Comparative effectiveness of pegfilgrastim, filgrastim, and sargramostim prophylaxis for neutropenia-related hospitalization: two US retrospective claims analyses

Abstract

Objective:

Few studies have compared the effectiveness of filgrastim (FIL), pegfilgrastim (PEG), and sargramostim (SAR) to reduce the risk of febrile neutropenia (FN) associated with myelosuppressive chemotherapy (M-CT). Two large commercial database analyses were separately conducted to examine the incidence of neutropenia-related and all-cause hospitalizations associated with FIL, PEG, and SAR prophylaxis for patients receiving M-CT for non-Hodgkin lymphoma (NHL), Hodgkin lymphoma, or solid tumors.

Methods:

Separate retrospective US claims database analyses utilized patient data from January 1, 2004 to April 30, 2010 using the HealthCore Integrated Research Database (HIRD^SM) and January 1, 2001 to August 31, 2009 using OptumInsight’s (formerly Ingenix) database. Patients were ≥18 years old and treated with M-CT for NHL, Hodgkin lymphoma, and solid tumors. All identified M-CT cycles with prophylactic (first 5 days of cycle) FIL, PEG, or SAR were included in the analysis. Patterns of administration and incidence rates of all-cause and neutropenia-related hospitalization were examined on a per-cycle basis.

Results:

In total, 9330 and 8762 patients with cancer, representing 30,264 and 24,215 chemotherapy cycles (28,189 and 22,649 (PEG), 1669 and 1351 (FIL), 406 and 215 (SAR)) from the HIRD^SM and OptumInsight databases, respectively, were included in the separate database analyses. Both the HIRD^SM and OptumInsight analysis showed that SAR and FIL prophylaxis had a higher risk of neutropenia-related hospitalization (SAR: OR = 3.48 [95%CI = 2.11, 5.74] and 2.81 [1.62, 4.87]; FIL: 1.78 [1.28, 2.48] and 2.36 [1.82, 3.06], respectively) and all-cause hospitalization (SAR: 2.18 [1.41, 3.36] and 2.41 [1.58, 3.68]; FIL:1.57 [1.25, 1.97] and 1.95 [1.60, 2.38], respectively) vs PEG.

Limitations:

Medical claims do not contain information about chemotherapy dose, and hospitalizations were not validated against the patient’s chart.

Conclusion:

In this comparative effectiveness study, use of PEG was associated with a lower risk of neutropenia-related and all-cause hospitalizations compared to use of FIL or SAR prophylaxis.

Keywords::

• Neutropenia
• Prophylaxis
• Colony-stimulating factor
• Outcomes
• Pegfilgrastim
• Filgrastim
• Sargramostim

Introduction

Neutropenia is the most frequent dose-limiting toxicity of systemic myelotoxic chemotherapy1. Febrile neutropenia (FN, oral temperature ≥38.0°C over 1 h and absolute neutrophil count [ANC] <500 neutrophils/mcl or <1000 neutrophils/mcl and a predicted decline to ≤500/mcl over the next 48 h)2 often results in hospitalization and treatment with broad-spectrum antibiotics and is associated with significant morbidity, mortality, and cost3–7.

Granulocyte colony-stimulating factor (G-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) are the two major types of CSFs in the US market8. The G-CSFs filgrastim (FIL) and pegfilgrastim (PEG) are hematopoietic growth factors that stimulate the proliferation and differentiation of neutrophils and promote hematopoietic recovery after chemotherapy9. FIL is rapidly cleared by the kidneys and requires daily subcutaneous injections for ∼11 days after chemotherapy10. PEG is the longer-acting pegylated form of FIL and requires a single dose administered subcutaneously once per chemotherapy cycle10–12. FIL and PEG are indicated to decrease the incidence of infection, as manifested by FN, in patients with non-myeloid malignancies receiving myelosuppressive chemotherapy (M-CT) associated with a clinically significant incidence of FN8.

The GM-CSF sargramostim (SAR) is a hematopoietic growth factor that stimulates the proliferation and differentiation of neutrophils and monocytes/macrophages. SAR is indicated (in the US) for use after induction chemotherapy in older patients with acute myelogenous leukemia to shorten the time to neutrophil recovery and to reduce the incidence of severe and life threatening infections and infections resulting in death13. It is also used in clinical practice to reduce the incidence of FN associated with myelosuppressive chemotherapy (M-CT), although it is not approved for this indication.

