Impact of primary prophylaxis on febrile neutropenia within community practices in the US

Abstract

Objective: To determine if granulocyte-colony-stimulating factor (G-CSF) primary prophylaxis is associated with a lower risk of febrile neutropenia (FN) than non-primary prophylaxis.

Methods: This was a retrospective, cohort study of medical records from a random sample of patients with solid tumours and lymphomas treated in 99 community oncology practices in 2003 (n=5319). Consecutively-sampled patients treated with chemotherapy and either filgrastim (Neupogen), pegfilgrastim (Neulasta) or no G-CSF were included (n=3123). Multivariate logistic regression estimated the odds of FN in patients receiving G-CSF primary prophylaxis (within 3 days of first chemotherapy cycle) compared with non-primary prophylaxis (delayed or no G-CSF).

Results: Patients receiving primary prophylaxis were less likely to develop FN than patients receiving non-primary prophylaxis (OR=0.49, 95% CI 0.34–0.71, p<0.001). Chemotherapy characteristics were associated with development of FN including, receipt of at least three chemotherapy drugs versus one (OR=2.13, 95% CI 1.17–3.89, p=0.014) and regimens with at least one myelosuppressive drug (OR=2.37, 95% CI 1.19–4.73, p=0.014).

Conclusion: Patients receiving G-CSF primary prophylaxis had significantly lower odds of developing FN than those receiving non-primary prophylaxis. Incidence of FN may be underestimated, as care not recorded in the medical oncologist's chart was not captured.

Key words::

• febrile neutropenia
• filgrastim
• G-CSF
• growth factors
• pegfilgrastim

Introduction

Neutropenia remains the dose-limiting toxicity of many chemotherapy regimens which places patients at increased risk for serious and potentially life-threatening infections. Neutropenia may also be associated with reductions in chemotherapy dose or dose intensity, as well as delays in treatment, that can have a negative impact on clinical outcomes^1–8.

Randomised controlled clinical trials (RCTs) have demonstrated that primary prophylaxis with granulocyte-colony-stimulating factors (G-CSFs) significantly decreases the risk of febrile neutropenia (FN) by 48–94% compared with placebo by promoting haematopoietic recovery after chemotherapy^9–11. Initially, the benefits of G-CSFs were demonstrated in RCTs with filgrastim (Neupogen) administered 24 hours following completion of chemotherapy for an average of 10–11 days during each cycle^12–14. These benefits were also observed with pegfilgrastim (Neulasta^†), a pegylated version of filgrastim with a sustained duration of action that is administered 24 hours after chemotherapy, only once per cycle^11,13–15. A recent meta-analysis of RCTs found that primary prophylaxis with G-CSFs was associated with a decreased risk of FN, a decreased risk of early mortality, and a higher likelihood of patients receiving the planned relative dose intensity of chemotherapy¹⁶.

While the efficacy of filgrastim and pegfilgrastim has been well-established in RCTs, the effectiveness of these therapies has not been fully evaluated in the community-oncology practice setting. RCTs are not always generaliseable to practice-based use and populations due to physician and patient preferences in drug therapy and usage, patient and disease characteristics, and other factors. It is therefore important to determine whether the benefits observed in controlled clinical trials extend to the potentially heterogeneous patient groups treated under the uncontrolled conditions of actual clinical practice.

This is particularly important in the case of G-CSFs, where substantial deviations in the initiation and duration of G-CSF treatment have been reported in community-based practices^17–21. These studies found that patients in community practice often initiated filgrastim therapy later in the cycle and received fewer days of therapy than in randomised controlled trials, where filgrastim was initiated 24 hours after the last dose of chemotherapy and continued for a mean or median of 10–11 days in each cycle^13–15. In contrast, pegfilgrastim was initiated more often in the first cycle of chemotherapy, and was associated with improved outcomes compared with filgrastim^17,18.

