Impact of G-CSF Prophylaxis on Chemotherapy Dose-Intensity, Link Between Dose-Intensity and Survival in Patients with Metastatic Pancreatic Adenocarcinoma.

BACKGROUND: In metastatic pancreatic adenocarcinoma, few data are available on the use of granulocyte-colony stimulating factor (G-CSF) prophylaxis and its impact on dose-intensity (DI), or the link between DI and progression-free survival (PFS). This study assessed the impact of G-CSF prophylaxis on the DI received by patients and the relationship between full DI and PFS according to chemotherapy regimens. PATIENTS AND METHODS: Patients from three first-line randomized phase II clinical trials were included in this retrospective cohort. G-CSF prophylaxis groups were identified and balanced according to baseline characteristics using a propensity score. Patients were classified into 2 treatment groups (FOLFIRINOX vs FOLFIRI/nab-paclitaxel (NAB)). DI was a binary variable (full/reduced). Adverse events were defined using NCI-CTCAE v4.0.
RESULTS: Of the 498 patients, 154 (31%) were in "prophylaxis" group; 179 (36%) were treated by FOLFIRINOX and 319 (64%) by FOLFIRI/NAB. In FOLFIRINOX group, G-CSF prophylaxis was significantly associated with a higher rate of full DI (OR, 5.07; 95% CI, 1.52-16.90; P < .01) while in FOLFIRI/NAB group, it was significantly associated with a lower rate of full DI (OR, 0.23; 95% CI, 0.06-0.83; P = .03). Full DI was associated with a non-significant increase in PFS (FOLFIRINOX group: HR 0.83; 95% CI, 0.59-1.16; P = .27; FOLFIRI/NAB group: HR 0.84; 95% CI, 0.63-1.11; P = .22).
CONCLUSION: Granulocyte-colony stimulating factor prophylaxis was associated with a higher rate of full DI with FOLFIRINOX. Full DI was associated with a non-significant increase in PFS. These results need to be confirmed prospectively.

This retrospective cohort study is one of the first to evaluate the impact of G-CSF prophylaxis on the dose-intensity of chemotherapy and not only on the incidence of adverse events such as febrile neutropenia. Its results, to be confirmed prospectively, could lead to a modification in the use of these expensive molecules in prophylaxis.

Introduction

Pancreatic adenocarcinoma (PAC) is a frequent digestive cancer with almost 500 000 new cases worldwide in 2020 and is associated with a very poor prognosis (IARC data 2020). In 1997, gemcitabine in monotherapy was established as the gold standard in the treatment of advanced PAC.[1] Starting in 2011, FOLFIRINOX[2] and gemcitabine plus nab-paclitaxel (GEMNAB)[3] have shown encouraging results, with a significant increase in overall survival (OS) in patients with metastatic pancreatic adenocarcinoma (mPAC), and became the first-line standard regimens. For each of these polychemotherapy regimens, efficacy is counterbalanced by a higher rate of neutropenia.[2,3] Febrile neutropenia (FN) and neutropenia <1000/mm3 are defined as grade ≥ 3 adverse events (AEs) according to the NCI-CTCAE v4.0.[4] These complications generally lead to reduced dose intensity (DI), defined as a dose reduction and/or a cycle postponement. As reported in other cancers,[5-8] this reduction in DI could affect chemotherapy efficacy.

The prophylactic use of granulocyte-colony stimulating factor (G-CSF) reduces the occurrence and the depth of grades 3-4 neutropenia and FN.[9,10] International consensus summarizes the indications for the use of G-CSF in primary prophylaxis (PP).[11-14] The last update of the European Organisation for Research and Treatment of Cancer (EORTC)[12] recommends the prescription of G-CSF as PP for chemotherapy associated with a high probability of FN (>20%). For an average FN probability (10%-20%), prophylaxis is only recommended for patients at risk (eg risk factors: age > 65 years, advanced disease, previous FN or female). Prophylaxis is not recommended in chemotherapy with low FN probability (<10%).

Few studies have evaluated the impact of G-CSF[15-17] on chemotherapy DI in digestive oncology. Moreover, the expected rates of FN and grades 3-4 neutropenia with the most widely used chemotherapies are only known through the results of clinical trials and hospital-based series, which by definition are subject to selection bias. The FN rate in recent clinical trials[2,18] is probably underestimated because of the widespread use of G-CSF in PP or secondary prophylaxis (SP).