Prophylactic FIL, PEG, and SAR can minimize the risk of severe neutropenia14–19. Few studies have compared the effectiveness of FIL, PEG, and SAR to reduce the incidence of FN associated with myelosuppressive chemotherapy. Two randomized trials14,15 and a prospective evaluation on medication use18 reported that SAR and FIL have comparable efficacy, safety, and tolerability; whereas two retrospective patient chart reviews demonstrated that FIL was better at preventing chemotherapy-induced neutropenic events than SAR20,21. The results of several large, retrospective, healthcare claims analyses have also been reported. One such analysis demonstrated that administration of SAR resulted in a lower risk of infection-related hospitalization and lower infection-related hospitalization costs than FIL and PEG22. Another large claims study found that the risk of hospitalizations was lower for PEG compared with FIL23, and in a different claims analysis, the odds of developing febrile neutropenia were reported to be 1.4-times higher among cancer patients receiving FIL vs PEG, a finding ascribed to shortened duration of use of FIL in a given chemotherapy cycle in community oncology practices24. Two additional US claims analyses reported that the risk of hospitalization for neutropenic complications during cancer chemotherapy is lower with PEG prophylaxis than with either FIL or SAR prophylaxis25. Finally, another US claims study26 evaluated the effectiveness of prophylactic PEG and FIL (i.e., used within the first 5 days of each cycle) and reported a significantly lower rate (1.1% vs 3.5%) of neutropenia hospitalizations for PEG vs FIL prophylaxis.

Due to the inconsistency in results reported above, the present study conducted two separate database analyses on large US commercial health plan populations, which included a strict definition of prophylaxis (within first 5 days of a cycle), in order to determine if there is a difference in neutropenia-related and all-cause hospitalizations between patients receiving FIL, PEG, and SAR prophylaxis and M-CT for non-Hodgkin lymphoma (NHL), Hodgkin lymphoma, or solid tumors.

Methods

Study design

Two retrospective US claims analysis using data from the HealthCore Integrated Research Database (HIRD^SM Wilmington, DE) and the OptumInsight (Eden Prairie, MN) claims database were conducted. All data were de-identified. All study materials were handled in compliance with the Health Insurance Portability and Accountability Act of 1996 regulations.

Claims data from January 1, 2004 to April 30, 2010 (HIRD^SM) and from January 1, 2001 to August 31, 2009 (OptumInsight) were used for this analysis. Eligible patients were ≥18 years old with NHL, Hodgkin lymphoma (OptumInsight only), breast, lung, colorectal, ovarian, or other solid tumors (OptumInsight only) cancer treated with chemotherapy. Cancer type was determined by the presence of two or more International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes (at least 7 days apart) for NHL (200.x, 202.x), Hodgkin lymphoma (201.xx), breast cancer (HIRD^SM: 174.x, 175.x, 233.0 x; OptumInsight: 174.x), lung cancer (HIRD^SM: 162.x, 231.2 x; OptumInsight: 162.x, 163 x), colorectal cancer (HIRD^SM: 153.x, 154.x, 230.3 x, 230.4; OptumInsight: 153.x, 154.0 x, 154.1, 154.8), ovarian cancer, or other cancer (OptumInsight: 140.xx–152.xx, 155.xx–161.xx, 164.xx–165.xx, 170.xx–172.xx, 175.xx–176.xx, 179.xx–182.xx, 184.xx–189.xx). Patients were also required to have at least one claim for FIL, and/or PEG, and/or SAR during the chemotherapy course. Patients were excluded if they had any claim for chemotherapy during 6 months (HIRD^SM) or 12 months (OptumInsight) prior to the initiation of chemotherapy; less than 6 months (HIRD^SM) or 12 months (OptumInsight) of continuous eligibility before the initiation of chemotherapy (or less than 60 days of continuous eligibility after the initiation of chemotherapy [HIRD^SM]); at least one medical claim for bone marrow or stem cell transplant during the study period; medical claims indicating services provided in a skilled nursing facility or hospice; or claims for the use of more than one type of CSF agent per chemotherapy course. Additional exclusion criteria for OptumInsight were: medical claims indicating primary cancers in more than one body site at baseline/during the chemotherapy course or radiation at the initiation of chemotherapy. Additional exclusion criteria for HIRD included: pharmacy claims for chemotherapy drugs.