These findings raise the possibility that delays in the initiation of G-CSF and reduced duration of filgrastim observed in community practice settings may negatively impact clinical outcomes. In an attempt to address this question, a retrospective observational cohort study was conducted to compare the effectiveness of any G-CSF as primary prophylaxis versus non-primary prophylaxis on FN events in patients with cancer treated in community-oncology practices. The overall rates of FN by treatment group for this cohort have been previously reported¹⁸. The hypothesis was that primary prophylaxis with any G-CSF would be associated with a lower incidence of FN than non-primary prophylaxis.

Additionally two secondary objectives were also studied. The first was to assess patient characteristics and patterns of care (chemotherapy and G-CSF) across treatment groups that were associated with receipt of G-CSF primary prophylaxis. The second was to compare, separately, the receipt of either G-CSF (filgrastim or pegfilgrastim) as primary prophylaxis versus receipt of non-primary prophylaxis on the incidence of FN.

Patients and methods

Study population and selection

A two-stage sampling procedure identified a retrospective cohort of patients with breast, lung, ovarian, or colon cancer, or lymphoma who initiated chemotherapy treatment in 2003. Patient cohorts were further defined based on new use of filgrastim, pegfilgrastim or no G-CSF²². For the first stage, oncology practices were selected using stratified random sampling from a commercial database owned by IMS Health. A probability sample of practices was identified. Practices were ineligible if they treated less than 150 patients with chemotherapy, were government affiliated, or were unable to identify a sample of patients. For practices that couldn't meet these criteria, declined to participate, or did not respond, another practice was randomly selected (n=1353). A random sample of 100 practices were selected; however one practice subsequently decided not to participated, so the final sample included 99 practices.

The second stage identified a consecutive sample of eligible patients from the 99 participating practices using one of the following data sources within the oncology practices: pharmacy records, billing records, appointment books, or chemotherapy administration records. The cohort was limited to patients who initiated chemotherapy regimens administered every 3–4 weeks (≥21-day regimens). Daily or weekly regimens are generally associated with a negligible risk of FN and every-2-week regimens require G-CSF prophylaxis; thus, patients treated with these regimens were excluded¹⁹. Patients were also excluded if they received a G-CSF after an FN event (secondary prophylaxis), were enrolled in a clinical trial that included a G-CSF or if they were receiving sargramostim (Leukine, GM-CSF, Bayer Health Care Pharmaceuticals, Wayne, NJ, USA). Details of the sampling procedure and patient selection have been previously published¹⁸.

From May to September 2004, trained research staff abstracted patient medical and laboratory records for up to eight chemotherapy cycles. Data on patient and tumour characteristics, chemotherapy, G-CSF use, laboratory test, FN events and hospitalisations were recorded. These included patient demographics and patterns of care including chemotherapy drugs and the G-CSF product selection and timing of initiation. Comorbid conditions were captured using the modified Charlson index²³. The number of chemotherapy drugs and the presence of myelosuppressive chemotherapy drugs in the regimen were recorded for assessing the patterns of care of chemotherapy. Myelosuppressive chemotherapy drugs included at least one of the following: cladribine, cytarabine, cyclophosphamide, docetaxel, fludarabine, ifosfamide, irinotecan, mechlorethamine, methotrexate, or topotecan. Data collection procedures complied with the United States Health Insurance Portability and Accountability Act (HIPAA) regulations²⁴. The study protocol was determined exempt from review by the Western Institutional Review Board (ethics committee) since no active drug was administered and patient charts were reviewed retrospectively.

Patients were initially categorised to one of three treatment groups based upon recorded G-CSF use. Patients who received a G-CSF within 3 days of chemotherapy in the first cycle were classified as having received primary prophylaxis. Patients who received a G-CSF after more than 3 days of chemotherapy in the first cycle or anytime during subsequent cycles were classified as delayed. Patients who received both filgrastim and pegfilgrastim were assigned to the treatment group according to the first G-CSF received to follow an intent-to-treat assignment¹⁸. Lastly, patients who never received a G-CSF were classified as no G-CSF. The decision on whether or not to use a G-CSF as primary prophylaxis is primarily based upon the chemotherapy regimen and patient factors (e.g. age) ³. When G-CSF is initiated later in the first or at any time in a subsequent cycle, it is generally in response to neutropenia, the main risk factor for FN. Patients who never received a G-CSF are therefore likely to have a lower risk of developing FN than patients in who received delayed G-CSF. As a conservative estimate of the effectiveness of G-CSF primary prophylaxis on the risk of FN, an intent-to-treat approach was used and the delayed and no G-CSF groups were categorised as non-primary prophylaxis.

