Granulocyte transfusion improves survival in pediatric febrile neutropenia: a 15-year cohort study

1. Study design and population

This retrospective cohort study was conducted at Siriraj Hospital, Thailand’s largest tertiary referral center, and included pediatric patients (≤18 years) with severe neutropenic infections admitted between January 2009 and July 2023. The study was approved by the Siriraj Institutional Review Board (SIRB Protocol No. 364/2566 [IRB4]).

2. Study context and methodological considerations

This cohort reflects the real-world implementation of GT in a resource-limited LMIC setting. Although most GT recipients were treated after 2018 and controls before 2018, institutional FN protocols, antimicrobial regimens, and intensive care unit practices remained consistent throughout the study period. Minor adjustments in antimicrobial agents or critical care technologies were applied based on clinical need and did not differ systematically between groups. To address potential confounders, multivariate logistic regression models were constructed adjusting for baseline ANC, MDR status, infection type, and disease status.

3. Inclusion and exclusion criteria

GT was considered based on a standardized institutional protocol implemented in 2018, which included the following indications: ANC<500 cells/mm³, receipt of active treatment aiming for remission or recovery, proven or probable invasive infection unresponsive to appropriate antimicrobial therapy, and anticipated marrow recovery in the near future. Exclusion criteria included irreversible marrow failure with no further treatment, non-neutropenic sepsis, pyrexia of unknown origin, and isolated viral infection.

Eligible patients were ≤18 years with severe neutropenia (ANC<500 cells/mm³) due to chemotherapy, hematopoietic stem cell transplantation (HSCT), or marrow failure. The infections were microbiologically documented or clinically suspected and refractory to initial therapy. Bacterial cases required ≥48 hours of persistent fever; fungal cases were defined clinically with supportive imaging/labs. Exclusion criteria included early death (<24 hours), incomplete data, ineligible diagnoses, or neutrophil dysfunction. The full criteria are in Supplementary Tables 1 and 2.

4. GT procedure and institutional protocol

GTs were administered according to a standardized institutional protocol implemented in 2018, with product preparation performed by the Department of Transfusion Medicine. Two types of product were used: (1) pooled granulocytes from leukocyte-depleted platelet concentrates, and (2) apheresis-derived single-donor units. The pooled units were prepared from 4 ABO-compatible whole blood donors screened for human immunodeficiency virus, hepatitis B virus, hepatitis C virus, and syphilis. Buffy coat remnants were separated, processed with platelet additive solution, centrifuged using an inverted technique, irradiated (25 Gy), and transfused within 24 hours. No mobilizing agents were used. For apheresis units, healthy donors underwent infectious screening and crossmatching, then received filgrastim (10 μg/kg subcutaneously) and dexamethasone (8 mg intravenously [IV]) one day earlier. Standard leukapheresis targeted a granulocyte yield ≥1×10¹⁰ cells. Units were leukoreduced, irradiated and transfused within 24 hours. Premedication included chlorpheniramine (0.1–0.15 mg/kg IV) and dexamethasone (0.1 mg/kg IV). All transfusions were administered over a 4-hour period with continuous monitoring. Additional details are provided in Supplementary Table 3.

5. Data collection

We retrospectively extracted clinical data from electronic medical records, including baseline demographics, primary diagnosis, ANC at febrile onset, infection characteristics, and treatment details. Key variables included:

(1) Infection-related variables: source of infection, pathogen type, MDR status, and fungal versus bacterial classification.

(2) Treatment-related variables: antimicrobial regimens, use of granulocyte colony-stimulating factor (G-CSF) and duration of G-CSF therapy.

(3) GT-specific variables: donor type (apheresis-derived, buffy coat-derived, or mixed), timing of GT relative to FN onset (≤3 days, 4–7 days, or >7 days), number of GTs administered, cumulative granulocyte dose (×10¹⁰ cells/kg), and premedications administered before transfusion.

(4) Adverse events: nonhemolytic febrile reactions, allergic reactions, transfusion-related acute lung injury (TRALI), cytomegalovirus (CMV) reactivation, and alloimmunization.

