Febrile neutropenia (FN) is a potentially life-threatening side effect of myelosuppressive chemotherapy; therefore, understanding the comparative effectiveness of colony-stimulating factors (CSFs) against FN is important for clinical practice. In a recent issue of Cancer, Heaney et al 1 presented a retrospective study suggesting that the risk of infection-related hospitalization may be lower for cancer chemotherapy patients who receive the CSF sargramostim (SAR) versus pegfilgrastim (PEG) or filgrastim (FIL). We believe serious design flaws in this study call into question their findings, and suggest that their results may not reflect the true comparative effectiveness of CSF agents.
For example, we believe the authors did not appropriately consider the timing of CSF use in a chemotherapy course (first vs later cycle) or the reason for CSF use (prophylaxis vs treatment), both of which may be associated with hospitalization risk. Systematic differences in these factors among CSF groups could confound results. The authors also defined follow-up differently across CSFs, which may have biased ascertainment of infection-related hospitalizations. For SAR and FIL, follow-up extended from the day of the initial administration to final dose (occurring ≤28 days from the previous administration). For PEG, follow-up included an additional 19 days after the final dose. Hospitalizations after final SAR or FIL administration were thus ignored, but counted (up to 19 days) after the final PEG administration. Moreover, the external validity of these findings was likely compromised by the use of a matched-cohort design (n = 990 SAR/FIL pairs, and n = 982 SAR/PEG pairs) that excluded data on the large majority of patients receiving CSF agents in clinical practice. 1
In summary, we believe the findings of the study by Heaney et al should be interpreted with caution and that research employing a more appropriate study design 2 is warranted for examining the comparative effectiveness of CSF agents.