Outcomes

Patient characteristics

The following patient characteristics were reported at the cycle level by CSF: the number of first and subsequent cycles received, mean number of days of prophylaxis, mean patient age, percent of female patients, geographic location within the US, tumor type, history of metastatic disease, neutropenia or infection in previous cycle, history of radiation therapy, and history of surgery. Diagnoses and procedures (comorbidities and a history of anemia, neutropenia, infections, radiation therapy, and surgery) were identified by diagnosis codes on medical claims, including ICD-9-CM, Current Procedural Terminology (CPT), and Healthcare Common Procedure Coding System (HCPCS) codes, for both outpatient visits and inpatient stays.

Chemotherapy cycles and courses

The first identified chemotherapy course for each patient during the intake period was used in the analysis of both databases. Each chemotherapy course may include several cycles. The first chemotherapy course began on the index date and ended with any of the following, whichever came first: absence of any chemotherapy claims within the 60 days after the index date, the end of insurance eligibility or study period, or the beginning of radiation initiation.

The first chemotherapy cycle and each unique cycle were identified for all patients and were defined according to previously published definitions23. Chemotherapy cycles were excluded if any of the subsequent administrations of chemotherapy occurred between 7–19 days following the cycle initiation. The end of the first chemotherapy cycle was defined as the first administration of the second cycle of chemotherapy if it occurred between 20–59 days after the initiation of the first cycle. In addition, the OptumInsight analysis excluded all first course chemotherapy cycles during which prophylactic FIL, PEG, or SAR was initiated after the 9th cycle. As well, HIRD^SM excluded cycles during which neutropenia-related hospitalization (broad definition) occurred prior to CSF initiation within the cycle; this was to ensure CSF use was prophylactic, not therapeutic.

Patterns of CSF administration were identified for each chemotherapy cycle using medical HCPCS, generic product identifier (GPI) codes (HIRD^SM), and outpatient pharmacy national drug codes (NDC) claims (OptumInsight). CSF administration was categorized as ‘prophylaxis’ when the first claim for FIL, PEG, or SAR appeared within the first 5 days of starting the chemotherapy cycle23. Switching among FIL, PEG, and SAR during a course of chemotherapy was not assessed and those with evidence of more than one CSF were removed from the study. Cycles in which the CSF claim appeared after day 5 of the cycle were excluded from the analyses. For medical claims, days of prophylaxis of CSF therapy is defined as the number of claims on unique days. For pharmacy claims, days of prophylaxis of CSF therapy was defined by HIRD^SM as the number of days supply; whereas OptumInsight did not include patients with pharmacy claims in this calculation.

Risk of hospitalization

The effects of prophylactic FIL, PEG, and SAR on neutropenia-related and all-cause hospitalizations for patients receiving chemotherapy were assessed. Neutropenia-related hospitalizations were defined using a ‘narrow’ or a ‘broad’ criterion. The narrow criterion was a primary or secondary diagnosis of neutropenia (ICD-9-CM code 288.0), and the ‘broad’ criterion was a primary or secondary diagnosis of neutropenia or fever of unknown origin (ICD-9: 780.6), or infection (codes available upon request). These definitions have been used to identify neutropenia-related hospitalization among cancer patients in previous studies3,25,27–29, and the neutropenia ICD-9-CM code 288.0 has been reported to have an 80% sensitivity for capturing neutropenia hospitalizations when compared to chart data30–34. All outcomes were examined for each cycle from the day following initiation of CSF prophylaxis to the end of the cycle, a maximum of 60 days.

Statistical analysis

Descriptive statistics are provided for all study variables. Numbers and percents are provided for dichotomous and polychotomous variables; means, medians, standard deviations, and percentiles are provided for continuous variables. Sub-group comparisons were performed using Wilcoxon rank sum/t-tests for continuous data and chi-square/Fisher’s exact tests for categorical data.