FN was determined based on documentation of any of the following in the patient medical records: FN, temperature of ≥38 °C with documentation of neutropenia or an absolute neutrophil count (ANC) <1.0 × 10⁹/L on the same day, or hospitalisation for FN.

Statistical analyses

The patient characteristics and patterns of care across treatment groups were compared using chi-square tests for categorical variables. All statistical tests were two-sided.

The bivariate associations between the incidence of FN and receipt of G-CSF primary prophylaxis, patient characteristics and chemotherapy regimen characteristics were assessed. The odds of FN [odds ratio (OR)] for patients receiving either G-CSF product as primary prophylaxis versus non-primary prophylaxis (delayed or no G-CSF) was examined using multivariate logistic regression adjusting for patient (age, gender, stage of disease, comorbidities, and baseline serum albumin, haemoglobin and ANC) and chemotherapy characteristics (number of chemotherapy and myelosuppressive drugs in the regimen).

A second multivariate logistic regression model was planned to estimate the effect of each G-CSF separately on the odds of FN. This included filgrastim primary prophylaxis versus the non-primary prophylaxis treatment group and pegfilgrastim primary prophylaxis versus non-primary prophylaxis. This model adjusted for the same covariates previously described.

Sensitivity analyses were also conducted to estimate the impact of excluding patients that never received a G-CSF from the non-primary prophylaxis treatment group. These analyses thereby provide a direct comparison of G-CSF timing of initiation, primary prophylaxis versus delayed, on the odds of developing FN. All analyses were conducted using SAS version 8.2. Statistical significance was set at p<0.05 for all comparisons.

Results

Study population

A total of 5,319 eligible patients were identified from the 99 participating community oncology clinics (Figure 1). A total of 3,123 of these patients underwent ≥21-day chemotherapy cycles and were included in the present analysis. Of these patients, 822 received primary prophylaxis with G-CSF (n=101 filgrastim, n=721 pegfilgrastim), 1,523 received delayed treatment with G-CSF (n=790 filgrastim, n=733 pegfilgrastim) and 778 patients did not receive a G-CSF (Figure 1).

Patient characteristics and patterns of care

As shown in Table 1, differences among patient characteristics were observed across these three treatment groups, including gender (p<0.001), cancer type (p<0.001), and comorbid conditions (p=0.039). Patients who received G-CSF primary prophylaxis were more likely to be treated with a chemotherapy regimen with at least three drugs (42%) as compared with patients who received delayed G-CSF initiation or never received a G-CSF (32% and 21%, respectively, p <0.001). Among patients who never received a G-CSF, 79% had at least one myelosuppressive drug in the chemotherapy regimen as compared with 94% of patients who received G-CSF primary prophylaxis and 91% of patients who had delayed CSF initiation (p<0.001).

By definition, all patients who received G-CSF primary prophylaxis received G-CSF in the first chemotherapy cycle (within 3 days of chemotherapy). Among patients who received delayed G-CSF, 39% initiated G-CSF in the first cycle. Patients who received G-CSF primary prophylaxis were more likely to receive pegfilgrastim (88%), whereas patients who received delayed G-CSF were more likely to receive filgrastim (52%).

Among patients who received filgrastim (n=891), the mean days of treatment was higher for patients who received primary prophylaxis versus those who received delayed initiation during the first cycle of chemotherapy (6.5 [SD 3.9] versus 3.4 days [SD 2.4], p<0.001). This same pattern was observed in subsequent cycles where filgrastim in primary prophylaxis was administered for 6.7 days [SD 4.0] versus 4.3 days [SD 2.9] with delayed initiation (p<0.001).

