Real-World Impact of Prophylactic Growth Factor Use on Timing of Febrile Neutropenia and Infection After High-Risk Chemotherapy

Background

Patients with breast cancer treated with myelosuppressive chemotherapy are at an increased risk of developing febrile neutropenia (FN), a medical emergency characterized by the occurrence of fever during a period of significant neutropenia.¹ To reduce the incidence of FN, NCCN recommends prophylactic treatment with granulocyte colony-stimulating factor (G-CSF) for adult patients with breast cancer who are at high risk of developing FN due to myelosuppressive chemotherapy.² Prior studies have investigated the effect of prophylactic G-CSF use on time to neutrophil nadir following chemotherapy.^3,4

However, little is known about the relationship between G-CSF prophylaxis and the timing of incident FN and infection within the first chemotherapy cycle. The present study assessed this relationship through analysis of real-world data among patients with breast cancer.

Methods

This study used all fee-for-service Medicare administrative claims between January 1, 2005, and December 31, 2020, to estimate the rates of primary prophylactic (PP) G-CSF use and first-cycle FN and infection among patients with breast cancer who initiated docetaxel-containing chemotherapy regimens with high FN risk. Sample patients were required to receive 1 of 3 chemotherapy regimens with a high risk of FN per NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Hematopoietic Growth Factors²: TAC (docetaxel/doxorubicin/cyclophosphamide), TC (docetaxel/cyclophosphamide), or TCH (docetaxel/carboplatin/trastuzumab). Notably, we could not reliably distinguish dose-dense AC followed by T (which requires PP G-CSF use) from AC followed by T with therapeutic or reactive use of G-CSF, so we did not include patients receiving this regimen in our analysis.

Rates of PP G-CSF use, defined as ≥1 claim for a G-CSF on or up to 3 days after the date of chemotherapy initiation (index date), were assessed overall and by regimen. Baseline characteristics were evaluated during the 6-month period immediately preceding the index date, and included patient demographics, comorbidities, and healthcare resource utilization.

To investigate the real-world protective benefit of PP G-CSFs, proportions of patients experiencing the primary outcome of FN or infection (FN/infection) and each of 2 secondary outcomes—“strictly defined” FN (sFN) and infection-related hospitalization (IRH)—in the first chemotherapy cycle and in week 1 versus beyond week 1 of the cycle were compared between patients receiving and not receiving PP G-CSF therapy. FN/Infection was defined as inpatient hospitalization with ≥1 neutropenia, infection, or fever diagnosis, based on a previously validated algorithm.⁵ sFN was defined more narrowly as inpatient hospitalization with either a primary diagnosis of neutropenia or ≥1 neutropenia diagnosis and ≥1 infection or fever diagnosis. IRH was defined as inpatient hospitalization with ≥1 diagnosis of infection or fever.

Adjusted odds ratios comparing incidence of first-cycle events in week 1 and beyond week 1 between patients with and without PP G-CSF use were estimated using multivariable logistic regression controlling for baseline differences between cohorts. Among the subset of patients experiencing each outcome, unadjusted time from chemotherapy initiation to the outcome event was also assessed.

Additional information on study methodology is provided in supplemental eMethods (available with this article at JNCCN.org).

Results

A total of 78,810 patients initiating high-risk chemotherapy for breast cancer and meeting all other selection criteria were identified (supplemental eTable 1). Of these, 79% received TC, 14% received TCH, and 7% received TAC (Table 1, supplemental eTable 2). Rates of PP G-CSF use were highest (87%) among patients receiving TAC and lowest (70%) among those receiving TCH.

Table 1.

Patient Characteristics

View Table

Baseline characteristics were generally balanced between patients receiving and not receiving PP G-CSF therapy despite several statistically significant differences (α<.05), owing partly to large patient counts in each cohort (Table 1). Patients receiving PP G-CSFs were older on average (69.1 vs 67.9 years) and experienced lower rates of hospitalization during the 6-month baseline period (25% vs 30%).

Unadjusted proportions of patients with first-cycle FN/infection ranged from approximately 5% among patients receiving TCH and PP G-CSFs to >19% among patients receiving TAC without PP G-CSFs (supplemental eTable 3A). Within each regimen subgroup, patients receiving PP G-CSFs were less likely to experience first-cycle FN/infection compared with those not receiving PP G-CSFs. Similar trends were observed for first-cycle sFN and IRH (supplemental eTable 3B, C).