Incidence rates of all-cause and neutropenia-related hospitalizations were assessed per chemotherapy cycle for the FIL, PEG, and SAR groups separately. Unadjusted and adjusted odds of hospitalization for patients receiving prophylaxis with SAR vs PEG and FIL were compared using generalized estimating equations (GEEs). GEE models were used on the cycle level to adjust for intra-observation correlation and provide more specific estimates given that patients may have multiple cycles of chemotherapy. Since the outcomes are binary (e.g., neutropenia vs not, infection vs no infection), a binomial distribution and logistic link function was specified for all GEEs. Models were adjusted for potential confounders, including age and sex of patient, geographic location, type of insurance, tumor characteristics (including type of cancer and existence of metastases (in HIRD^SM models)), treatment characteristics (including myelotoxicity of chemotherapy), chronic comorbidities (using the Deyo-Charlson comorbidity index), and prior neutropenic event. Models of all-cause and neutropenia-related hospitalizations were adjusted for potential confounders, including age and sex of patient, geographic location, type of insurance, tumor characteristics (including type of cancer and existence of metastases (in HIRD^SM models)), treatment characteristics (including myelotoxicity of chemotherapy), chronic comorbidities (using the Deyo-Charlson comorbidity index), and prior neutropenic event. In order to adjust for multiple comparisons of all-cause and neutropenia-related hospitalizations within a study (HIRD^SM and OptumInsight), comparisons were made at a statistical significance level based on the Bonferroni correction (0.05/6 = 0.008).

All statistical analyses was performed using SAS version 9.1 (SAS corporation, Cary, NC) and Stata version 10 SE (StataCorp, College Station, TX).

Results

Patient characteristics

The HIRD^SM and OptumInsight analyses included 9330 and 8762 patients with cancer meeting the inclusion criteria, representing 30,264 and 24,215 cycles during which FIL (1669 and 1351 cycles), PEG (28,189 and 22,649 cycles), or SAR (406 and 215 cycles) were delivered, respectively (Table 1).

Table 1.  Patient characteristics: cycle level results.