Using data from 3 randomized clinical trials (RCTs) (PRODIGE 35,[19] PRODIGE 37[20] and AFUGEM[21]), which tested more or less neutropenic polychemotherapy, with no recommendations in protocols regarding G-CSF use, we report here, for each regimen, practices regarding G-CSF use and the occurrence of neutropenic AEs (FN plus neutropenia grades 3-4).

We then evaluated the impact of G-CSF prophylaxis on DI and the relationship between DI and progression-free survival (PFS).

Materials and Methods

Study Population and Definition of Groups

Our study population included all patients who received at least one chemotherapy dose in one of the three mentioned RCTs, which evaluated different first-lines of chemotherapy in mPAC. The PRODIGE 35 trial evaluated FOLFIRINOX alone versus FOLFIRINOX (8 cycles) followed by simplified LV5FU2 in maintenance versus FOLFIRI3/gemcitabine alternately (every 2 months); the PRODIGE 37 trial evaluated GEMNAB/FOLFIRI3 alternately versus GEMNAB; the AFUGEM trial evaluated GEMNAB versus nab-paclitaxel + simplified LV5FU2 (detailed regimens in Supplementary Table S1). During these RCTs, prescriptions of G-CSF were prospectively registered at each cycle (type of G-CSF, date of initiation, number of injections). In the 3 RCTs, tumors were evaluated using CT-scans every 8 weeks.

Patients were classified into one of the 3 following treatment groups according to the first chemotherapy sequence (defined as the first combination of chemotherapy treatment administered to patients): FOLFIRINOX group (PRODIGE 35 arms A and B), FOLFIRI group (PRODIGE 35 arm C), and NAB group (either GEMNAB (PRODIGE 37 and AFUGEM) or LV5FU2 plus nab-paclitaxel (AFUGEM)).

Patients were also classified into 3 G-CSF prophylaxis groups. The PP group consisted of patients who received G-CSF from the first cycle and during at least 2 consecutive cycles. The SP group consisted of patients who received G-CSF from cycle n > 1 and during at least 2 consecutive cycles (also including those who received G-CSF for only 1 cycle from cycle n > 1 and who discontinued study treatment at cycle n+1). The PP and SP groups were merged into one “prophylaxis” group when assessing the impact of G-CSF prophylaxis on the DI. The “no prophylaxis” (ØP) group consisted of patients who did not receive G-CSF or were given G-CSF for curative purposes during one cycle only.

Study Objectives

The first objective of the study was to describe the use of G-CSF in PP or SP as well as rates of FN or grades 3-4 neutropenia regarding chemotherapy.

The second objective was to evaluate the impact of G-CSF prophylaxis (PP+SP) on DI received by patients treated with FOLFIRINOX or FOLFIRI/NAB. DI was considered a repeated binary variable. For one cycle, we regarded DI as full when a patient received at least 90% of all theoretical doses of each anticancer agent of the polychemotherapy regimen with no delay longer than 7 days (all delays longer than 7 days, all causes combined were considered). Otherwise, DI was considered reduced.

The last objective was to evaluate the impact of DI on PFS, defined as the time between the first day of the first chemotherapy cycle and the date of first progression (clinical or radiological) or death from any cause.

Statistical Analysis

Qualitative variables were reported with frequencies and percentages and quantitative variables with medians and ranges or interquartile ranges (IQR).

Numbers and percentages of FN and grades 3-4 neutropenia were described according to chemotherapy and G-CSF use (no G-CSF vs G-CSF). The “no G-CSF” profile consisted of ØP patients and SP patients before they received any G-CSF, and the “G-CSF” profile consisted of PP patients and SP patients after they had received G-CSF.

Propensity Score to Balance Patients’ Baseline Characteristics Between the “Prophylaxis” Group and the “No Prophylaxis” Group

Some patients’ baseline characteristics are known to be risk factors for FN and could influence G-CSF prophylaxis.[11,12] Because patients in the different treatment groups were not randomized according to G-CSF prophylaxis, their baseline characteristics may vary substantially between prophylaxis groups. These differences can lead to a biased evaluation of the effect of prophylaxis on the DI received. The propensity score,[22,23] defined here as the conditional probability of receiving G-CSF prophylaxis given the patients’ baseline characteristics, was used to balance these covariates in the “prophylaxis” and “no prophylaxis” groups, and therefore reduce this bias. A logistic regression model was built to estimate the propensity score. The model was adjusted for the baseline characteristics. A categorical balancing score variable was then derived by grouping the estimated probabilities into 5 categories, based on the quintiles of their distribution.[23]

Modeling the Relationship Between Prophylaxis and DI for Each Treatment Group

To analyze the impact of prophylaxis on DI, we modeled the probability of receiving a full DI according to prophylaxis through a mixed logistic regression model[24] since DI was considered a repeated binary variable (1 measurement at each cycle for each patient).