Effectiveness of G-CSF primary prophylaxis on the risk of FN

After adjusting for patient characteristics and chemotherapy regimen characteristics, patients receiving primary prophylaxis were less likely to develop FN than patients categorised as non-primary prophylaxis (i.e. received delayed or no G-CSF) (OR=0.49, 95% CI 0.34–0.71, p<0.001) (Table 2). Chemotherapy characteristics were a significant predictor of FN incidence. Compared with regimens with one drug, the odds of FN was higher with three or more drugs in the regimen (OR=2.13, 95% CI 1.17–3.89, p=0.014). The inclusion of at least one myelosuppressive drug in the chemotherapy regimen increased the odds of FN by a factor of approximately 2 (OR=2.37, 95% CI 1.19–4.73, p=0.014).

For the secondary objective of comparing each G-CSF separately, the incidence of FN was the highest for patients not receiving primary prophylaxis (7.5%) followed by filgrastim (6.9%), then pegfilgrastim (4.2%). As compared to patients not receiving primary prophylaxis the odds of FN for patients who received primary filgrastim prophylaxis were 0.76 (95% CI 0.34–1.71, p=0.509) and for those receiving pegfilgrastim, the odds were 0.46 (95% CI 0.31–0.68, p<0.001). This model yielded the same significant predictors of FN incidence as described above with similar variability in the estimate.

The sensitivity analyses demonstrated a greater impact of primary prophylaxis on development of FN, after excluding patients that never received a G-CSF. For the comparison of G-CSF primary prophylaxis versus non-primary prophylaxis, the odds of FN decreased to 0.35 (95% CI 0.24–0.51, p<0.0001). Similarly the odds of FN decreased for both G-CSFs separately as primary prophylaxis when compared to non-primary prophylaxis. For pegfilgrastim the odds were 0.33 (95% CI 0.22–0.49, p<0.0001) and with filgrastim the odds were 0.54 (95% CI 0.24–1.23, p=0.144).

Discussion

In this retrospective cohort study of patients treated in community-oncology practices, primary G-CSF prophylaxis was associated with a significant reduction in the risk of FN compared with non-primary prophylaxis of delayed or no G-CSF initiation. After adjusting for patient and chemotherapy characteristics, patients receiving primary prophylaxis with either filgrastim or pegfilgrastim had an approximate 50% reduction in the odds of developing FN compared with those who did not receive primary prophylaxis. This reduction in FN associated with primary prophylaxis in a community-based setting is consistent with that observed in RCTs^9,11,26.

When evaluated independently, primary prophylaxis with pegfilgrastim was associated with a statistically significant 54% reduction in the odds of FN as compared to patients that received delayed or no G-CSF. This is in contrast with the 24% reduction with filgrastim that was not statistically significant. The lack of significant reduction in the risk of FN with filgrastim primary prophylaxis may be related to several factors. There were a relatively low number of patients that received filgrastim as primary prophylaxis and these patients received a shorter mean duration of therapy of approximately 7 days, in contrast to the mean or median of 10–11 days reported in the registrational randomised, controlled trials^9,10. Since pegfilgrastim is only administered once per chemotherapy cycle, duration of use is not an issue. This difference in product characteristic may enable better adherence to treatment recommendations with pegfilgrastim than filgrastim, which may impact the development of FN.

While treatment with a more myelosuppressive chemotherapy regimen was associated with a greater likelihood of primary prophylaxis, patient characteristics appeared to have less of an impact on use of primary prophylaxis in this study. The only patient characteristics associated with receipt of primary prophylaxis observed in this patient population was gender, cancer type and comorbid conditions. Physicians may not have considered patient characteristics in their use of primary prophylaxis because the updated guidelines stating these as a reason for initiating primary prophylaxis were published in 2005 and 2006 after data collection for this study was completed^27–29. There is little guidance on the extent to which patient risk factors increase a patient's baseline FN risk, making it difficult for physicians to act upon these recommendations in clinical practice. Further research quantifying the FN risk associated with patient characteristics may enable their use in clinical algorithms for determining which patients will benefit most from primary prophylaxis.