Among patients receiving PP G-CSFs, unadjusted and adjusted odds of FN/infection, sFN, and IRH in week 1 were higher for TAC and TC and lower for TCH compared with patients not receiving PP G-CSFs (Figure 1, supplemental eTables 3A and 4A). In contrast, odds of all 3 outcome events beyond week 1 of cycle 1 among patients receiving PP G-CSFs were less than half compared with patients not receiving PP G-CSFs. Results were similar in sensitivity analyses assessing effects of additional patient selection criteria on unadjusted rates of FN/infection (supplemental eTable 5A–C).

Adjusted^a proportions of patients experiencing FN/infection^b in week 1 and beyond week 1 of the first chemotherapy cycle^c, by PP G-CSF status^d.

Abbreviations: aOR, adjusted odds ratio; FN, febrile neutropenia; G-CSF, granulocyte colony-stimulating factor; PP, primary prophylactic; TAC, docetaxel/doxorubicin/cyclophosphamide; TC, docetaxel/cyclophosphamide; TCH, docetaxel/carboplatin/trastuzumab.

^aAdjusted proportions and aORs were estimated using multivariable logistic regression controlling for observed differences in patient characteristics at baseline, cycle duration (models with outcome measure evaluated beyond week 1 of cycle 1 only), and first-cycle PP G-CSF use. aORs reflect regression-estimated associations between PP G-CSF use and the given outcome of interest.

^bFN/infection was defined as inpatient hospitalization with an associated diagnosis code for neutropenia, infection, or fever.

^cThe first chemotherapy cycle was defined individually for each patient as the period beginning on the index date and ending the earlier of the day before the first observed chemotherapy administration occurring at least 13 days later, if applicable, and day 27 post index date.

^dPP G-CSF use was defined as G-CSF use on or up to 3 days after index date.

Citation: Journal of the National Comprehensive Cancer Network 21, 9; 10.6004/jnccn.2023.7044

View Full Size

Adjusted^a proportions of patients experiencing FN/infection^b in week 1 and beyond week 1 of the first chemotherapy cycle^c, by PP G-CSF status^d.

Abbreviations: aOR, adjusted odds ratio; FN, febrile neutropenia; G-CSF, granulocyte colony-stimulating factor; PP, primary prophylactic; TAC, docetaxel/doxorubicin/cyclophosphamide; TC, docetaxel/cyclophosphamide; TCH, docetaxel/carboplatin/trastuzumab.

^aAdjusted proportions and aORs were estimated using multivariable logistic regression controlling for observed differences in patient characteristics at baseline, cycle duration (models with outcome measure evaluated beyond week 1 of cycle 1 only), and first-cycle PP G-CSF use. aORs reflect regression-estimated associations between PP G-CSF use and the given outcome of interest.

^bFN/infection was defined as inpatient hospitalization with an associated diagnosis code for neutropenia, infection, or fever.

^cThe first chemotherapy cycle was defined individually for each patient as the period beginning on the index date and ending the earlier of the day before the first observed chemotherapy administration occurring at least 13 days later, if applicable, and day 27 post index date.

^dPP G-CSF use was defined as G-CSF use on or up to 3 days after index date.

Citation: Journal of the National Comprehensive Cancer Network 21, 9; 10.6004/jnccn.2023.7044

Adjusted^a proportions of patients experiencing FN/infection^b in week 1 and beyond week 1 of the first chemotherapy cycle^c, by PP G-CSF status^d.

Abbreviations: aOR, adjusted odds ratio; FN, febrile neutropenia; G-CSF, granulocyte colony-stimulating factor; PP, primary prophylactic; TAC, docetaxel/doxorubicin/cyclophosphamide; TC, docetaxel/cyclophosphamide; TCH, docetaxel/carboplatin/trastuzumab.

^aAdjusted proportions and aORs were estimated using multivariable logistic regression controlling for observed differences in patient characteristics at baseline, cycle duration (models with outcome measure evaluated beyond week 1 of cycle 1 only), and first-cycle PP G-CSF use. aORs reflect regression-estimated associations between PP G-CSF use and the given outcome of interest.