Because DI is influenced not only by FN and grades 3-4 neutropenia, which could be corrected by G-CSF prophylaxis, but also by other grades 3-4 toxicities, we included in the model: hematological AEs including anemia, thrombocytopenia, excluding neutropenia, and leukopenia; gastrointestinal AEs including diarrhea, nausea, vomiting, and mucositis; neuro-allergic AEs including paresthesia, neuropathy, allergic reaction, and palmoplantar erythrodysesthesia syndrome; deterioration in general health including anorexia; and liver AEs including increased ALT and AST and hyperbilirubinemia.

Fixed effects were estimated for prophylaxis and these severe toxicity variables. The model was adjusted for the categorical propensity score variable. The model allowed a random effect at both the patient level (to account for intra-patient correlated measurements of DI) and the chemotherapy cycle level (to account for intra-cycle correlated measurements of DI). Only DI for the first chemotherapy sequence was considered in this analysis.

Modeling the Relationship Between DI and PFS in Each Treatment Group

We fitted a Cox proportional hazard model with mixed effects[25] to explore the relationship between PFS and DI. The model allowed a chemotherapy cycle-specific random effect, and for the FOLFIRI/NAB treatment, a study-specific random effect. DI was treated as a time-varying covariate as it was evaluated at each successive chemotherapy cycle. The model was adjusted for patients’ baseline characteristics. DI was considered only for the first 3 months of treatment.

Modeling Kaplan-Meier Curves Between the 3 G-CSF Prophylaxis Groups and PFS or OS

For explanatory purpose in the FOLFIRINOX group, we computed and plotted the Kaplan-Meier estimations of the PFS and OS according to previously defined prophylaxis groups (PP, SP, or ØP). The associated hazard-ratios (HRs) were calculated based on a Cox proportional hazard model.

R software version 3.6.3 (R Foundation) was used for all statistical analyses. An estimated effect associated with a P-value of < .05 was considered significant.

Results

Study Population and Neutropenic Events

A total of 498 patients were included in our analysis (Fig. 1). Dose reductions per drug according to each chemotherapy regimen used in the 3 RCTs are detailed in Supplementary Table S2.

Figure 1.

Study flowchart patients’ classification by prophylaxis group. PP, primary prophylaxis; ØP, no prophylaxis; SP, secondary prophylaxis; FN, febrile neutropenia.

The “no G-CSF” group included 417 (73.0%) patients; the “G-CSF” group included 154 (27.0%) patients (Table 1). More patients in the “no G-CSF” group than in the “G-CSF” group presented at least one FN (3.4% and 0.6%, respectively, P = .13). At least one grade 3 or 4 neutropenia occurred in 21.6% of patients in the “no G-CSF” group versus 18.8% of patients in the “G-CSF” group (P = .55).

Table 1.

Rates of FN and grades 3-4 neutropenia according to “no G-CSF” and “G-CSF” groups.

	Episodes of FN/patient			Episodes of grades 3-4 neutropenia/patient
	0	1	2	0	1	2	>2
	Percentage of patients			Percentage of patients
No G-CSF (N = 417)a	96.6	3.4	0	78.4	19.2	1.7	0.7
FOLFIRINOX (N = 112)	96.4	3.6	—	70.5	27.7	1.8	—
FOLFIRI (N = 83)	100	—	—	85.5	12.1	2.4	—
NAB (N = 222)	95.5	4.5	—	79.7	17.5	1.4	1.4
G-CSF (N = 154)b	99.4	0	0.6	81.2	13.6	3.3	1.9
FOLFIRINOX (N = 113)	99.1	—	0.9	85.0	10.6	2.7	1.7
FOLFIRI (N = 11)	100	—	—	100	—	—	—
NAB (N = 30)	100	—	—	60.0	30.0	6.7	3.3

ØP and SP patients before prophylaxis.

PP patients and SP patients after prophylaxis.

FN, febrile neutropenia.

Patients’ Characteristics and Prophylaxis Groups

Patients’ baseline characteristics and treatments were balanced between the “ØP”, “PP”, and “SP” groups except for treatment, age, and absolute neutrophil count (ANC; Table 2). The proportion of patients over 65 years was higher in the “PP” group than in the “SP” or “ØP” groups. All patients had ANC > 1500/mm3 at baseline but the median ANC was higher in the “ØP” than in the “PP” and “SP” groups.

