Comparative efficacy and safety of four long-acting granulocyte colony-stimulating factors for primary prophylaxis in patients with breast cancer: a prospective observational cohort study in China Authors: Zihan Guo, Jinwei Hu, Miao Mo, Mengmeng Wang, Shuohan Huang, Tongtong Liu, Jiyong Liu, Qiong Du Published: 2026 ABSTRACT SummaryBackgroundLong-acting granulocyte colony-stimulating factors (G-CSFs) are the standard of care for primary prophylaxis against chemotherapy-induced neutropenia in breast cancer, but comparative data remain limited.MethodsThis prospective observational cohort study was conducted at a single center in Shanghai, China. Patients with breast cancer scheduled for high-febrile-neutropenia-risk chemotherapy and planned for long-acting G-CSF primary prophylaxis were eligible. Patients received subcutaneous pegfilgrastim, mecapegfilgrastim, telpegfilgrastim, or efbemalenograstim alfa once per cycle per clinical practice. The primary outcome was the incidence of grade 3/4 neutropenia (leukopenia), assessed during chemotherapy cycles and at follow-up visits. Safety was assessed via bone pain-related treatment-emergent adverse events (BPR TEAEs).FindingsBetween March 25 and September 30, 2025, 407 patients were analyzed. Grade 3/4 neutropenia (leukopenia) occurred in 24.1%, varying numerically across agents (telpegfilgrastim 30.2%, mecapegfilgrastim 16.4%; overall P = 0.061). After multivariable adjustment, telpegfilgrastim was associated with a higher risk of grade 3/4 neutropenia (leukopenia) versus mecapegfilgrastim (OR = 2.46, P = 0.011). BPR TEAEs occurred in 62.4%, with similar incidence across groups (P = 0.32). Lower BPR TEAE risk was associated with ddEC chemotherapy (OR = 0.32, P = 0.0010) and postmenopausal status (OR = 0.53, P = 0.031). In exploratory analyses, administration at 24–48 h (versus <24 h) was associated with lower grade 4 neutropenia (leukopenia) (9.9% versus 27.3%, P = 0.011) and severe BPR TEAEs (0.5% versus 4.5%, P = 0.032), and a 3 mg pegfilgrastim dose was associated with comparable myeloprotection to 6 mg (29.4% versus 26.5%, P = 0.81) but fewer BPR TEAEs (29.4% versus 60.3%, P = 0.022).InterpretationAll four long-acting G-CSFs demonstrated clinical efficacy. Numerical variations and adjusted analyses indicated distinct clinical profiles, supporting tailored prophylaxis. Optimizing administration timing and considering dose reduction could improve benefit-risk balance. Larger prospective studies are needed to confirm these exploratory findings.FundingNone.Keywords: Breast cancer, Long-acting granulocyte colony-stimulating factor (G-CSF), Prospective cohort study, Chemotherapy-induced neutropenia, Bone pain Research in context Evidence before this study We searched PubMed from inception to February 28, 2025, using the terms “long-acting granulocyte colony-stimulating factor (G-CSF)”, “pegfilgrastim”, “neutropenia”, and “bone pain”. Existing evidence, mainly from registration trials and single-agent studies, supports the efficacy of long-acting G-CSFs. However, direct head-to-head comparative data to inform agent selection remain limited. While guidelines recommend administration 24–72 h after chemotherapy, administration within 24 h is frequently used in clinical practice for convenience. Comparative data on the efficacy and safety of these two strategies are still limited and inconsistent. Similarly, well-powered data on the feasibility and outcomes of dose reduction for agents such as pegfilgrastim are scarce. Added value of this study This prospective observational cohort study of 407 patients provides a direct comparison of four contemporary long-acting G-CSFs. It delineates their distinct clinical profiles to support agent selection and offers hypothesis-generating findings on the potential benefits of administering G-CSF at 24–48 h versus within 24 h. It also provides exploratory evidence suggesting that a reduced 3 mg dose of pegfilgrastim may maintain efficacy while reducing bone pain, and that routine short-acting G-CSF rescue does not significantly accelerate hematologic recovery. Finally, it identifies patient- and regimen-specific factors associated with bone pain risk, which may aid pretreatment counseling. Implications of all the available evidence The combined evidence from previous research and this study supports a trend toward more personalized primary prophylaxis with long-acting G-CSFs. The four agents demonstrate distinct clinical profiles that may inform agent selection. Exploratory findings on administration timing (24–48 h) and a reduced 3 mg pegfilgrastim dose are hypothesis-generating. Future prospective trials are needed to confirm these observations and define the optimal administration strategy for balancing efficacy and tolerability. Introduction Chemotherapy-induced neutropenia (CIN) is one of the most common adverse events and dose-limiting toxicities of cancer therapy. Severe neutropenia can progress to febrile neutropenia (FN), which significantly increases the risk of infection and may lead to hospitalization, chemotherapy delays, or dose reductions. These complications not only compromise the continuity and efficacy of anticancer therapy but also pose a life-threatening risk to patients.1 Statistics indicate that the incidence of FN can reach 20% in high-risk chemotherapy regimens, with an associated mortality rate as high as 5–10%.2,3 Granulocyte colony-stimulating factor (G-CSF) has been demonstrated to significantly reduce the incidence of FN and related hospitalizations. For high-risk regimens, guidelines