in this definition). Late fever was defined as a temperature greater than 38.2 °C occurring after the fever had initially resolved. The time to resolution of febrile neutropenia was defined as the number of days elapsed before the first day with both a temperature less than 37.5 °C and a neutrophil count of 1.0 x 10⁹/L or less (subsequent days of fever or neutropenia were not included in this definition). The prospectively defined secondary end points of the study were days on study (hospitalization) and the incidence of the change of antibiotics. This was an inpatient study; therefore, all days on study were spent in the hospital. Although some patients were inpatients when fever and neutropenia developed and some remained in the hospital after completing the study, the days on study equalled the period of hospitalization for most patients. Patients who remained in the hospital after completing the study did so for reasons other than continued morbidity related to fever and neutropenia.

Safety

All adverse events were graded according to World Health Organization criteria [19]. Data collected included daily clinical evaluation, toxicity grading, and blood examinations; and urea, electrolytes, creatinine, liver function tests, clotting tests, and urinalysis, all of which were done every fifth day. The numbers of red blood cell and platelet transfusions were also recorded. Deaths that occurred during the study and within 30 days of study completion were recorded, and a blinded assessment was made of any possible contribution to the death by the study drug.

Statistical Analyses

We used a 1:1 ratio of patients treated with filgrastim to patients receiving placebo. The sample size was chosen to detect a 25% difference in the protocol-specified primary end points of days of fever and days of neutropenia with 80% power at a significance level of 0.05. However, an unplanned interim analysis was done for administrative purposes, rather than with the intent of early study termination, when 87 patients had completed the study. Therefore, the inferential testing of the primary outcomes for the final analysis was done at a significance level of 0.048 (two-tailed) [20].

All patients randomly assigned to receive and treated with at least one dose of the study drug were analyzed for efficacy (n = 216). All patients randomly assigned were analyzed for safety outcomes (n = 218). We did not do an evaluable patient subset analysis. We calculated descriptive statistics for the study end points defined above.

Differences between the filgrastim and placebo groups in days of neutropenia, days of fever, and days on study (hospitalization) were evaluated with the Wilcoxon rank-sum test. We calculated point estimates and approximate 95% confidence intervals for the differences between the values in the two groups for each of these primary outcome variables [21]. The chi-square test was used to measure the association of categorical variables. We used the Kaplan-Meier product-limit method [22] and the Mantel-Cox log-rank test [23] to estimate and compare the time-to-resolution distributions. The Cox proportional-hazards model was used in univariate analyses to relate known or suspected covariates to each time-to-resolution end point. Interactions among treatment covariates were assessed by multivariate analyses [24]. Patients who were withdrawn from the study without fulfilling the criteria for study completion contributed to the study end points until their withdrawal, after which they were censored. We adjusted all analyses for influential covariates using the Mantel-Haenszel procedure [25]. These covariates were age ( 55 years vs. >55 years); baseline Eastern Cooperative Oncology Group performance status (1 to 2 vs. 3 to 4); tumor type (solid vs. lymphoma vs. acute lymphocytic leukemia); and days since the last chemotherapy treatment (<10 days vs. 10 days).

Although stratification was not done before randomization, subset analyses were done based on clinically relevant subgroups determined before the analysis. These subgroups were tumor type (solid tumor, lymphoma, or acute lymphocytic leukemia), baseline neutrophil count (<0.1 x 10⁹/L, 0.1 to <0.5 x 10⁹/L, or 0.5 to <1.0 x 10⁹/L), days elapsed since completion of chemotherapy (<10 days or 10 days), and infection type (microbiologically or clinically documented infection, or possible or unlikely infection). The group "lymphoma" included patients with all grades of non-Hodgkin lymphoma and Hodgkin disease, 1 patient with B-cell chronic lymphocytic leukemia, and 2 patients with multiple myeloma. A microbiologically or clinically documented infection was defined as one for which a pathogenic organism was cultured from an appropriate site, such as blood, urine, or sputum, or for which clinical or radiologic localizing signs of infection other than fever and tachycardia were present.

