In this retrospective United States claims analysis, use of prophylactic pegfilgrastim was associated with a decreased risk of neutropenia-related hospitalization (narrow definition: OR = 0.43, 95% CI: 0.16–1.13; broad definition: OR = 0.38, 95% CI: 0.24–0.59) and all-cause hospitalization (OR = 0.50, 95% CI: 0.35–0.72) when compared to that seen with the use of prophylactic filgrastim. Patients who received filgrastim prophylaxis in this study had a mean 4.8 days of prophylaxis, which is much shorter than the duration which has been demonstrated to be non-inferior to pegfilgrastim prophylaxis in clinical trials [14–18]. These results are consistent with previous findings that hospitalization risk was more significantly reduced by pegfilgrastim prophylaxis than by filgrastim prophylaxis in clinical practice [20–23]. Key findings regarding utilization included an increase in neutropenia-related ambulatory visits and hospitalizations with filgrastim prophylaxis compared to pegfilgrastim prophylaxis. In addition, mean per-cycle neutropenia-related costs were greater with filgrastim prophylaxis than with pegfilgrastim prophylaxis.
In a previous study that used a large managed care claims database in the United States, Weycker et al. performed a retrospective cohort analysis of cancer patients who received filgrastim or pegfilgrastim during their first course of chemotherapy (2003–2005) . Both all-cause (OR = 0.73, 95% CI: 0.59–0.91) and neutropenia-related (narrow definition: OR = 0.64, 95% CI: 0.41–0.99; broad definition: OR = 0.69, 95% CI: 0.52–0.92) hospitalization risk was lower with pegfilgrastim prophylaxis compared to filgrastim prophylaxis. Two subsequent claims studies reported similar results [22, 23]. The consistent finding that filgrastim prophylaxis is associated with higher hospital risks than pegfilgrastim prophylaxis may be related to the less than optimal use of filgrastim in clinical practice. Specifically, while in the comparative clinical trials, filgrastim was given for an average of 10–11 days [14–18]; in the clinic, filgrastim is often given for far fewer days, resulting in suboptimal prevention of neutropenia and thus greater rates of hospitalization for neutropenic complications [19–23].
One distinct aspect of this study is that comparative filgrastim and pegfilgrastim costs and resource utilization data were reported, including details such as the number of hospitalizations, ambulatory and ER visits, and per-cycle costs. Comparative studies assessing cost and resource utilization of filgrastim and pegfilgrastim in the United States are limited. A retrospective single-time-point survey conducted by Fortner et al. assessed the human resource costs required for administering filgrastim or pegfilgrastim . They concluded that a single administration with filgrastim or pegfilgrastim had equivalent human resource costs, but because of the greater number of visits required with filgrastim, the total time and human resource cost with filgrastim (14.8 hours and $364.66) in a 21-day chemotherapy cycle were more than those with pegfilgrastim (2.4 hours and $57.30).
In this study utilizing a large claims database, all-cause costs were roughly equivalent for filgrastim and pegfilgrastim on a per-cycle basis, with more spent on hospitalizations with filgrastim and more spent on ambulatory care with pegfilgrastim. It is interesting to note that although pegfilgrastim cycles had greater costs for ambulatory care, there were proportionally more ambulatory visits for filgrastim cycles. The increased all-cause per-cycle cost of pegfilgrastim ambulatory care may reflect the greater drug costs of pegfilgrastim compared to filgrastim, especially when fewer than the recommended number of filgrastim doses were administered. Neutropenia-related costs were greater for filgrastim than pegfilgrastim because of the greater costs of both inpatient and ambulatory care during filgrastim cycles. Overall, these data suggest that the greater drug costs with pegfilgrastim are offset by decreased hospitalization costs.
There are several sources of bias and limitations inherent in the study design that could influence interpretation of these results. As this was a database of a large employment-sponsored managed care population, patients would most likely be in the 18–64 year age range. Thus, the effects of G-CSF on outcomes in the population of patients aged 65 years or above were not fully captured. Additionally, the data are dependent on the accuracy of claims coding and hence contain any errors or omissions that occurred during that coding. By using the broad criteria for neutropenia, as well as all-cause utilization, we were able to more accurately provide upper limits for our estimates. It should be noted that this study compared risk between treatment cohorts. The potential under-coding and mistakes in coding of febrile neutropenia are unlikely to be associated with G-CSF selection and thus do not affect the estimates of interest (ie, ORs). Sample size is another issue that affects the statistical power of our estimates, as there were fewer than 400 cycles with prophylactic filgrastim use. Another source of bias is the assumption that G-CSF administration by day 5 of a cycle represents prophylaxis rather than treatment. Although this definition has been used in other studies [19, 21], its validity has not yet been confirmed in the literature. Thus, it is uncertain whether earlier or later onset of administration may represent prophylaxis or treatment in a clinical setting. Likewise, our categorization of certain cycles as containing highly myelosuppressive chemotherapy based on the presence of individual agents used in that cycle may not adequately capture the various factors that affect the myelosuppressive effects of a chemotherapy regimen, such as combination chemotherapy and doses of specific agents. Furthermore, potential differences across health plans covered in the study sample were not adjusted for comparison of costs between filgrastim and pegfilgrastim cycles.
It is unclear whether this study adequately captured the various known patient, disease, and treatment characteristics that are risk factors for developing febrile neutropenia [6, 12]. The claims database contained several of these, such as age, sex, comorbidities, recent history of anemia, history of radiation, tumor type, and number of myelosuppressive agents. To reduce the effect of possible selection bias, data were adjusted for those covariates in the GEE model. However, the claims database did not include other potential predictors of febrile neutropenia, such as treatment intent, disease stage, chemotherapy dose, previous febrile neutropenia events, laboratory values, and concomitant medications. Some of these factors could influence selection of G-CSF as either filgrastim or pegfilgrastim. Thus the study results may still be confounded by possible differences in those unobserved characteristics between filgrastim and pegfilgrastim groups.
The use of per-cycle analyses for utilization and costs has the disadvantage of not capturing costs associated with cycles in which G-CSF was not administered. However, per-patient analyses would not allow us to temporally associate use of either G-CSF with utilization. For instance, in a per-patient analysis, a patient hospitalized in cycle 1 who subsequently received G-CSF in cycle 2 would have those two events associated, when clearly the hospitalization was independent of any effects from G-CSF. Another disadvantage of per-patient analyses is that while on average, each patient who received filgrastim received it for a mean of 2.3 cycles, while each patient who received pegfilgrastim received it for a mean of 3.5 cycles. The accompanying drug costs would make it difficult to discern any possible effect of either medication on all-cause medical costs.
In conclusion, the results of this claims analysis indicate that prophylactic use of filgrastim as compared with pegfilgrastim is associated with an increased risk of hospitalization from all causes as well as neutropenia-related causes. Results from this analysis and others indicate that other factors influence the risk of hospitalization, including comorbidities, history of anemia, age, and presence of metastatic disease [2, 6, 11, 12, 29]. Future studies that explore the role of these characteristics will help further clarify the various factors that lead to febrile neutropenia and associated complications.