recommend primary prophylaxis with G-CSF,3 which stimulates the bone marrow to enhance neutrophil production, thereby significantly reducing the incidence of FN and ensuring the timely execution of chemotherapy plans.4 In recent years, long-acting G-CSFs (e.g., pegfilgrastim, mecapegfilgrastim, telpegfilgrastim, and efbemalenograstim alfa) have gradually replaced short-acting formulations as the preferred option for FN prophylaxis in clinical practice. Their advantage lies in an extended half-life, allowing a single injection to cover the entire neutropenic risk period post-chemotherapy.5 However, several challenges persist in the real-world application of long-acting G-CSFs. First, variations in pharmacokinetic properties (e.g., half-life, molecular structure) among different agents may influence their clinical efficacy and safety profiles.6, 7, 8, 9, 10 While current evidence is largely derived from randomized controlled trials (RCTs), direct comparative studies of these different formulations in routine clinical practice remain scarc e. Second, although guidelines recommend administration within 24–72 h after chemotherapy,3 injection within 24 h (same-day) is frequently adopted in clinical practice for logistical convenience. For many patients, same-day administration reduces hospital visits and avoids a return trip to the clinic the following day. Consequently, the comparative efficacy and safety of same-day versus next-day administration remain a subject of ongoing debate, with existing studies reporting conflicting results.11, 12, 13 Third, despite prophylaxis, a subset of patients still experiences severe neutropenia. Robust evidence regarding the clinical benefit, indications, and optimal timing of rescue therapy with short-acting G-CSF in these cases is currently lacking. Furthermore, long-acting G-CSF administration is associated with several adverse effects, the most common being bone pain, myalgia, fatigue, and rash, of which bone pain is the most prevalent.14 However, its true incidence and impact may be underestimated, as data from routine clinical practice are limited. These adverse events may compromise patient adherence to prophylactic or therapeutic regimens. The incidence and risk factors for adverse events in the real world still require further exploration. This prospective observational cohort study systematically evaluates the comparative effectiveness and safety profiles of different long-acting G-CSFs in patients with breast cancer. We hypothesized that, owing to their distinct molecular structures, the four agents would differ in real-world effectiveness and safety, and that patient- and treatment-related factors would be associated with bone pain risk. Its primary objectives were to assess their effects on neutropenia incidence and G-CSF-related adverse events (e.g., bone pain, myalgia) and to explore optimal administration and rescue strategies. The ultimate goal is to identify a more appropriate G-CSF administration strategy that ensures both efficacy and safety, thereby contributing to the development of individualized prophylaxis for neutropenia. Additionally, potential risk factors associated with bone pain will be explored to support clinical risk assessment and optimize patient management. Methods Study design and participants This was a prospective observational cohort study conducted at Fudan University Shanghai Cancer Center (FUSCC) to evaluate the comparative efficacy and safety of four long-acting G-CSFs for primary prophylaxis in patients with breast cancer receiving high-risk FN chemotherapy regimens. The inclusion criteria were as follows: (1) diagnosis of breast cancer and planned to receive a chemotherapy regimen at high risk of febrile neutropenia; (2) absence of significant neutropenia prior to chemotherapy, defined as an absolute neutrophil count (ANC) ≥ 1.5 × 109/L or a white blood cell count (WBC) ≥ 3.0 × 109/L; (3) scheduled to receive long-acting G-CSF as primary prophylaxis after chemotherapy; (4) completion of at least two cycles of chemotherapy. Exclusion criteria included: (1) a history of severe infection within two weeks before the first long-acting G-CSF administration; (2) known bone marrow involvement by tumor; (3) severe dysfunction of the liver, kidney, heart, or lungs; (4) receipt of systemic bone radiotherapy within 15 days before the first dose of long-acting G-CSF. Before the first chemotherapy cycle, all baseline demographic and clinical characteristics were recorded. All patients received one of the following specified high-risk regimens: docetaxel and cyclophosphamide (TC), dose-dense doxorubicin and cyclophosphamide (ddEC), or docetaxel and carboplatin with or without trastuzumab and pertuzumab (TCb[HP]). After chemotherapy, all patients received long-acting G-CSF as primary prophylaxis. The group assignment was based on routine clinical practice. Complete blood counts were assessed on days 6, 8, and 10 of each chemotherapy cycle according to routine clinical practice. Each patient had a minimum of two follow-up visits. For patients scheduled to receive four cycles, a mid-treatment face-to-face follow-up was conducted on day 1 of cycle 3 (C3D1); for those scheduled to receive six cycles, a mid-treatment face-to-face follow-up was conducted on day 1 of cycle 4 (C4D1). All patients also received a telephone follow-up 30 ± 7 days after the last cycle. The timing of these mid-treatment visits in days from treatment initiation depended on the cycle length of the chemotherapy regimen (14 days for ddEC; 21 days for TC and TCb[HP]). During