	HIRD^SM						OptumInsight
	FIL (1)	PEG (2)	SAR (3)	p-value			FIL (1)	PEG (2)	SAR (3)	p-value
				1 vs 3	2 vs 3	1 vs 2				1 vs 3	2 vs 3	1 vs 2
Patients, n (%)	621 (6.66)	8569 (91.84)	140 (1.50)	–	–	–	628 (8.44)	6719 (90.3)	94 (1.26)	–	–	–
Cycles, n (%)
All cycles	1669 (100)	28,189 (100)	406 (100)	–	–	–	1351 (100)	22,649 (100)	215 (100)	–	–	–
First cycles	284 (17.02)	5888 (20.89)	85 (20.94)	–	–	–	268 (19.84)	5009 (22.12)	45 (20.93)	–	–	–
Subsequent cycles	1385 (82.98)	22,301 (79.11)	321 (79.06)	–	–	–	1083 (80.16)	17,640 (77.88)	170 (79.07)	–	–	–
Mean number of days of early prophylaxis, (SD)^a	6.41 (5.85)	–	7.48 (3.79)	0.001	–	–	4.73 (2.98)	–	6.64 (4.80)	<0.001	–	–
Age, mean (SD)	56.59 (11.32)	56.05 (11.33)	59.99 (11.76)	0.000	0.000	0.056	54.97 (11.20)	54.84 (10.61)	57.11 (11.11)	0.009	0.002	0.676
Female, n (%)	1380 (82.68)	23,519 (83.43)	324 (79.80)	0.174	0.051	0.424	1016 (75.20)	18,044 (79.67)	149 (69.30)	0.066	0.000	0.000
Geographic Region, n (%)				0.000	0.000	0.000				<0.001	<0.001	<0.001
Northeast	–	–	–	–	–	–	162 (11.99)	1407 (6.21)	0	–	–	–
Midwest	–	–	–	–	–	–	376 (27.83)	5793 (25.58)	49 (22.79)	–	–	–
East	249 (14.92)	4189 (14.86)	22 (5.42)	–	–	–	–	–	–	–	–	–
South	300 (17.97)	8638 (30.64)	52 (12.81)	–	–	–	675 (49.96)	12,007 (53.01)	134 (62.33)	–	–	–
Central	222 (13.30)	7909 (28.06)	133 (32.76)	–	–	–	–	–	–	–	–	–
West	898 (53.80)	7453 (26.44)	199 (49.01)	–	–	–	138 (10.21)	3442 (15.20)	32 (14.88)	–	–	–
Deyo-Charlson Comorbidity Index, mean (SD)	4.93 (3.15)	5.26 (3.16)	4.71 (2.96)	0.182	0.000	0.000	4.80 (2.52)	4.57 (2.36)	5.47 (2.45)	<0.001	<0.001	<0.001
Tumor type, n (%)				0.000	0.000	0.000				0.001	0.000	0.000
Breast cancer	1000 (59.92)	17,551 (62.26)	207 (50.99)	–	–	–	774 (57.29)	12,935 (57.11)	106 (49.3)	–	–	–
Lung cancer	231 (13.84)	4885 (17.33)	90 (22.17)	–	–	–	178 (13.18)	2990 (13.2)	45 (20.93)	–	–	–
NHL	213 (12.76)	2904 (10.30)	62 (15.27)	–	–	–	215 (15.91)	4227 (18.66)	39 (18.14)	–	–	–
HL	0	0	0	–	–	–	22 (1.63)	26 (0.11)	0 (0)	–	–	–
Colorectal cancer	60 (3.59)	335 (1.19)	19 (4.68)	–	–	–	19 (1.41)	99 (0.44)	8 (3.72)	–	–	–
Ovarian cancer	165 (9.89)	2514 (8.92)	28 (6.90)	–	–	–	85 (6.29)	1445 (6.38)	5 (2.33)	–	–	–
Other cancer	0	0	0	–	–	–	58 (4.29)	927 (4.09)	12 (5.58)	–	–	–
Metastatic disease, n (%)^b,c	660 (39.54)	12,590 (44.66)	130 (32.02)	0.117	0.000	0.000	2271 (30.93)	9537 (26.74)	320 (34.52)	0.026	0.000	0.000
Neutropenia or infection in previous cycle^c,d,e	1134 (67.94)	16,783 (59.54)	273 (67.24)	0.786	0.002	0.000	863 (63.88)	10,036 (44.31)	135 (62.79)	0.758	0.000	0.000
History of radiation therapy, n (%)^c,f	70 (4.19)	1825 (6.47)	36 (8.87)	0.000	0.052	0.000	182 (13.47)	4090 (18.06)	30 (13.95)	0.848	0.119	0.000
History of surgery, n (%)^c,f	1050 (62.91)	17,784 (63.09)	243 (59.85)	0.254	0.180	0.885	861 (63.73)	13729 (60.62)	120 (55.81)	0.026	0.152	0.023

^aOptumInsight: patients with pharmacy claims were not included in this calculation.

^bOptumInsight: if an individual has at least two of the same 3-digit codes for metastases in any position 365 days prior to first cycle start date. Codes for identifying metastases are 196.xx, 197.xx, or 198.xx. Codes must be separated by at least 7 days. NOTE: Claims from laboratories, diagnostic testing centers, or any diagnostic tests were not considered when identifying metastases

^cHIRD^SM: For first cycle, during the 6 months prior to cycle start date.

^dUnspecified fever, infection, or ICD-9 = 288.xx.

^eOptumInsight: prior 90-days if 1st cycle occurrence of neutropenia or infection in previous cycle (prior 90-days of 1st cycle).

^fOptumInsight: For first cycle, during 12 months prior to cycle start date.

Breast cancer was the most common tumor type among cycles with CSF prophylaxis in both databases; the second-most common was lung for HIRD^SM and NHL for OptumInsight. The majority of patients had experienced neutropenia or infection in either the 6 months prior to the start of the first cycle (HIRD^SM) or for the 90 days prior to the start of the first cycle (OptumInsight) as well as a history of surgery (6 months prior to start of first cycle (HIRD^SM) or 12 months prior to start of first cycle (OptumInsight); whereas few had a history of radiation therapy or chemotherapy cycles with CSF prophylaxis surgery (6 months prior to start of first cycle (HIRD^SM) or 12 months prior to start of first cycle (OptumInsight) (Table 1).