Table 2.

Patients’ baseline characteristics and treatments according to prophylaxis groups.

	ØP (N = 344)	PP (N = 81)	SP (N = 73)	PP+SP (N = 154)	Total (N = 498)
	%	%	%	%	%
Sex
Male	59.0	58.0	53.4	55.8	58.0
Female	41.0	42.0	46.6	44.2	42.0
Treatment
FOLFIRINOX	19.2	82.7	63.0	73.4	35.9
FOLFIRI	22.1	4.9	9.6	7.1	17.5
NAB	58.7	12.4	27.4	19.5	46.6
WHO PS
0	39.0	45.7	34.2	40.3	39.4
1	52.3	51.8	65.8	58.4	54.2
2	8.7	2.5	0	1.3	6.4
Age
<65 years	48.0	40.7	53.4	46.7	47.6
≥65 years	52.0	59.3	46.6	53.2	52.4
BSA
<2 m²	86.1	88.9	87.7	88.3	86.8
≥2 m²	13.9	11.1	12.3	11.7	13.2

ØP, no prophylaxis; PP, primary prophylaxis; SP, secondary prophylaxis; WHO PS, WHO performance status; BSA, body surface area; IQR, interquartile range; ANC, absolute neutrophils count.

Patients’ demographic and disease characteristics according to the 3 RCTs are summarized in Supplementary Table S3.

Patients treated with FOLFIRINOX more frequently received G-CSF (“PP”: 37%; “SP”: 26%; “ØP”: 37%) than did those treated with FOLFIRI (“PP”: 5%; “SP”: 8%; “ØP”: 87%) or NAB (“PP”: 4%; “SP”: 9%; “ØP”: 87%; Supplementary Table S4). Moreover, regarding only patients with G-CSF prophylaxis, patients in the FOLFIRINOX group more frequently received G-CSF PP than did patients in the FOLFIRI/NAB group (59% vs 34%; P < .01; see Supplementary Table S5).

Impact of G-CSF Prophylaxis on DI

All available risk factors, except albumin due to missing data, were included in the propensity score. The logistic model results used for propensity score building is presented in Supplementary Table S6. Results from the multivariate DI analysis, adjusted for the propensity score variable, are presented in Table 3.

Table 3.

Estimated OR from multivariate DI analysis (full vs reduced) adjusted for the propensity score variable.

Variables		OR [95% CI]	P-value
FOLFIRINOX group
G-CSF prophylaxis	Yes vs no	5.07 [1.52-16.90]	<.01
Hematological AEs (grades 3-4)	Yes vs no	0.52 [0.08-3.37]	.50
Gastrointestinal AEs (grades 3-4)	Yes vs no	4.72 [1.48-15.10]	.01
Neuro-allergic AEs (grades 3-4)	Yes vs no	0.67 [0.06-6.92]	.73
Deterioration in general health (grades 3-4)	Yes vs no	1.11 [0.29-4.18]	.88
Liver AEs (grades 3-4)	Yes vs no	1.32 [0.15-11.76]	.80
FOLFIRI/NAB group
G-CSF prophylaxis	Yes vs no	0.23 [0.06-0.83]	.03
Hematological AEs (grades 3-4)	Yes vs no	0.76 [0.19-3.04]	.69
Gastrointestinal AEs (grades 3-4)	Yes vs no	2.32 [0.78-6.90]	.13
Neuro-allergic AEs (grades 3-4)	Yes vs no	0.49 [0.07-3.45]	.47
Deterioration in general health (grades 3-4)	Yes vs no	0.32 [0.12-0.87]	.03
Liver AEs (grades 3-4)	Yes vs no	0.45 [0.15-1.39]	.17

OR, odds-ratio; CI, confidence interval; AEs, adverse events.

In patients treated with FOLFIRINOX with (N = 113) or without (N = 66) G-CSF prophylaxis, G-CSF prophylaxis or the occurrence of at least one grades 3-4 gastrointestinal AE was significantly associated with a greater likelihood of receiving full DI. In contrast, in patients treated with FOLFIRI/NAB with (N = 41) or without (N = 278) G-CSF prophylaxis, G-CSF prophylaxis and a grades 3-4 deterioration in general health were associated with a lower likelihood of receiving full DI.