these visits, data were collected on G-CSF administration, blood counts, interventions, and bone pain-related treatment-emergent adverse events (BPR TEAEs). Patients were also encouraged to report any symptoms spontaneously between visits. Clinical endpoints The primary efficacy endpoint was the incidence of grade 3/4 neutropenia (leukopenia), an accepted surrogate marker for FN risk and a key determinant of chemotherapy dose modifications in clinical practice.3,15,16 Grade 3 was defined as 0.5 × 109/L ≤ ANC <1.0 × 109/L or 1.0 × 109/L ≤ WBC <2.0 × 109/L, and grade 4 as ANC < 0.5 × 109/L or WBC < 1.0 × 109/L, according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Secondary efficacy outcomes included the incidence of FN, differences in symptomatic management and hematologic recovery, and the occurrence of chemotherapy dose adjustments or delays due to neutropenia or FN. Safety assessments focused on bone pain and related symptoms. Endpoints included the occurrence, duration, location, and severity of bone pain, which was graded according to CTCAE v5.0 as mild (tolerable), moderate (limiting instrumental activities of daily living [ADL]), or severe (limiting self-care ADL). BPR TEAEs were captured through structured questioning at predefined clinic visits and telephone follow-ups, as well as spontaneous patient reports. Additional safety outcomes included analgesic use and response, as well as modifications to the G-CSF regimen, such as switching agents, dose reduction, or discontinuation, due to adverse events. To ensure comprehensive capture of all relevant events, BPR TEAEs were defined to include any of the following preferred terms: arthralgia, back pain, bone pain, musculoskeletal pain, myalgia, neck pain, and pain in extremity. BPR TEAEs were assessed using three measures: time to first pain report within each cycle (days from the end of long-acting G-CSF administration to the first pain report), number of cycles with reported pain (reflecting affected cycles), and duration of pain in days (from symptom onset to resolution). For analysis, patients were counted only once for overall BPR TEAEs regardless of multiple symptom reports, and the highest severity grade across all reported subdomains was assigned as the overall severity. Subdomain-specific incidences were reported separately for each pain type. Statistical analysis A pre-study sample size calculation was performed for the primary endpoint using PASS 2025 software (NCSS, LLC, Kaysville, UT, USA) (α = 0.05, power = 0.90). Based on expected proportions derived from the literature, the required evaluable sample size was estimated to be 196 patients (49 per group). Accounting for a 20% dropout rate, the target enrollment was set at 248 patients (62 per group). The study was designed as a prospective consecutive cohort, aiming to enroll all eligible patients until the target sample size was achieved. Normality was assessed using the Shapiro–Wilk test. As the data were not normally distributed, continuous variables were summarized as the median (interquartile range, IQR) and compared among groups using the Kruskal–Wallis test. Categorical variables were presented as frequencies and percentages and compared using the chi-square test. For the primary efficacy endpoint, univariable logistic regression analyses with robust standard errors were first performed to assess associations between exposures and binary outcomes. Multivariable logistic regression models with robust standard errors were then used to adjust for potential confounding. Covariates were selected based on clinical relevance and generally included age, BMI, ECOG performance status, menopausal status, comorbidity status, history of previous radiotherapy or chemotherapy, type of chemotherapy, chemotherapy regimen, and timing of long-acting G-CSF administration. For the analysis of risk factors for BPR TEAEs, univariable analyses were first performed, and covariates with P < 0.10 were then included in the multivariable model. To assess the robustness of the primary findings, inverse probability of treatment weighting (IPTW) was used to adjust for baseline confounding. To evaluate the impact of missing covariates on the IPTW analysis, random forest single imputation was performed prior to refitting the multinomial propensity score model and re-estimating the IPTW weights. All subgroup analyses were prespecified in the study protocol; no post-hoc analyses were performed. All analyses were conducted using StataMP 17 (StataCorp, College Station, TX, USA). A two-sided P-value of <0.050 was considered statistically significant. Ethics The study was conducted in accordance with the Declaration of Helsinki, and the protocol (Supplementary material) was approved by the Institutional Review Board and Ethics Committee of Fudan University Shanghai Cancer Center (Approval Number: 2506-Exp210). Written informed consent was obtained from all participants before enrollment. This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.17 Role of the funding source No funding was received for this work. Results Between March 25, 2025, and September 30, 2025, a total of 526 patients with breast cancer scheduled to receive chemotherapy regimens at high risk of FN were screened. Of these, 61 patients were excluded: 48 did not meet the inclusion criteria, eight declined study participation, four returned to local hospitals, and one had influenza A infection within two weeks. The remaining 465 patients were eligible and received clinically prescribed long-acting G-CSF as primary prophylaxis. After