^bFN/infection was defined as inpatient hospitalization with an associated diagnosis code for neutropenia, infection, or fever.

^cThe first chemotherapy cycle was defined individually for each patient as the period beginning on the index date and ending the earlier of the day before the first observed chemotherapy administration occurring at least 13 days later, if applicable, and day 27 post index date.

^dPP G-CSF use was defined as G-CSF use on or up to 3 days after index date.

Citation: Journal of the National Comprehensive Cancer Network 21, 9; 10.6004/jnccn.2023.7044

Among patients experiencing a given outcome event during the first chemotherapy cycle, time from chemotherapy initiation to the event was more heavily right-skewed for patients receiving PP G-CSFs relative to those not receiving PP G-CSFs (Figure 2, supplemental eFigures 1 and 2). Nearly half (49%) of those treated with PP G-CSFs who developed FN/infection in cycle 1 experienced FN/infection within the first 6 days following chemotherapy initiation compared with only 18% of those not treated with PP G-CSFs who experienced FN/infection. Analogous proportions were 61% and 15% for sFN and 47% and 17% for IRH.

Time to first-cycle^a FN/infection^b among patients experiencing first-cycle FN/infection, by PP G-CSF status^c,d,e.

Abbreviations: FN, febrile neutropenia; G-CSF, granulocyte colony-stimulating factor; PP, primary prophylactic; pts, patients; sFN, febrile neutropenia under strict definition.

^aThe first chemotherapy cycle was defined individually for each patient as the period beginning on the index date and ending the earlier of the day before the first observed chemotherapy administration occurring at least 13 days later, if applicable, and day 27 post index date.

^bFN/Infection was defined as inpatient hospitalization with an associated diagnosis code for neutropenia, infection, or fever.

^cPP G-CSF use was defined as G-CSF use on or up to 3 days after the index date.

^dWilcoxon rank sum tests were used to compare time to each event among those experiencing the given event between patients receiving and not receiving PP G-CSFs.

^eTime intervals were selected to maximize the level of detail with necessitating suppression of data representing <11 patients per cell.

Citation: Journal of the National Comprehensive Cancer Network 21, 9; 10.6004/jnccn.2023.7044

View Full Size

Time to first-cycle^a FN/infection^b among patients experiencing first-cycle FN/infection, by PP G-CSF status^c,d,e.

Abbreviations: FN, febrile neutropenia; G-CSF, granulocyte colony-stimulating factor; PP, primary prophylactic; pts, patients; sFN, febrile neutropenia under strict definition.

^aThe first chemotherapy cycle was defined individually for each patient as the period beginning on the index date and ending the earlier of the day before the first observed chemotherapy administration occurring at least 13 days later, if applicable, and day 27 post index date.

^bFN/Infection was defined as inpatient hospitalization with an associated diagnosis code for neutropenia, infection, or fever.

^cPP G-CSF use was defined as G-CSF use on or up to 3 days after the index date.

^dWilcoxon rank sum tests were used to compare time to each event among those experiencing the given event between patients receiving and not receiving PP G-CSFs.

^eTime intervals were selected to maximize the level of detail with necessitating suppression of data representing <11 patients per cell.

Citation: Journal of the National Comprehensive Cancer Network 21, 9; 10.6004/jnccn.2023.7044

Time to first-cycle^a FN/infection^b among patients experiencing first-cycle FN/infection, by PP G-CSF status^c,d,e.

Abbreviations: FN, febrile neutropenia; G-CSF, granulocyte colony-stimulating factor; PP, primary prophylactic; pts, patients; sFN, febrile neutropenia under strict definition.

^aThe first chemotherapy cycle was defined individually for each patient as the period beginning on the index date and ending the earlier of the day before the first observed chemotherapy administration occurring at least 13 days later, if applicable, and day 27 post index date.

^bFN/Infection was defined as inpatient hospitalization with an associated diagnosis code for neutropenia, infection, or fever.

^cPP G-CSF use was defined as G-CSF use on or up to 3 days after the index date.

^dWilcoxon rank sum tests were used to compare time to each event among those experiencing the given event between patients receiving and not receiving PP G-CSFs.