Statistical analyses were done using SAS software version 6.07 [26] or BMDP software version 1990 [27] on a SUN/UNIX operating system.

Results

Top Methods Results Discussion Author & Article Info References

Two of the 218 patients entered and randomly assigned were removed from the trial before receiving the study drug. Of the 216 patients remaining, 109 received filgrastim and 107 received placebo. Seventeen patients in the filgrastim group and 14 patients in the placebo group were withdrawn before study completion, but all were included in the intention-to-treat analyses. Of these 31 patients, 11 died, 5 were considered treatment failures, 8 had persistent fever no longer thought to be caused by infection, 4 were removed at the patient's request, and 3 were withdrawn for technical reasons. The enrollment characteristics of the patients are shown in Table 1. The two treatment groups were well matched, with only minor differences in age, sex ratio, and extent of previous chemotherapy.

Filgrastim significantly reduced the median number of days of neutropenia. The filgrastim group had 1 less day with a neutrophil count of less than 0.5 x 10⁹/L (a 25% reduction) and 2 fewer days with a neutrophil count of less than 1.0 x 10⁹/L (a 40% reduction) (Table 2 and Figure 1). The median neutrophil count over time showed an earlier, steeper increase and a higher plateau in the filgrastim group (Figure 1). The proportion of patients with a neutrophil count of less than 1.0 x 10⁹/L in relation to time was statistically significantly less in the filgrastim group than in the placebo group (Figure 2).

View larger version (22K): [in this window] [in a new window]   Figure 1. Neutrophil recovery in patients with febrile neutropenia and proportion remaining on study by number of days from study entry. A. Piecewise linear plot of median neutrophil counts of patients remaining on study by study days 1 to 7 and (B) by days 1 to 28. C. Number of patients on study by study day.

View larger version (21K): [in this window] [in a new window]   Figure 2. Resolution of neutropenia, fever, and febrile neutropenia by number of days from study entry. A. Probability of not achieving a neutrophil count of 1.0 x 109/L. B. Probability of unresolved fever. C. Probability of failure to resolve febrile neutropenia. Resolution of fever was defined as the first complete day with a temperature less than 37.5 °C. Resolution of febrile neutropenia was defined as the first day of having both a temperature less than 37.5 °C and a neutrophil count greater than 1.0 x 109/L. No further events occurred after day 17. Numbers on plots indicate the number of patients remaining at risk.

The median number of days of fever were the same, but a 1-day reduction in the means reflected fewer patients with prolonged fever in the filgrastim group (Table 2). No difference in the time to resolution of fever was seen between the groups Figure 2, although the time to resolution of febrile neutropenia was reduced from a median of 6.0 to 5.0 days in the filgrastim group (P = 0.01; adjusted log-rank test) (Figure 2). The incidence and duration of late fever was nearly identical in each of the treatment groups.

Hospitalization

The median number of days on study (hospitalization) was the same, and although the mean was lower in the filgrastim group, the differences were not statistically significant (P = 0.09) (Table 2). Prolonged hospitalization, however, occurred predominantly in the placebo group (Figure 1). When the distribution of the number of days on study was divided into quartiles, the relative risk for requiring prolonged hospitalization (4th quartile, >11 days) was approximately twice as high in the placebo group as the risk in the filgrastim group (relative risk, 2.1; CI, 1.1 to 4.1; P = 0.02).

Antibiotic and Filgrastim Therapy

The number of days of treatment and the total dose of piperacillin and tobramycin delivered were almost identical in the two groups. No difference was seen in the use of other intravenous antibiotics: Forty-six percent of patients in the filgrastim group and 41% of patients in the placebo group received alternative antimicrobial therapy (P = 0.48). Six percent of patients treated with filgrastim and 11% of the patients receiving placebo were given intravenous antifungal therapy (P = 0.13). Filgrastim was given for a median of 7.0 days, with a mean total dose of 4965 µg.