excluding 58 patients with missing data or loss to follow-up, 407 patients were included in the final analysis cohort (Fig. 1). Patients were categorized into four groups based on the specific long-acting G-CSF administered: pegfilgrastim (n = 88), mecapegfilgrastim (n = 110), telpegfilgrastim (n = 149), and efbemalenograstim alfa (n = 60). All patients were Asian. At baseline, no patient reported long-term analgesic use. Baseline demographic and clinical characteristics were well balanced across the groups, with no statistically significant differences. The detailed characteristics are presented in Table 1. Study profile. G-CSF: granulocyte colony-stimulating factor. Table 1Baseline characteristics.VariablePegfilgrastim (n = 88)Mecapegfilgrastim (n = 110)Telpegfilgrastim (n = 149)Efbemalenograstim alfa (n = 60)PAge, median (IQR), years49 (40–58)49 (42–57)48 (42–56)45 (39–54)0.26BMI, median (IQR), kg/m222.8 (21.1–25.1)24.0 (21.3–26.0)23.4 (21.2–25.6)23.4 (21.1–25.2)0.33Sex, n (%)0.60 Female88 (100.0)109 (99.1)147 (98.7)60 (100.0) Male01 (0.9)2 (1.3)0ECOG performance status, n (%)0.56 05 (5.7)5 (4.5)9 (6.0)6 (10.0) 183 (94.3)105 (95.5)140 (94.0)54 (90.0)Menopausal status, n (%)0.85 Premenopausal49 (55.7)65 (59.1)91 (61.1)37 (61.7) Postmenopausal39 (44.3)45 (40.9)58 (38.9)23 (38.3)History of previous radiotherapy or chemotherapy, n (%)0.99 No85 (96.6)107 (97.3)145 (97.3)58 (96.7) Yes3 (3.4)3 (2.7)4 (2.7)2 (3.3)Type of chemotherapy, n (%)0.29 Adjuvant therapy79 (89.8)97 (88.2)122 (81.9)49 (81.7) Neoadjuvant therapy9 (10.2)13 (11.8)27 (18.1)11 (18.3)Chemotherapy regimen, n (%)0.89 TC52 (59.1)59 (53.6)82 (55.0)33 (55.0) ddEC12 (13.6)12 (10.9)22 (14.8)7 (11.7) TCb(HP)24 (27.3)39 (35.5)45 (30.2)20 (33.3)Comorbidities, n (%)a0.50 No72 (81.8)81 (73.6)116 (77.9)44 (73.3) Yes16 (18.2)29 (26.4)33 (22.1)16 (26.7)IQR: interquartile range; BMI: body mass index; ECOG: Eastern Cooperative Oncology Group; TC: docetaxel/cyclophosphamide; ddEC: dose-dense doxorubicin/cyclophosphamide; TCb(HP): docetaxel/carboplatin, with or without trastuzumab and pertuzumab.aComorbidities included cardiovascular, metabolic, pulmonary, hematologic, infectious, and immune/endocrine-related conditions and other medical conditions. Baseline characteristics. IQR: interquartile range; BMI: body mass index; ECOG: Eastern Cooperative Oncology Group; TC: docetaxel/cyclophosphamide; ddEC: dose-dense doxorubicin/cyclophosphamide; TCb(HP): docetaxel/carboplatin, with or without trastuzumab and pertuzumab. Comorbidities included cardiovascular, metabolic, pulmonary, hematologic, infectious, and immune/endocrine-related conditions and other medical conditions. We compared the efficacy of the four agents in preventing neutropenia (leukopenia) and further evaluated the utility of rescue therapy. A total of 98 (24.1%) of 407 patients experienced at least one episode of grade 3/4 neutropenia (leukopenia), including 54 of grade 3 and 44 of grade 4. Among the 98 patients with grade 3/4 neutropenia (leukopenia), 89 (90.8%) developed it after the first chemotherapy cycle. Of these 89 patients, 80 (89.9%) did not have recurrent grade 3/4 neutropenia (leukopenia) after completing two cycles of treatment. The incidence of grade 3/4 neutropenia (leukopenia) was highest in the telpegfilgrastim group (45 [30.2%] of 149) and lowest in the mecapegfilgrastim group (18 [16.4%] of 110). The incidence of grade 4 neutropenia (leukopenia) was highest in the pegfilgrastim group (11 [12.5%] of 88) and lowest in the mecapegfilgrastim group (8 [7.3%] of 110). FN occurred in two patients in the telpegfilgrastim group and one patient in the pegfilgrastim group. The incidence of chemotherapy dose reduction due to neutropenia (leukopenia) was lowest in the efbemalenograstim alfa group (2 [3.3%] of 60) (Table 2). Univariate analysis indicated that the telpegfilgrastim group had a significantly higher risk of developing grade 3/4 neutropenia (leukopenia) (versus mecapegfilgrastim OR = 2.21, P = 0.011, 95% CI: 1.20–4.09). This association remained significant after multivariable adjustment (versus mecapegfilgrastim OR = 2.46, P = 0.011, 95% CI: 1.23–4.91) (Supplementary Table S1). Table 2Comparison of efficacy endpoints.Efficacy endpoints, n (%)Pegfilgrastim (n = 88)Mecapegfilgrastim (n = 110)Telpegfilgrastim (n = 149)Efbemalenograstim alfa (n = 60)PGrade 3/4 neutropenia (leukopenia)23 (26.1)18 (16.4)45 (30.2)12 (20.0)0.061 Grade 312 (13.6)10 (9.1)27 (18.1)5 (8.3)0.11 Grade 411 (12.5)8 (7.3)18 (12.1)7 (11.7)0.58FN1 (1.1)02 (1.3)00.14Chemotherapy modifications due to neutropenia or FN7 (8.0)5 (4.5)12 (8.1)2 (3.3)0.46 Dose reduction7 (8.0)5 (4.5)12 (8.1)2 (3.3)0.46 Schedule delay0000 Early discontinuation0000FN: febrile neutropenia. Comparison of efficacy endpoints. FN: febrile neutropenia. We then evaluated the efficacy of short-acting G-CSF rescue therapy in patients with grade 3/4 neutropenia (leukopenia). Of the 98 eligible patients, 42 received short-acting G-CSF rescue, whereas 55 did not; one patient was excluded due to unavailable subsequent treatment information. The day 10 hematologic recovery rate (WBC ≥ 3.0 × 109/L and ANC ≥ 1.5 × 109/L) was 95.2% (40 of 42) versus 92.7% (51 of 55) in the rescue and non-rescue groups, respectively (P = 0.61). Stratified analysis by neutropenia grade showed comparable recovery rates between the two groups in both grade 3 (95.5% [21 of 22] versus 90.6% [29 of 32], P = 0.51) and grade 4 subgroups (95.0% [19 of 20] versus 95.7% [22 of 23], P = 0.92) (Supplementary Table S2). BPR TEAEs were reported in 254 (62.4%) of 407 