Chemotherapy cycles and courses

First cycle use of CSF prophylaxis was 58.85% (17.02% FIL, 20.89% PEG, and 20.94% SAR) in the HIRD^SM population and 62.89% (19.84% FIL, 22.12% PEG, and 20.93% SAR) for the OptumInsight population. The mean (SD) days of FIL prophylaxis was 6.41 days (5.85) vs 7.48 days (3.79) for SAR prophylaxis in the HIRD^SM population and 4.73 (2.98) days vs 6.64 days (4.80) in the OptumInsight population (Table 1).

Risk of hospitalization

SAR and FIL prophylaxis had a higher risk of neutropenia-related hospitalization (broad definition) in the HIRD^SM analysis (SAR: OR = 3.48, 95% CI = 2.11–5.74; FIL: OR = 1.78, 95% CI = 1.28–2.48) as well as the OptumInsight analysis (SAR: OR = 2.81, 95% CI = 1.62–4.87; FIL: OR = 2.36, 95% CI = 1.82–3.06) vs PEG (Table 2).

Table 2.  GEE model^a of risk of hospitalizations (neutropenia-related hospitalization (broad and narrow definition) and all-cause hospitalizations)—prophylaxis of FIL, PEG, and SAR.

Cycle-level analysis	HIRD^SM, All			OptumInsight^b, All
	OR	95% CI	p-value^c	OR	95% CI	p-value^c
Number of unique patients	9330			8762
Number of cycles	30,264			24,215
Neutropenia-related hospitalization—broad definition
Pegfilgrastim	(Reference group)			(Reference group)
Filgrastim	1.783	[1.280, 2.483]	0.001	2.357	[1.816, 3.059]	0.000
Sargramostim	3.481	[2.111, 5.740]	0.000	2.813	[1.624, 4.873]	0.000
Neutropenia-related hospitalization—narrow definition
Pegfilgrastim	(Reference group)			(Reference group)
Filgrastim	3.179	[2.030, 4.980]	0.000	1.215	[0.569, 2.595]	0.615
Sargramostim	6.365	[3.525, 11.495]	0.000	4.277	[1.657, 11.035]	0.003
All-cause hospitalizations
Pegfilgrastim	(Reference group)			(Reference group)
Filgrastim	1.573	[1.253, 1.974]	0.000	1.949	[1.596, 2.379]	0.000
Sargramostim	2.178	[1.411, 3.361]	0.000	2.408	[1.578, 3.679]	0.000

GEE, generalized estimating equation.

^a Models were adjusted for potential confounders: age and sex of patient, geographic location, type of insurance, tumor characteristics (including type of cancer and existence of metastases (in HIRD^SM models)), treatment characteristics (including myelotoxicity of chemotherapy), chronic comorbidities (using the Deyo-Charlson comorbidity index), and prior neutropenic event.

^bOptumInsight multivariate analysis at the cycle level included up to the 9th cycle. This restriction was included because failure to do so resulted in GEE model estimation that failed to converge.

^cIn order to adjust for multiple comparisons within the study, comparisons were made at a statistical significance level based on the Bonferroni correction (0.05/6 = 0.008).

Both HIRD^SM and OptumInsight analyses reported a higher risk for neutropenia-related hospitalizations (narrow definition) during which cycles of SAR (OR = 6.37, 95% CI = 3.53–11.50 and OR = 4.28, 95% CI = 1.66–11.04) or FIL (OR = 3.18, 95% CI = 2.03–4.98 and OR = 1.22, 95% CI = 0.57–2.60) were used vs PEG, respectively (Table 2).

Finally, the risk of all-cause hospitalization was also higher during cycles of SAR (OR = 2.18, 95% CI = 1.41–3.36) or FIL (OR = 1.57, 95% CI = 1.25–1.97) vs PEG in the HIRD^SM analysis. This was confirmed in the OptumInsight analysis (SAR: 2.41, 95% CI = 1.58–3.68; FIL: 1.95, 95% CI = 1.60–2.60) (Table 2).

Discussion

Patients receiving myelosuppressive chemotherapy are at risk for developing FN, a major dose-limiting toxicity associated with hospitalization, morbidity, and mortality. Prophylactic use of a G/GM-CSF (FIL, PEG, and SAR) can decrease the incidence of FN. This study examined real-world effects of G/GM-CSF on hospitalization risk; where prophylaxis with PEG was associated with a lower risk of neutropenia-related (both broad and narrow definitions) and all-cause hospitalizations compared to FIL and SAR. We chose not to analyze costs in this study as other similar claims database studies31–34 have provided relative costs (including hospitalization costs) on these CSFs.