6. Outcomes and analysis

Primary outcome: 30-day overall survival

Secondary outcomes: (1) 90-day overall survival, (2) time to fever resolution, (3) time to FN recovery, (4) time to ANC recovery, (5) incidence and types of transfusion-related adverse events

In addition to the predefined primary and secondary outcomes, prespecified subgroup and exploratory analyses were carried out to examine the effect of GT in various clinical circumstances (e.g., transplant status, hematologic risk, type of infection, pattern of resistance, and host factors). To keep things concise, summary findings alone are included in the principal results, but detailed model specifications and results are provided in the Supplementary Material (Supplementary Tables 2, 4–10; Supplementary Figs. 1 and 2).

7. Statistical analysis

Descriptive statistics summarized baseline characteristics. Categorical variables were contrasted by chi-square or Fisher exact tests, and continuous variables by the Mann-Whitney U test. Survival endpoints were compared using Kaplan-Meier curves with log-rank tests, multivariable logistic regression, and Cox proportional hazards models adjusted for clinically significant covariates. To adjust for potential bias and confounding, propensity score matching (PSM) and inverse probability of treatment weighting (IPTW) with stabilized weights were used. Sensitivity analysis by landmark analysis, E-value estimation, and doubly robust models was done to examine robustness of results. Detailed analytic steps and supplementary results appear in Supplementary Text 1 and Supplementary Tables 6, 10–12.

8. Definitions

Severe FN was defined as a single oral temperature ≥38.3°C or a sustained temperature ≥38.0°C for at least one hour, with an ANC <500 cells/mm³ [23]. MDR infection followed CDC criteria: resistance to ≥1 agent in ≥3 antimicrobial classes. The definitions for remission status, infection source and the GT parameters are detailed in Supplementary Table 13.

1. Patient characteristics

A total of 58 pediatric patients with severe neutropenic infections were initially identified, comprising 30 in the control group and 28 in the GT group. Four patients (2 from each group) who died within 24 hours after initiating treatment were excluded, resulting in a final analytic cohort of 54 patients (GT, n=26; control, n=28).

Baseline demographic and clinical characteristics were comparable between the groups, including age, sex, body surface area, underlying hematologic diagnoses, duration and etiology of neutropenia, initial ANC, and the use of G-CSF. Most patients had high-risk hematologic disorders, and the distribution of disease stage and prior treatment exposure was similar between groups. The causes of neutropenia included intensive chemotherapy, HSCT, and primary marrow failure syndromes. G-CSF was administered in both groups with comparable baseline usage rates (Table 1).

Table 1.

Patient characteristics

2. GT characteristics, timing, and safety

Among the 26 patients who received GT, most had high-risk hematologic disorders, including acute myeloid leukemia (n=9), acute lymphoblastic leukemia (n=6), biphenotypic leukemia (n=1), and severe aplastic anemia (n=5). Three patients had undergone HSCT. The median age at transfusion was 8.7 (range, 0.8–16.9) years, and the median duration of neutropenia before GT initiation was 16 (range, 2–88) days. The median ANC at the time of transfusion was 10 cells/mm³.

Donor sources included apheresis (42.3%), pooled buffy coat donation (42.3%), or both (15.4%). The median number of transfusions per patient was 2 (range, 1–13), with a median granulocyte dose of 0.1727 × 10¹⁰ cells/kg. Transfusions were administered on consecutive days. Twenty-two patients received a single course of GT, while 4 patients received multiple courses for separate infection episodes. To maintain consistency in outcome attribution, only the final course was included in the outcome analyses.

GT was well tolerated in all cases. No serious transfusion-related adverse events (SAEs) were reported, including TRALI, anaphylaxis, hemolysis, or alloimmunization. One patient developed transient fever during infusion, which was attributed to concurrent infection. Among 15 patients monitored for CMV reactivation, 1 case (6.7%) of asymptomatic reactivation without organ involvement was observed.

The effect of GT timing on survival was explored using stratified analysis. No deaths occurred among patients receiving GT within 3 days or after 7 days of FN onset. However, a 29% 30-day mortality rate was observed among those receiving GT between days 4–7 (P=0.053), potentially reflecting indication bias, as sicker patients may have received delayed transfusions (Supplementary Tables 14 and 15).