patients. Among these patients, 236 (92.9%) experienced pain symptoms after the first administration of the long-acting G-CSF. The most common events were bone pain (n = 182) and myalgia (n = 99); others included back pain, musculoskeletal pain, neck pain, and extremity pain. The overall incidence of any BPR TEAEs was highest in the efbemalenograstim alfa group (40 [66.7%] of 60) and lowest in the pegfilgrastim group (48 [54.5%] of 88) (Table 3). No statistically significant differences were observed in the incidence of these pain conditions among the four treatment groups in the primary analysis. However, multivariable analysis identified that mecapegfilgrastim was associated with a higher risk of moderate bone pain (versus pegfilgrastim OR = 2.91, P = 0.029, 95% CI: 1.12–7.62) (Supplementary Table S3). Differences in the occurrence of other pain conditions showed no significant variation in either univariate or multivariable analyses. The median time to first reported pain was one day after long-acting G-CSF administration; the median pain duration was three days, and the median number of cycles with reported pain was two overall (2.5 in the efbemalenograstim alfa group). The comparative efficacy and safety profiles of the four agents are summarized in Fig. 2. Table 3Incidence and severity of BPR TEAEs.Safety endpoints, n (%)Pegfilgrastim (n = 88)Mecapegfilgrastim (n = 110)Telpegfilgrastim (n = 149)Efbemalenograstim alfa (n = 60)PBPR TEAEs Any48 (54.5)68 (61.8)98 (65.8)40 (66.7)0.32 Mild37 (42.0)43 (39.1)73 (49.0)29 (48.3)0.38 Moderate11 (12.5)25 (22.7)22 (14.8)11 (18.3)0.22 Severe003 (2.0)00.16Bone pain Any35 (39.8)51 (46.4)63 (42.3)33 (55.0)0.27 Mild28 (31.8)33 (30.0)49 (32.9)23 (38.3)0.74 Moderate7 (8.0)18 (16.4)11 (7.4)10 (16.7)0.050 Severe003 (2.0)00.16Myalgia Any16 (18.2)29 (26.4)41 (27.5)13 (21.7)0.38 Mild12 (13.6)16 (14.5)29 (19.5)11 (18.3)0.60 Moderate4 (4.5)13 (11.8)12 (8.1)2 (3.3)0.14 Severe0000Back pain Any7 (8.0)5 (4.5)9 (6.0)1 (1.7)0.39 Mild4 (4.5)2 (1.8)7 (4.7)00.23 Moderate3 (3.4)3 (2.7)2 (1.3)1 (1.7)0.72 Severe0000Pain in extremity Any15 (17.0)16 (14.5)21 (14.1)4 (6.7)0.33 Mild10 (11.4)11 (10.0)16 (10.7)4 (6.7)0.80 Moderate5 (5.7)5 (4.5)5 (3.4)00.31 Severe0000Musculoskeletal pain Any3 (3.4)9 (8.2)6 (4.0)4 (6.7)0.38 Mild2 (2.3)2 (1.8)4 (2.7)3 (5.0)0.66 Moderate1 (1.1)7 (6.4)2 (1.3)1 (1.7)0.052 Severe0000Arthralgia Any2 (2.3)8 (7.3)5 (3.4)6 (10.0)0.095 Mild2 (2.3)5 (4.5)4 (2.7)5 (8.3)0.22 Moderate03 (2.7)1 (0.7)1 (1.7)0.31 Severe0000Neck pain Any01 (0.9)1 (0.7)00.75 Mild01 (0.9)1 (0.7)00.75 Moderate0000 Severe0000BPR TEAEs: bone pain-related treatment-emergent adverse events. Incidence and severity of BPR TEAEs. BPR TEAEs: bone pain-related treatment-emergent adverse events. Efficacy and safety profiles of four long-acting G-CSF agents. G-CSF: granulocyte colony-stimulating factor; BPR TEAEs: bone pain-related treatment-emergent adverse events. Of the 254 patients with BPR TEAEs, 83 (32.7%) used analgesics. Among them, 79 patients used ibuprofen (65 [82.3%] effective), two used celecoxib (both ineffective), one used acetaminophen (ineffective), and one received an unspecified analgesic and reported symptom relief. 20 (7.9%) of 254 switched to a different long-acting G-CSF brand, leading to symptom resolution in 12 (60.0%) of 20. Dose reduction was implemented in seven (2.8%) of 254, resolving symptoms in five (71.4%); and two patients (2.4%) had persistent pain, one of whom discontinued treatment. Six patients stopped long-acting G-CSF prophylaxis entirely with no symptom recurrence (Supplementary Table S4). In the univariate analysis, advanced age, comorbidities, the ddEC chemotherapy regimen, and postmenopausal status were associated with lower odds of BPR TEAEs. Variables with a univariate P-value < 0.10 were subsequently entered into a multivariable logistic regression model. After adjustment for these potential confounders, only two factors remained independently associated with a lower risk: receipt of the ddEC regimen (versus TC, OR = 0.32, P = 0.0010) and postmenopausal status (versus premenopausal, OR = 0.53, P = 0.031). No other variables, including age, comorbidities, type of long-acting G-CSF, or timing of G-CSF administration, showed significant associations in the final model (Fig. 3, Supplementary Table S5). Multivariable analysis of risk factors for BPR TEAEs. BPR TEAEs: bone pain-related treatment-emergent adverse events; OR: odds ratio; 95% CI: 95% confidence interval; TC: docetaxel/cyclophosphamide; ddEC: dose-dense doxorubicin/cyclophosphamide; TCb(HP): docetaxel/carboplatin, with or without trastuzumab and pertuzumab; G-CSF: granulocyte colony-stimulating factor. In further exploratory analyses, we assessed the impact of administration timing and pegfilgrastim dose on efficacy and safety outcomes. The majority of patients (385 [94.6%] of 407) received long-acting G-CSF within 24–48 h post-chemotherapy, while a smaller cohort (22 [5.4%] of 407) received it within 24 h. The incidence of grade 3/4 neutropenia (leukopenia) was numerically higher in the <24 h group (8 [36.4%] of 22) compared with the 24–48 h group (90 [23.4%] of 385). Administration at 24–48 h was associated with a lower incidence of grade 4 neutropenia (leukopenia) (38 [9.9%] of 385 versus 6 [27.3%] of 22 in the <24 h group, P = 0.011). Regarding safety outcomes, the overall incidence of BPR TEAEs was higher in the 24–48 h group (244 [63.4%] of 385 versus 10 [45.5%] of 22). However, the incidence of severe BPR TEAEs was lower in the 24–48 h group (2 [0.5%] of 385 versus 1 [4.5%] of 22, P = 