The combined HIRD^SM and OptumInsight databases provided data on 18,187 unique patients which included 58,083 cycles of CSF. The databases were similar in terms of patient characteristics, where the majority was female, the mean age ranged from 55–60 across all of the CSF groups, and the most common tumor type was breast cancer (49–62%). Both databases also showed FIL to have a significantly shorter number of mean days of prophylaxis (i.e., administrations within the first 5 days of a cycle) vs SAR. More importantly, the average days of FIL prophylaxis (5–6 days) observed in the two databases was fewer than required for recommended ANC recovery (10–11 days in registration trials)35,36. However, it is important to note that this study examined effectiveness which can be impacted by the adherence of both patients and physicians. Physicians may not test daily for ANC recovery, as is typically done in clinical trials. Consequently, laboratory monitoring of CSFs in a real-world clinical setting may not be done in accordance with label recommendations for regular blood count monitoring. Nevertheless, this study describes and provides real-world effectiveness data based on how physicians actually use CSFs in the prevention of neutropenia.

The results of the present study are different from the results of another retrospective claims database study by Heaney et al.22, who reported the risk of infection-related hospitalization was lower with patients receiving SAR vs PEG or FIL. The results of the Heaney et al. study may have been impacted by the study design which did not take into consideration the timing of CSF use within the chemotherapy course (i.e., prophylactic vs treatment); therefore the data may have included both prophylactic and treatment use when comparing the CSFs for the prevention of FN. In addition, the study also used a matching design, rather than propensity score (PS) matching, and this may have imposed a bias as it was based on only two demographic variables. The results of the present study are however consistent with the results reported by Weycker et al.25, which also pooled data from two healthcare claims databases: Thomson Reuters MarketScan Commercial Claims and Encounters and Medicare Supplemental and Coordination of Benefits Database (MarketScan database) as well as Intercontinental Marketing Services PharMetrics Patient-Centric Database (PharMetrics Database). The methods used by Weycker et al. were replicated in the current study (i.e., for each patient, every unique chemotherapy cycle was identified, along with each cycle in which FIL, PEG, or SAR was administered by the fifth day of the cycle) in order to have comparable results. The Weycker et al.25 study reported results for 77,269 cycles of FIL (n = 8286), PEG (n = 67,247), and SAR (n = 1736). Similar to the current study, Weycker et al. reported that the risk of hospitalization for neutropenia complications (narrow definition, broad definition, and all-cause) were lower with PEG prophylaxis than FIL or SAR. In addition, a large retrospective chart review37 of patient records from 99 community oncology practices in the US also validated the results of the present study by reporting a 40% increase in the odds of developing FN for patients taking FIL vs PEG (absolute rates of 6.5% and 4.7%, respectively). Furthermore, a clinical trial38 which compared the efficacies of four regimens (ciprofloxacin, G-CSF (FIL or lenograstim), PEG, or PEG plus ciprofloxacin) for primary prophylaxis of FN in breast cancer patients receiving six-to-eight cycles of neoadjuvant TAC (docetaxel/doxorubicin/cyclophosphamide) reported that PEG with or without ciprofloxacin was significantly more effective than daily G-CSF or ciprofloxacin in preventing FN (5% and 7% vs 18% and 22% of patients)38. Therefore, apart from the Heaney et al.22 study, other published analyses using multiple study designs, including community practice chart review37, prospective clinical trial38, and another large claims analysis25 all concur with the results of the present study in that PEG is associated with a lower risk of FN compared to FIL25,38 or SAR25.

The claims database captures many of the important risk factors known to be associated with the development of FN. These variables include age, sex, comorbidities, history of anemia, neutropenia, or infection, history of radiation, tumor type, and number of myelosuppressive agents. Therefore, this study, which uses the same variables as the Weycker et al.23 study, is able to control for the strongest predictors of FN.