Impact of DI on PFS

In both groups, full DI was associated with a trend toward an increase in PFS (Table 4). In the FOLFIRINOX group, a WHO performance status equal to 1 and age below 65 years were associated with a trend toward a decrease in PFS. In the FOLFIRI/NAB group, a WHO performance status of 2 was significantly associated with a decrease in PFS.

Table 4.

Cox proportional hazard mixed model results for PFS analysis.

Variables		HR [95% CI]	P-value
FOLFIRINOX group
Dose-intensity	Full vs reduced	0.83 [0.59-1.16]	.27
Age (year)	≥65 vs <65 years	0.74 [0.54-1.01]	.06
WHO PS	1 vs 0	1.33 [0.98-1.81]	.07
Sex	Male vs female	0.95 [0.69-1.30]	.73
FOLFIRI/NAB group
Dose-intensity	Full vs reduced	0.84 [0.63-1.11]	.22
Age (year)	≥65 vs <65 years	0.79 [0.60-1.06]	.11
WHO PS	1 vs 0	1.16 [0.86-1.57]	.32
	2 vs 0	2.70 [1.61-4.51]	<.01
Sex	Male vs female	1.07 [0.81-1.41]	.66

HR, hazard-ratio; CI, confidence interval; WHO PS, WHO performance status.

Kaplan-Meier Curves, Prophylaxis Types, and Survival in the FOLFIRINOX Group

In the FOLFIRINOX group (N = 179), the median PFS was 7.85 months in the SP group, 6.43 months in the PP group and 4.00 months in the ØP group. Compared with the ØP group, the PFS hazard-ratio was 0.52 [0.35-0.76] (P < .01) for SP group and 0.56 [0.39-0.80] (P < .01) for PP group. No significant difference for SP versus PP was observed (HR = 0.92; 95% CI [0.62-1.36]; P = .67; Fig. 2). The same approach was used for OS, and the results are presented in Supplementary Figure S1; Kaplan-Meier curves for prophylaxis groups (PP + SP) versus no prophylaxis group (ØP) are also presented in Supplementary Figure S1.

Figure 2.

Kaplan-Meier estimates of PFS according to prophylaxis group. Kaplan-Meier curve for PFS. PP, primary prophylaxis; ØP, no prophylaxis; SP, secondary prophylaxis.

Discussion

In our study on patients with mPAC, G-CSF prophylaxis was associated with a higher rate of full DI in the FOLFIRINOX group. In contrast, G-CSF prophylaxis was associated with a lower rate of full DI in the FOLFIRI/NAB group. Even although the 2 treatment groups we compared were not randomized according to G-CSF prophylaxis, the propensity score allowed us to build 2 well-balanced groups for our analysis. Nevertheless, we found a non-significant association between full DI and better PFS in both groups.

Studies conducted in onco-hematology,[26] breast cancer[27] or urothelial cancer[28] have highlighted the positive impact of G-CSF on DI and/or survival. A systematic review of randomized controlled trials in patients with non-Hodgkin’s malignant lymphoma treated with intensive chemotherapy showed that G-CSF PP versus no G-CSF PP was significantly associated with higher DI.[29] An observational study conducted in patients mainly with hematological malignancies (63.8%) showed that G-CSF prophylaxis (no type specified) divided by 5 in-hospital mortality from sepsis or pneumonia.[30] In a meta-analysis including results from 59 randomized G-CSF trials in non-digestive cancers,[31] G-CSF use was significantly associated with both higher DI and a reduced risk of death [RR = 0.93; 95% CI: 0.90-0.96; P < .001].

To the best of our knowledge, there are very few data[32,33] in digestive oncology on the impact of G-CSF on DI and/or survival. The major strength of our study is to be the first to bring original data on the use of G-CSF, to assess the relationship between G-CSF prophylaxis (all types) and DI received by patients with mPAC. This is of major interest considering the recent and widespread use of hematotoxic polychemotherapy in PAC.

Using RCT data, we showed that the use of G-CSF prophylaxis depended not only on the chemotherapy and age but also, probably, on practices in the different centers. We could not take into account a center effect because of the large number of centers, and the fact that some of them had included only 1 or 2 patients. We also observed that G-CSF prophylaxis was used outside EORTC recommendations[12] in some patients, probably with the intention to maintain a full DI.