0.032). Furthermore, patients receiving G-CSF at 24–48 h experienced a shorter median duration of BPR TEAEs compared to those in the <24 h group (three days versus seven days, P = 0.0003) (Fig. 4A, Supplementary Table S6, Supplementary Table S8). Efficacy and safety by G-CSF administration timing and pegfilgrastim dose. (A) Grade 4 neutropenia (leukopenia) and severe BPR TEAEs: 24–48 h versus <24 h post-chemotherapy. (B) Grade 3/4 neutropenia (leukopenia) and BPR TEAEs: 3 mg versus 6 mg pegfilgrastim. G-CSF: granulocyte colony-stimulating factor; BPR TEAEs: bone pain-related treatment-emergent adverse events. Among the 88 patients receiving pegfilgrastim in this study, 17 (19.3%) received the 3 mg dose, and 68 (77.3%) received the standard 6 mg dose. The remaining three patients had switched to a different long-acting G-CSF at follow-up, so they were excluded from the analysis. In this exploratory analysis, the incidence of grade 3/4 neutropenia (leukopenia) was 29.4% (5 of 17) in the 3 mg cohort and 26.5% (18 of 68) in the 6 mg cohort (P = 0.81), and the incidence of BPR TEAEs was 29.4% (5 of 17) versus 60.3% (41 of 68) (P = 0.022) (Fig. 4B). Furthermore, the median duration of BPR TEAEs was similar between the two dose groups, with no difference (P = 0.84) (Supplementary Table S7, Supplementary Table S9). In sensitivity analyses, IPTW-weighted comparisons showed no significant difference in grade 3/4 neutropenia (leukopenia) across groups (P = 0.25) or in BPR TEAEs (P = 0.51), consistent with the primary analysis. After random forest single imputation of missing covariates, IPTW-weighted results remained consistent with the complete-case analysis for both grade 3/4 neutropenia (leukopenia) (P = 0.24) and BPR TEAEs (P = 0.40) (Supplementary Table S10). Discussion This prospective observational study provides a direct head-to-head comparison of four long-acting G-CSF agents used for primary prophylaxis in breast cancer. By simultaneously evaluating these agents in a routine clinical setting, our study offers pragmatic insights into their comparative efficacy and safety profiles. Our findings confirm the overall efficacy of all agents in preventing FN. More importantly, they reveal clinically relevant descriptive differences in neutropenia (leukopenia) risk and bone pain profiles among the formulations, alongside statistically supported insights on administration strategy. These observations underscore the potential for more individualized supportive care, where efficacy, safety, and patient tolerability can be balanced. However, the primary efficacy endpoint was a surrogate marker (grade 3/4 neutropenia/leukopenia), not FN itself. With only three FN events (0.7%) observed, the study was not designed to detect differences in FN incidence between agents. Therefore, the comparative efficacy findings should be interpreted as hypothesis-generating based on surrogate hematologic endpoints, not as definitive evidence. While all four long-acting G-CSF formulations effectively prevented FN, with a very low overall incidence, clinically meaningful descriptive differences in the incidence of grade 3/4 neutropenia (leukopenia) were observed. The rate was highest with telpegfilgrastim (30.2%) and lowest with mecapegfilgrastim (16.4%). Multivariable analysis identified telpegfilgrastim as independently associated with an increased risk of grade 3/4 neutropenia (leukopenia) after adjustment for clinical factors. These differences may be partly attributed to molecular configurations and pharmacokinetic properties, although residual confounding cannot be excluded. Pegfilgrastim, the earliest long-acting G-CSF, is a PEGylated molecule produced in E. coli. Mecapegfilgrastim and telpegfilgrastim represent optimizations of the PEGylated structure with different linkers or conjugation sites.6,18 In contrast, efbemalenograstim alfa exhibits a distinct molecular design as a dimer consisting of two G-CSF-Fc molecules covalently linked via disulfide bonds between Fc fragments.19 Efbemalenograstim alfa is produced in Chinese hamster ovary (CHO) cells with glycosylations, rendering its structure more similar to endogenous G-CSF and lacking PEGylation. This characteristic may contribute to its potentially increased receptor-binding potency compared with bacterially produced agents.20 Although anti-PEG antibodies have been documented and have been hypothesized to potentially compromise the efficacy of PEGylated drugs in some patients, the clinical relevance of this mechanism in the context of G-CSF prophylaxis remains uncertain.21 In our study, efbemalenograstim alfa and mecapegfilgrastim were associated with lower incidences of grade 3/4 neutropenia (leukopenia) than the other two agents. These observations suggest that variations in the core protein source (bacterial versus mammalian) and the presence or type of PEGylation may contribute to differences in clinical outcomes; however, this hypothesis requires direct investigation in future pharmacodynamic and mechanistic studies.22 In clinical practice, short-acting G-CSF is often used as a rescue therapy for established neutropenia (leukopenia). In our analysis, no significant difference in hematologic recovery was observed between patients who received rescue therapy and those who did not. Current guidelines highlight the absence of studies addressing rescue G-CSF use after prophylactic pegfilgrastim. Our study contributes to addressing this evidence gap by providing data on this question, suggesting that routine rescue therapy does not accelerate recovery.3 However, given the