The study results should be interpreted with acknowledgement of additional limitations. Data were obtained by abstracting patients' charts and therefore are dependent on the completeness and accuracy of medical record documentation. Thus, we may have underestimated the true FN incidence since patients who sought care in settings where care was not documented in their oncologist's chart (i.e., emergency room or hospital) may have been missed. Additionally, the FN incidence reflects the real world effectiveness of the drugs and not reflective of patients receiving the recommended duration of filgrastim prophylaxis. Lastly, the relatively recent approval of pegfilgrastim in January 2002 prior to this study period may have momentarily changed filgrastim prescribing patterns. Studies evaluating prescribing patterns of filgrastim post-2003 and with a longer study period, may address this potential limitation.

Conclusions

Among patients treated in the ‘real-world’ of community oncology practices, primary prophylactic use of G-CSF within 3 days of chemotherapy was associated with a decrease in FN compared with G-CSF administered after 3 days in any chemotherapy cycle or no G-CSF administered. These findings also demonstrated that pegfilgrastim as the choice of G-CSF in primary prophylaxis was associated with a significant decrease in FN compared with no primary prophylaxis. This study provides information about the use of G-CSFs in actual clinical practice, where treatment does not always adhere to dosing derived from RCT studies. Communicating the benefit of primary prophylaxis for patients receiving myelosuppressive chemotherapy in a community practice setting may improve prevention of FN, especially for patients at increased risk.

Figure 1. Study Sample.

Table 1. Patient characteristics and patterns of care.

	Primary G-CSF prophylaxis (n=822) n (%)	Delayed G-CSF initiation (n=1523) n (%)	No G-CSF (n=778) n (%)	p-value
Patient characteristics
Age
<65 years	505 (61.4)	961 (63.1)	454 (58.4)	0.087
65 years and older	317 (38.6)	562 (36.9)	324 (41.7)
Gender
Male	212 (25.8)	320 (21.0)	257 (33.0)	<0.001
Female	610 (74.2)	1203 (79.0)	521 (67.0)
Cancer type
Breast cancer	416 (50.6)	783 (51.4)	267 (34.3)	<0.001
Lung cancer	176 (21.4)	348 (22.9)	250 (32.1)
non-Hodgkin's lymphoma	173 (21.1)	221 (14.5)	60 (7.7)
Other *	57 (6.9)	171 (11.2)	201 (25.8)
Stage^†
Limited or unknown	692 (84.2)	1264 (83.0)	626 (80.5)	0.130
Advanced (IV)	130 (15.8)	259 (17.0)	152 (19.5)
Comorbid conditions
0	591 (71.9)	1070 (70.3)	507 (65.2)	0.039
1	169 (21.6)	325 (21.3)	190 (24.4)
2 or more	62 (7.5)	128 (8.4)	81 (10.4)
Baseline serum albumin
at least 3.5 g/dL and higher	717 (87.2)	1356 (89.0)	666 (85.6)	0.054
<3.5 g/dL	105 (12.8)	167 (11.0)	112 (14.4)
Baseline anaemia
at least 10 g/dL and higher	797 (97.0)	1465 (96.2)	751 (96.5)	0.6274
Haemoglobin <10 g/dL	25 (3.0)	58 (3.8)	27 (3.5)
Baseline absolute neutrophil count
at least 1.5 × 10^9 cells/L and higher	811 (98.7)	1509 (99.1)	775 (99.6)	0.129
<1.5 × 10^9 cells/L	11 (1.3)	14 (0.9)	3 (0.4)
Patterns of care
Number of chemotherapy drugs in regimen
1	65 (7.9)	117 (7.7)	121 (15.6)	<0.001
2	409 (49.8)	920 (60.4)	494 (63.5)
3 or more	348 (42.3)	486 (32.0)	193 (21.0)
At least one myelosuppressive drug in regimen
No	49 (6.0)	133 (8.7)	161 (20.7)	<0.001
Yes	773 (94.0)	1390 (91.3)	617 (79.3)
G-CSF received
Pegfilgrastim	721 (87.7)	733 (48.1)	—	<0.001
Filgrastim	101 (12.3)	790 (51.9)	—
No G-CSF	0	0	778 (100.0)
Cycle G-CSF initiated
1	822 (100)	580 (38.9)	—	<0.001
2	—	556 (36.5)	—
3	—	203 (13.3)	—
4 or later	—	184 (12.1)	—