Subgroup Analyses

It was of interest to explore the potential benefits of filgrastim within subsets of the study group to better understand its actions and to help design future studies that would further evaluate the role of filgrastim in patients with fever and neutropenia. Baseline neutrophil count, infection type, tumor type, and days elapsed since completion of chemotherapy were considered useful clinical measures for dividing the study group into subsets.

Filgrastim decreased the number of days of neutropenia and accelerated neutrophil recovery regardless of the baseline neutrophil count (Table 3). Statistically significant differences occurred in the patients who had baseline neutrophil counts between 0.1 x 10⁹/L and less than 0.5 x 10⁹/L and in the patients whose counts were less than 0.1 x 10⁹/L. Statistical significance was not reached for the few patients who had baseline neutrophil counts of 0.5 to less than 1.0 x 10⁹/L. Filgrastim resulted in a reduction in the time to resolution of fever only in patients with baseline neutrophil counts of less than 0.1 x 10⁹/L (P = 0.02) (Figure 3). The acceleration in neutrophil recovery in response to filgrastim was particularly evident in patients with solid tumors and in patients for whom 10 or more days had elapsed between completion of chemotherapy and the onset of fever and neutropenia (Table 4, Figure 4). Although filgrastim decreased the number of days of neutropenia in patients with lymphoma, the number of days of fever and days on study (hospitalization) were similar in the treatment groups in this subset of patients. By contrast, filgrastim appeared to reduce the number of days of fever and days on study (hospitalization) in patients with solid tumors, although these differences did not reach statistical significance after adjustment for days elapsed since chemotherapy (Table 4).

View larger version (20K): [in this window] [in a new window]   Figure 3. Subset analysis according to baseline neutrophil count of probability of unresolved fever by number of days from study entry. A. Baseline neutrophil count of less than 0.1 x 109/L. B. Baseline neutrophil count of 0.1 to less than 0.5 x 109/L. C. Baseline neutrophil count of 0.5 to less than 1.0 x 109/L. Resolution of fever was defined as the first complete day with a temperature of less than 37.5 °C. No further events occurred after day 17. Numbers on plots indicate the number of patients remaining at risk.

View larger version (17K): [in this window] [in a new window]   Figure 4. Neutrophil recovery in patients with febrile neutropenia remaining on study by the number of days from study entry according to tumor type and number of days elapsed since completion of chemotherapy. A. Solid tumors. B. Lymphomas. C. Less than 10 days since the completion of chemotherapy. D. Ten days or more since the completion of chemotherapy. The proportions of patients remaining on study at day 7 were 71% in the filgrastim group and 79% in the placebo group.

Safety

As expected of patients receiving chemotherapy for cancer, several potential adverse events were recorded in both study groups. The double-blind, placebo-controlled design allowed us to identify those events likely to be caused by filgrastim therapy. No significant difference (P > 0.2) in adverse events was seen between the two treatment groups except for musculoskeletal symptoms, which were more common in the filgrastim group. Thirty-five patients (32%) in the filgrastim group reported musculoskeletal symptoms (bone, muscle, or joint pain) compared with 24 (22%) patients in the placebo group (P = 0.13). These symptoms were mild to moderate in severity (World Health Organization grades 1 and 2) in 32 (91%) patients in the filgrastim group and in 23 (96%) patients in the placebo group. There was no difference between groups in the incidence or duration of thrombocytopenia (platelet count <50 x 10⁹/L) (58 patients in the filgrastim group [53%] and 56 patients in the placebo group [52%]); anemia (hemoglobin concentration <80 g/L) (62 patients in the filgrastim group [57%] and 64 patients in the placebo group [60%]); the need for red blood cell transfusions (71 patients in the filgrastim group [65%] and 66 patients in the placebo group [62%]) or platelet transfusions (39 patients in the filgrastim group [35%] and 36 patients in the placebo group [34%]). There was no significant difference in the death rates between the treatment groups, and the study drug was not considered to have contributed to any of the deaths: Nine patients in the filgrastim group and 7 in the placebo group died during the study or within 7 days of its completion, and 3 and 8 patients, respectively, died within 30 days of study completion (12 [11%] deaths in the filgrastim group and 15 [14%] deaths in the placebo group; P = 0.5).