observational design and the lack of detailed data on ANC kinetics, duration of severe neutropenia, and per-cycle variations, this finding should be considered exploratory rather than conclusive. These results suggest that routine use of rescue therapy may not provide a clear benefit, but potential efficacy in selected high-risk patients cannot be ruled out. From a health economics perspective, unnecessary rescue therapy also carries cost implications, including drug acquisition costs, additional clinic visits, nursing time, and patient burden. Current guidelines recommend short-acting G-CSF rescue only in selected high-risk patients. Our exploratory findings are consistent with this cautious approach, but confirmation in prospective studies is needed. Bone pain remains the most common and treatment-limiting adverse event associated with G-CSF. In our study cohort, the overall incidence of BPR TEAEs was substantial (62.4%), exceeding the 7.2–45.7% range reported in some earlier studies.23, 24, 25, 26, 27, 28 This difference is likely attributable to several factors: first, our prospective design with systematic assessment at predefined time points (face-to-face during mid-treatment and telephone follow-up 30 days after chemotherapy) using CTCAE v5.0 ensured comprehensive capture of events; second, our broad definition of BPR TEAEs encompassed multiple related pain terms, which, as noted by Bondarenko et al.,27 can substantially affect reported incidence rates across studies; and third, observational settings may yield higher reporting rates than tightly controlled trial populations. These methodological considerations should be taken into account when interpreting the observed incidence. Events were primarily mild to moderate in severity. Incidence varied among agents, with efbemalenograstim alfa showing the highest rate (66.7%) and pegfilgrastim the lowest (54.5%). The etiology of G-CSF–induced bone pain is multifactorial: G-CSF administration promotes proliferation and inhibits apoptosis of immature progenitor and mature cells in the bone marrow, potentially leading to pain. Concurrently, these cells may induce bone marrow cell expansion and recruit inflammatory cells such as monocytes and macrophages, which release cytokines including TNF-α, interleukins, and endothelin, thereby triggering pain.29,30 This mechanistic understanding underscores why NSAIDs are often effective. Importantly, our data highlight that most BPR TEAEs were well tolerated and manageable through interventions such as NSAIDs, dose reduction, or switching to an alternative long-acting G-CSF; only a minimal proportion of patients ultimately required cessation of prophylaxis.31,32 In the present study, menopausal status emerged as a protective factor for G-CSF-related bone pain in multivariable analysis, whereas chronological age lost statistical significance after adjustment. This finding likely reflects the close correlation between age and menopausal status. Previous studies reported a higher risk of bone pain among younger patients receiving pegfilgrastim, which may be attributed to a stronger bone marrow proliferative response and lower pressure pain threshold of young patients.33, 34, 35 Although postmenopausal status has been associated with chronic bone pain in the context of osteoporosis, the mechanism underlying G-CSF-related bone pain is distinct and primarily driven by acute marrow expansion rather than chronic pain caused by decreased estrogen levels.29,36 The specific mechanistic interpretation warrants further investigation. Additionally, risk factor analysis identified that patients receiving the ddEC chemotherapy regimen have a lower risk of BPR TEAEs. A recent randomized trial similarly reported that taxane-based chemotherapy was associated with higher bone pain scores than anthracycline-based regimens.37 The consistency between these independent studies strengthens the evidence that the choice of chemotherapy backbone influences G-CSF-related bone pain and may inform risk-stratified supportive care. The observed efficacy-safety profiles present a practical consideration: mecapegfilgrastim was associated with the lowest incidence of grade 3/4 neutropenia (leukopenia) but a higher rate of moderate to severe bone pain, whereas telpegfilgrastim showed the opposite trend. These profiles underscore the need to balance efficacy and tolerability when selecting a G-CSF agent in clinical practice. Beyond the choice of agent, our data indicated that the benefit-risk profile of long-acting G-CSF prophylaxis can be optimized through evidence-based administration strategies regarding timing and dose. Our exploratory analysis suggests that administration of long-acting G-CSF within 24–48 h post-chemotherapy may be associated with a lower incidence of grade 4 neutropenia (leukopenia) and severe BPR TEAEs compared with administration within 24 h, findings that align with biological rationale and the results reported by Weycker et al.11 However, given the small size of the <24 h subgroup (n = 22), these data should be interpreted as exploratory and hypothesis-generating rather than confirmatory. A recent well-powered randomized trial by Wang et al. provides more definitive evidence on this question, demonstrating that extending the administration window to 72 h post-chemotherapy significantly reduces the incidence and severity of pegfilgrastim-induced bone pain compared to 24- or 48-h administration, without compromising efficacy.38 While our data, despite its limitations, are directionally consistent with this randomized