^eTime intervals were selected to maximize the level of detail with necessitating suppression of data representing <11 patients per cell.

Citation: Journal of the National Comprehensive Cancer Network 21, 9; 10.6004/jnccn.2023.7044

Discussion

To our knowledge, this is the first analysis of real-world data to assess both incidence and timing of FN or infection within the first chemotherapy cycle for patients receiving moderate (ie, FN risk 10%–20%) or high-risk (FN risk >20%) myelosuppressive chemotherapy for breast cancer based on NCCN chemotherapy risk classifications.² Findings from our analysis suggest that these patients remain at risk for developing FN/infection, sFN, and IRH in the first week after chemotherapy initiation, regardless of whether prophylactic G-CSF treatment is administered. Although prior research has not evaluated the timing of incident FN/infection during the first cycle, incidence proportions of first-cycle FN/infection in our sample are generally consistent with prior real-world estimates.^6–11 Similarly, the proportion of patients in our study who received G-CSF prophylaxis is consistent with findings from previous real-world research.^6,8,11–14

Current NCCN Guidelines recommend PP G-CSF for patients receiving adjuvant chemotherapy with TC based on findings of randomized clinical trials, including a Japanese study conducted in 2011 through 2012 showing FN incidence after TC without PP G-CSF as high as 68.8%.¹⁵ The NCCN Guidelines recommendation for TC as a possible regimen for treating breast cancer contains a footnote stating, “Risk for FN has been reported variably as intermediate risk (10%–20%) or high risk (>20%) depending on the study.” As an average 50% to 60% of all FN events occur in the first chemotherapy cycle in patients with solid tumors, our findings of 10% to 12% first-cycle FN risk in patients not receiving PP G-CSFs and TC chemotherapy are consistent with an approximate ≥20% FN risk.¹⁶ We also confirm that TAC chemotherapy is a high–FN risk regimen.

FN risk in the first week after adjuvant chemotherapy remains an unmet clinical need. Recent data suggest that antibiotic use (including for prophylaxis) may be detrimental and prophylactic antibiotics should be used with caution.¹⁷ Data supporting the use of currently available agents that mitigate myelotoxicity in adjuvant breast cancer treatment with mechanisms of action different from G-CSF agents, including plinabulin and trilaciclib, are not available.^18–20

Our study has several limitations. First, it relies on real-world administrative claims data to assess the timing of FN/infection following chemotherapy initiation. Prior validation notwithstanding, methods to identify FN using administrative claims can only identify FN that is confirmed based on a medical encounter. Therefore, rates of FN/infection reported here may underestimate the true incidence of FN/infection, because patients may not always seek medical attention resulting in detectable claims. Similarly, because a specific diagnosis code for FN does not currently exist, our primary study outcome required algorithmic identification of FN/infection based on diagnoses of neutropenia, infection, or fever during hospitalization. However, in addition to relying on a commonly used algorithm for FN/infection identification,²¹ our findings were robust with regard to an alternative strict definition of FN that we developed. Additionally, although our study used chemotherapy regimen as the sole determinant of FN risk, other patient-specific factors may also inform risk. To minimize the degree to which study findings were driven by physician selection bias in identifying candidates for G-CSF prophylaxis, we conducted multivariable regression analyses with covariates for select available patient characteristics, including baseline healthcare resource use and preexisting comorbidities. Our findings remained consistent after controlling for observed differences between cohorts at baseline. Nevertheless, residual confounding by chemotherapy dosing and other patient characteristics undocumented in administrative claims may remain. We did not specifically exclude patients recently postoperative, though current oncology practice defers chemotherapy initiation until wounds are clearly healed. Because of the limitations of claims data, we could not exclude dose reductions or other chemotherapy modifications, prophylactic antibiotic use, or other nonreimbursable measures used to reduce FN in the first and subsequent cycles, as was done in other studies on this topic.^22,23 Finally, patients receiving dose-dense regimens were omitted from this study due to designed-in G-CSF use.

Conclusions

The present study is the first to assess both incidence and timing of FN and infection within the first chemotherapy cycle. Patients with breast cancer have a risk of FN and infection early in the first chemotherapy cycle despite prophylactic treatment with G-CSFs.