There are other potential predictors of FN that the claims database does not contain, including stage of disease, chemotherapy dose, race, performance status, nutritional status, body surface area, lab values (bilirubin, hemoglobin, serum albumin, glomerular filtration rate, alkaline phosphatase) and concomitant medications. Stage is an important variable that the claims database cannot control for, although it can capture tumor type and secondary malignancies (identified by ICD-9 codes) as evidence for metastases. If an omitted predictor of FN is also predictive of treatment choice (PEG, FIL, or SAR) confounding may occur. The direction of the resulting bias would depend on whether this association was positive or negative. For example, if patients with higher chemotherapy doses were more likely to be put on PEG and higher chemotherapy doses are associated with greater risk of FN hospitalization, the odds ratio for FIL and SAR would be larger than the true OR.

In addition, relevant patient histories regarding such factors as chemotherapy, hospitalizations, and FN events occurring before the study period are unobservable. The impact of these limitations cannot be ascertained. Additional limitations include a lack of available information about the daily dose of FIL. Thus, days of therapy was used to represent dose. Coding for chemotherapy regimens was done based on prior studies, clinical input, and code description of chemotherapy. Given the low sample size for SAR, several of our estimates suffer from lack of precision resulting in very wide confidence intervals.

Following previous studies, it is assumed that FIL and SAR administration by day 5 of a cycle represents prophylaxis rather than treatment, as it would be unlikely to have a low ANC or white blood cell count at day 5 of a chemotherapy cycle. Although this definition has been used in other studies23, its validity has not yet been confirmed in the literature.

Furthermore, physician selection of FIL, PEG, or SAR may be influenced by differences in patient treatment history and other risk factors of FN such as chemotherapy regimen, treatment intent (palliative, curative, extend survival), age, comorbid conditions. Insurance coverage may also influence physician selection. As well, some biases may not be adequately controlled for due to selection bias for the use of PEG over FIL without having more detailed clinical information. Also, switching among FIL, PEG, and SAR during a course of chemotherapy was not assessed in this study, and patients with evidence of more than one CSF were eliminated from the analyses.

In addition, the claims algorithm that was used to identify hospitalizations as neutropenia-related was not validated against the patient’s chart with respect to the reason for the hospitalization. However, electronic medical record (EMR) data is not a reliable source for FN since the majority of patients are treated inpatient31–35, and outpatient EMR does not record these data well unless the patient was sent to the hospital from the clinic or treated outpatient at the clinic. Therefore, the lack of validation would have most likely resulted in an under-estimate of neutropenia-related hospitalizations. Finally, this study primarily focused on a large US employment-sponsored managed care population. Thus, the results of this study may not be generalized to all patients receiving SAR, FIL, or PEG in the US.

Conclusion

In this comparative effectiveness study, use of PEG was associated with a lower risk of neutropenia-related and all-cause hospitalizations compared to use of FIL or SAR.

Transparency

Declaration of funding

This study was funded by Amgen, Inc. Amgen sponsored the retrospective analysis of claims data from the HealthCore Integrated Research Database (HIRD^SM Wilmington, DE) and the OptumInsight (Eden Prairie, MN) claims database.

Declaration of financial/other relationships

H.T., J.Y., and A.K. as well as H.J.H. and L.B. have disclosed that they are employees of HealthCore, Inc. and OptumInsight, respectively, which provided consultancy for the project funded by Amgen. N.A. has declared grant funding from Genetech, as well as Amgen Speaker and Consultant services.

Acknowledgment

The authors would like to acknowledge the writing assistance of Tara Cowling, of Medlior Health Outcomes Ltd. These results were presented in part at the American Society of Clinical Oncology Annual Meeting, Chicago, IL, June 3–7, 2011.

Notes

*Study period was defined as the 1-year period prior to course start date (baseline) through end of course.

†Patients were excluded if they received their medication through pharmacy (as opposed to medical HCPCS codes), as the exact chemotherapy administration date cannot be determined, subsequently affecting their chemotherapy cycle logic based on claims.

*Like Weycker et al.²⁵, OptumInsight excluded patients with pharmacy claims for FIL and SAR from this calculation and therefore may have smaller sample sizes compared to HIRD^SM results which used pharmacy claims.

†FIL vs PEG adjusted odds ratio in OptumInsight data is not statistically significant, likely due to the low incidence of the outcome in its narrow definition. The use of the narrow definition may under-estimate the real FN incidence rate because of potential under-coding of neutropenia.