evidence in suggesting that later administration may improve tolerability, the optimal window for balancing efficacy and tolerability may extend to 72 h. Together, these findings underscore the need for further prospective studies to define the optimal administration window. Furthermore, our exploratory analysis of pegfilgrastim dosing revealed a dissociation between hematopoietic efficacy and pain induction. In this small, non-randomized exploratory analysis, a 3 mg dose was associated with a numerically lower incidence of BPR TEAEs than the standard 6 mg dose, without a statistically significant difference in grade 3/4 neutropenia. However, because of the small sample size, lack of body weight adjustment, and non-randomized allocation, these findings are hypothesis-generating only. They should not be interpreted as evidence of non-inferiority or as practice-changing. The reasons why some patients received a 3 mg dose were not systematically recorded. Potential unmeasured factors—such as lower body weight, perceived frailty, physician preference, or logistical convenience—may have influenced treatment selection and could partly explain the observed differences. Future prospective, adequately powered non-inferiority trials are needed to assess whether dose reduction of pegfilgrastim is safe and effective. It is also worth noting that the blood monitoring schedule in this study reflects routine clinical practice at our institution, where complete blood counts are assessed on days 6, 8, and 10 of each chemotherapy cycle for patients with breast cancer receiving high-risk myelosuppressive regimens. This practice is supported by pivotal randomized controlled trials of G-CSF therapies, which have consistently demonstrated that the neutrophil nadir occurs between days 7 and 9 following chemotherapy.9,18,25,26,28 By sampling on days 6, 8, and 10, this schedule comprehensively covers this critical window of neutropenia (leukopenia) risk while avoiding daily blood draws, reflecting a pragmatic balance between adequate monitoring and patient comfort in routine clinical practice. Although this monitoring frequency may differ from practices in some countries, it was applied uniformly to all enrolled patients, ensuring that the comparative analyses in this observational study remain internally valid and unbiased. Several limitations of this study should be acknowledged. First, the single-center design and the China-specific use of two agents without FDA/EMA approval limit generalizability to Western populations; thus, caution is warranted when extrapolating these findings beyond China. Second, the observational nature precludes definitive causal inference, and residual confounding may persist despite multivariable adjustment. Additionally, the choice of G-CSF in routine clinical practice may be influenced by physician preference, drug availability, institutional formularies, and economic factors, which were not controlled for in this study. Furthermore, the exploratory analyses of administration timing and dose reduction involved very small subgroups and should be interpreted as hypothesis-generating. Moreover, recall bias and observer-based CTCAE grading may affect the accuracy of pain severity and timing. Finally, the follow-up was limited to the chemotherapy period; long-term safety or effects of repeated G-CSF exposure were not evaluated. This observational study confirms the overall efficacy of long-acting G-CSFs in preventing grade 3/4 neutropenia (leukopenia), with a very low observed incidence of FN across all treatment groups. It further reveals numerical variations in efficacy and safety profiles among the four agents. Exploratory analyses suggest that administration within 24–48 h post-chemotherapy and a reduced dose of 3 mg pegfilgrastim may be associated with improved outcomes; however, these findings are hypothesis-generating and require confirmation in prospective, adequately powered studies. These observations highlight opportunities for a more personalized approach to prophylaxis, where the choice of agent, timing, and dose can be tailored based on individual patient risk factors and preferences, aiming to optimize both clinical effectiveness and quality of life during chemotherapy. Future studies incorporating biomarkers and patient-reported outcomes are warranted to refine these personalized strategies further. Contributors ZG and QD contributed to the study conception, design, and critical revision of the manuscript for important intellectual content. ZG led and was directly involved in the data collection and overall analysis. JH performed the detailed statistical analysis and data interpretation. MM provided statistical guidance and contributed to data analysis. JH, MW, and SH assisted with data collection. JH and MW drafted the initial manuscript. TL and JL accessed and verified the underlying data. QD, as the corresponding author, had full access to all data and final responsibility for the decision to submit for publication. All authors read and approved the final version of the manuscript. Data sharing statement Data from this study will be shared upon reasonable request. Individual participant data will not be shared. Declaration of interests We declare no competing interests. Acknowledgements None. Footnotes Supplementary data related to this article can be found at https://doi.org/10.1016/j.eclinm.2026.104106. Supplementary data Supplementary Tables S1–S10 Supplementary Protocol References Associated Data Supplementary Materials Supplementary Tables S1–S10 Supplementary Protocol