Discussion

Top Methods Results Discussion Author & Article Info References

Our study clearly showed that filgrastim accelerated neutrophil recovery in patients who received antibiotics for febrile neutropenia that developed after chemotherapy. The study patients were heterogeneous, had nonmyeloid malignancies, and received various chemotherapy regimens ranging from standard adjuvant treatment for breast cancer to intensive salvage treatments for lymphoma and acute lymphoblastic leukemia. Based on the medians of the two treatment groups, filgrastim resulted in 2 fewer days of a neutrophil count of less than 1.0 x 10⁹/L and 1 less day of a neutrophil count of less than 0.5 x 10⁹/L. This benefit was seen regardless of the baseline neutrophil count, although statistical significance was not achieved for the relatively few patients with baseline neutrophil counts of 0.5 to less than 1.0 x 10⁹/L. Despite this acceleration in neutrophil recovery, filgrastim treatment was not associated with a statistically significant reduction in the number of days of fever. However, filgrastim significantly reduced the time to resolution of febrile neutropenia. Although the medians for the number of days on study (hospitalization) were identical, there was a 1.3-day reduction in the means that reflects the higher proportion of patients receiving placebo who required a prolonged hospital stay.

As neutrophil recovery has been shown to be important in resolving infection associated with neutropenia [6, 7], why did the accelerated neutrophil recovery induced by filgrastim not result in a statistically significant reduction in the number of days of fever? Further, the risk for developing a second bacterial infection or a fungal infection is increased during prolonged periods of neutropenia [28, 29], yet filgrastim did not reduce the need to change antibiotics or to use antifungal agents. Several factors may have contributed to this discordance: the important role of antibiotics to control infections in these patients (which contributes to defervescence), the confounding effect of tumor-related fever, and possibly insufficient statistical power of the study. Interestingly, in the exploratory analyses, the subset of patients with microbiologically or clinically documented infection (fever unrelated to tumor) had a statistically significant reduction in the number of days of fever with filgrastim, and patients with severe neutropenia (neutrophil count <0.1 x 10⁹/L) had a reduction in the time to resolution of fever.

The days on study in the hospital were determined by the protocol and largely depended on fever resolution; it is therefore not surprising that filgrastim did not substantially shorten hospital stay for the study group overall. Nevertheless, a reduction in the risk for prolonged hospitalization in the filgrastim group was statistically significant, suggesting that patients at risk for delayed recovery, such as those presenting with low neutrophil counts, may benefit most.

Two features of the study design require comment. First, it was planned in 1987 when only limited phase I data for filgrastim were available. The protocol required 4 consecutive afebrile days and neutrophil recovery (neutrophil count, >0.5 x 10⁹/L) before discharge because this was typical of clinical practice at the time. This might now be considered conservative because recent studies of various antibiotic regimens in patients with febrile neutropenia have reported median durations of therapy as short as 6 days [6]. Nevertheless, guidelines of the Infectious Diseases Society of America published in 1990 still recommend a minimum of 7 days of intravenous antibiotic treatment [5].