Furthermore, recommendations on G-CSF use[11,12,34] are based on the results of studies dealing mainly with breast, lung, gynecological, urological, hematological, and colorectal cancers. The only regimen included in these recommendations for PAC is the combination of irinotecan and gemcitabine. This regimen is not very hematotoxic, and today, is no longer used. The risk of FN with other regimens is not clearly established, in particular with respect to FOLFIRINOX. In Conroy’s study,[2] FOLFIRINOX was associated with 5.4% FN and 45.7% grades 3-4 neutropenia. In this study, G-CSF prescription was left to the discretion of the investigator and 42.5% of patients in the FOLFIRINOX group received G-CSF prophylaxis during the trial. It was therefore impossible to have an accurate idea of the FN rate induced by the regimen in the absence of G-CSF prophylaxis. In our cohort, the rate prophylaxis reached 63%, showing that G-CSF prophylaxis is widely used with FOLFIRINOX.

We found a significant relationship between G-CSF prophylaxis and a higher proportion of full DI for the FOLFIRINOX group, while G-CSF prophylaxis was associated with a lower proportion of full DI for the FOLFIRI/NAB group. The hypothesis is that, as FOLFIRINOX is expected to be hematotoxic, G-CSF prophylaxis was mainly used as a PP to decrease the occurrence of severe white blood cell toxicity and to maintain full DI. Conversely, as FOLFIRI/NAB is not expected to be hematotoxic, G-CSF was given as SP in patients with previously reduced DI. This hypothesis is based on the observed proportions of PP and SP according to the treatment group (Supplementary Table S5).

We also found a positive but non-significant relationship between PFS and full DI regardless of the chemotherapy. This exploratory analysis was carried out on robust data from a relatively small sample of patients. The small sample size may explain the absence of significant differences; however, the results seem to be interesting with regard to the optimization of patient care, in particular at a time when FOLFIRINOX is being evaluated in numerous trials as a neo-adjuvant treatment for resectable PAC or as an induction treatment for borderline/locally advanced PAC. The results of this study are in accordance with those of the meta-analysis conducted by Lyman,[31] which showed a decrease in the relative risk of all-cause mortality in patients who received G-CSF. Given the high cost of G-CSF treatments in France,[35] as well as the AEs and the impact on quality of life that these drugs can have, a prospective study evaluating the efficacy and the benefit-risk ratio of prophylaxis on dose-intensity and survival (PFS and OS) seems essential. Indeed, if the improvements in PFS and OS with G-CSF prophylaxis in hematotoxic protocols were confirmed, with a favorable benefit-risk ratio, this would be a strong argument for recommending systematic prophylaxis with hematotoxic chemotherapy.

In a completely exploratory and non-generalizable way, we plotted Kaplan-Meier curves modeling the relationship between the different types of prophylaxis as we defined them (PP, SP, and ØP) and survival (PFS or OS). Only the FOLFIRINOX group was included in this analysis because the number of patients receiving prophylaxis (PP or SP) in the FOLFIRI/NAB group was too small (see Supplementary Data 4). This analysis did not correspond to one of our objectives and is there only to illustrate a possible relationship that could exist between the 2 variables. Despite the exploratory nature of the analysis, we can see a trend toward improved survival when patients receive G-CSF prophylaxis. These curves should be interpreted with caution, and only a RCT would show the impact of prophylaxis on survival and be able to distinguish between the impact of PP and that of SP.

Our study has several limitations. First, it was a retrospective study based on clinical trial data. There is therefore a selection bias and the results cannot be extrapolated to the general population. Due to its retrospective design, some interesting data not collected at the time of the trials, such as the neutrophil-lymphocyte ratio, which has been shown to be of interest in other studies,[36,37] could not be studied. Second, the clinical trials considered for this study evaluated heterogeneous first-line chemotherapy. This required us to limit our analysis to the first chemotherapy sequence received by the patients. Finally, in order to harmonize data and to allow their inclusion in complex models, the simplification of some variables (eg such as considering DI a binary variable) made the data less informative.

Conclusion

G-CSF prophylaxis is prescribed in almost two-thirds of patients treated with FOLFIRINOX for mPAC. However, it was mainly prescribed outside recommendations, and principally to maintain a high DI. G-CSF prophylaxis in patients treated with FOLFIRINOX was significantly associated with a higher proportion of those receiving full DI and for those treated with FOLFIRI/NAB G-CSF was associated with a lower proportion of full DI. There was a non-significant association between full DI and increased PFS whatever the chemotherapy. These data highlight the interest of analyzing G-CSF use in clinical trials. Future studies that include more patients and/or other chemotherapy regimens (FLOT, DCF, FOLFOXIRI, etc.) could confirm these data and determine the interest of and best protocols for G-CSF use in digestive oncology.

Click here for additional data file.

Click here for additional data file.