p-Values calculated using chi-square statistics.

*Other cancer types included small-cell lung cancer, non-small cell lung cancer, ovarian, colorectal, and Hodgkin's Disease; ^†advanced stage includes III, IV and limited stage includes I and II.

ANC, absolute neutrophil count; G-CSF, granulocyte-colony-stimulating factor.

Table 2. Odds of febrile neutropenia with G-CSF primary prophylaxis versus non-primary prophylaxis (delayed or no G-CSF), adjusted for patient and chemotherapy regimen characteristics (n=3123)^*.

	Adjusted odds ratio (95% confidence interval)
	Incidence of FN n (%)	G-CSF primary prophylaxis	p-value
G-CSF primary prophylaxis
Yes	37 (4.5)	0.49 (0.34–0.71)	<0.001
Non-primary prophylaxis	172 (7.5)	Ref	—
Patient characteristics
Age <65 years	127 (6.6)	Ref	—
Age ≥65 years	82 (6.8)	0.96 (0.71–1.30)	0.802
Male	59 (7.5)	Ref	—
Female	150 (6.4)	0.86 (0.63–1.19)	0.364
Stage		Ref	—
Limited or unknown	174 (6.7)	0.93 (0.63–1.38)	0.719
Advanced (stage IV)	35 (6.5)
Comorbid conditions		Ref	—
0	134 (6.2)	1.24 (0.88–1.75)	0.219
1	52 (7.6)	1.44 (0.90–2.31)	0.130
≥2	23 (8.5)
Baseline serum albumin		1.54 (1.00–2.36)	0.048
<3.5 g/dL	35 (9.1)	Ref	—
≥3.5 g/dL or unknown	174 (6.4)
Baseline anaemia			0.716
Haemoglobin <10.0 g/dL	10 (9.1)	1.14 (0.56–2.33)	—
Haemoglobin ≥10.0 g/dL	199 (6.6)	Ref
Baseline ANC		1.04 (0.30–3.64)	0.949
<1.5 × 10^9/L	2 (7.1)	Ref	—
≥1.5 × 10^9/L	207 (6.7)
Chemotherapy characteristics
Number of chemotherapy drugs in regimen
1	14 (4.6)	Ref	—
2	92 (5.1)	0.94 (0.52–1.72)	0.851
≥3	103 (10.3)	2.13 (1.17–3.89)	0.014
At least one myelosuppressive drug in regimen?
No	10 (2.9)	Ref	—
Yes	199 (7.2)	2.37 (1.19–4.73)	0.014

G-CSF, granulocyte-colony-stimulating factor; ANC, absolute neutrophil count; FN, febrile neutropenia; Ref, reference.

Acknowledgments

Declaration of interests: This research was sponsored by Amgen, Inc. J.L.M. has disclosed that she was an employee of Amgen at the time of this writing, but is still an Amgen stockholder; D.L.H. is an Amgen stockholder. M.W. and V.A.M. have no relevant financial relationships to disclose. The authors thank Beiying Ding for support in developing the statistical analysis plan and i3 Statprobe for conducting the statistical analyses.

Notes

*,† Neupogen and Neulasta are registered tradenames of Amgen Inc., Thousand Oaks, CA, USA.