Second, the filgrastim was administered at a dose of 12 µg/kg per day subcutaneously because available phase I data showed this to be safe and effective in stimulating neutrophil production in patients with cancer and neutropenia [12-14, 30]. The continuation of filgrastim therapy until neutrophil counts were greater than 20 x 10⁹/L in patients with persisting fever would now be considered unnecessary but was based on theoretical concerns of an abrupt decrease in neutrophil counts after filgrastim therapy was stopped. Continuous subcutaneous infusion was chosen for convenience because the dose of 12 µg/kg per day would have required multiple bolus injections each day [30]. Subsequent studies have shown filgrastim at 5 µg/kg per day by bolus subcutaneous injection to be effective in other clinical settings [16, 17]. Only one study has directly compared the efficacies of bolus and infusional subcutaneous administration [30]. The optimal route of administration, dose, and duration of treatment with filgrastim required for benefit in patients with febrile neutropenia induced by chemotherapy is yet to be determined.

Is filgrastim cost-effective in the clinical setting of established febrile neutropenia? Our study was not designed to prospectively collect relevant cost–benefit data; therefore, an accurate assessment cannot be made. The average cost per patient of filgrastim, which was used conservatively and perhaps excessively in this study, was $1986 (4965 µg x $0.4/µgrams of Neupogen [Amgen, Inc., Thousand Oaks, California]) against a 1.3-day reduction in the mean hospital stay. However, to account for the full benefits of filgrastim to the patient, the reduced risk for prolonged hospitalization should be considered.

Although several potential adverse events were recorded, only musculoskeletal symptoms, 94% of which were mild to moderate, appeared to be associated with filgrastim treatment. These symptoms occurred particularly during neutrophil recovery. We could identify no other filgrastim-related adverse effects even after correcting for drug exposure. These observations are consistent with previous studies that showed filgrastim to be safe with few adverse effects [15]. In particular, filgrastim does not cause fever, and this makes it an appropriate hematopoietic factor for use in febrile neutropenic patients in whom fever and continued fever are the best clinical signs of sepsis and failure of response to therapy, respectively [29, 31]. Death rates were similar for the two groups and similar to those reported in other series. In this clinical setting, with a mortality rate of only 5% to 10%, many more patients would be required to show a statistically significant difference in death rates.

Phase II studies have shown a reduction in the severity and duration of neutropenia when filgrastim is administered soon after chemotherapy [12-14, 30]. In two phase III studies [16, 17], filgrastim given to patients with small-cell lung cancer from the day after chemotherapy reduced the incidence of febrile neutropenic episodes by approximately 50%. In patients in whom febrile neutropenic episodes still occurred, the duration of the episodes was shorter in the filgrastim group, although this was not statistically significant [16]. In our study, filgrastim treatment was started within 12 hours of the initiation of antibiotic therapy at the onset of fever. This was a median of 10 days after the last chemotherapy dose, indicating that starting filgrastim treatment later could still substantially reduce the duration of neutropenia.

Recombinant human granulocyte-macrophage colony-stimulating factor has also been shown to accelerate neutrophil recovery in patients presenting with neutropenia after chemotherapy [32], but its use in patients presenting with established fever and neutropenia has only been reported in a small, double-blind, randomized trial in 30 patients [33]. In that study, therapy with granulocyte-macrophage colony-stimulating factor (2.8 µg/kg per day by continuous intravenous infusion) appeared to accelerate neutrophil recovery but did not reduce the duration of fever in 24 patients who could be evaluated for this measurement. A larger study is required to determine if granulocyte-macrophage colony-stimulating factor is beneficial in this clinical setting.

It seems probable that patients at risk for morbidity from febrile neutropenia developing after chemotherapy will benefit maximally from filgrastim if it is given before neutropenia develops. Further studies are being done to identify the most efficient preventive schedule of filgrastim with respect to timing of therapy after various chemotherapy regimens [34]. Our data show that in patients not receiving preventive filgrastim treatment, the addition of filgrastim to empiric therapy with broad-spectrum antibiotics for the treatment of established fever and neutropenia results in accelerated neutrophil recovery and a shorter duration of febrile neutropenia. Although these effects may not result in a significant reduction in the number of days of fever and hospital stay, the risk for prolonged hospitalization appears to be decreased.