3. Infection characteristics and G-CSF use

Infectious profiles and antimicrobial resistance patterns were evaluated and found to be comparable between the GT and control groups. In the GT group, the most common infection sites were sepsis of unknown origin (31%), disseminated infection (19%), and skin/soft tissue infection (16%). Bacterial infections accounted for 54% of cases and were equally distributed between Gram-positive and Gram-negative organisms. Fungal infections were present in 33% of GT patients, with candidemia (63%) and invasive pulmonary aspergillosis (37%) being predominant. Mixed infections were observed in 13% of cases. Notably, MDR Gram-positive pathogens (e.g., vancomycin-resistant enterococcus (VRE), MDR viridans streptococci) were detected only in the GT group. In contrast, among control patients, sepsis of unknown origin (39%) and skin/soft tissue infections (25%) were most frequent; fungal infections occurred in 38% (primarily pulmonary aspergillosis, 74%), and Gram-negative organisms accounted for 75% of bacterial infections. MDR pathogens were identified in 25% of control cases and all were MDR Gram-negative (Table 2).

Table 2.

Causative agents identified in patients during febrile neutropenia

In addition, the use of G-CSF was analyzed to assess its potential interaction with GT. The rate of G-CSF administration was similar between the GT and control groups (58% vs. 46%, respectively; P=0.408). There were no significant differences in the duration of G-CSF therapy (median, 7 days vs. 8 days; P=0.595) or the total G-CSF dose per kilogram (median, 8.33 ×10⁶/kg vs. 6.37 ×10⁶/kg; P=0.489) (Supplementary Tables 16 and 17; Supplementary Figs. 3 and 4). Although univariate analysis indicated that high-dose G-CSF might be associated with improved 30-day survival (P=0.029), this finding was not sustained in multivariate models adjusting for GT (adjusted OR, 0.42; P=0.512), and no significant interaction between GT and G-CSF dose was detected (P=0.999) (Supplementary Tables 18 and 19). This suggests that the survival benefit observed in our cohort is attributable primarily to GT rather than differences in G-CSF use.

4. Primary outcome: 30-day survival

1) Full cohort analysis (n=54)

In the full cohort of 54 pediatric patients with severe neutropenic infections, GT was significantly associated with improved 30-day survival. Multivariate logistic regression adjusting for fungal infection, MDR bacterial infection, and baseline ANC revealed that GT independently reduced the risk of 30-day mortality (odds ratio [OR], 0.105; 95% confidence interval [CI], 0.016–0.700; P=0.020). Similarly, multivariate Cox regression supported a significant survival benefit with GT (hazard ratio [HR], 0.191; 95% CI, 0.041–0.885; P=0.034) (Table 3).

Table 3.

Summary of multivariate regression analyses of granulocyte transfusion (GT) on 30 and 90-day survival outcomes

Unadjusted Kaplan-Meier analysis demonstrated higher 30-day survival among GT recipients compared to controls (76.9% vs. 50.0%, log-rank P=0.035), and this benefit remained robust in the PSM analysis. In the matched cohort (n=26 pairs), GT was associated with improved 30-day survival (92.3% vs. 65.4%) with an OR of 0.157 (95% CI, 0.022–0.706; P=0.028) and HR of 0.191 (95% CI, 0.041–0.885; P=0.034) (Supplementary Table 5–5.1). Covariate balance before and after matching is shown in Supplementary Fig. 5.

A composite outcome of death or failure to achieve recovery by day 30 was also evaluated. GT was significantly associated with improved composite outcomes (adjusted OR, 7.89; P=0.011), while fungal infection showed a borderline negative effect (P=0.050) (Table 3; Supplementary Fig. 6).

2) Subgroup analyses

Subgroup analyses confirmed that GT was independently associated with improved 30-day survival in non-HSCT individuals, whilst fungal disease and active disease predicted adverse outcomes (Supplementary Table 4). In high-risk hematologic disease, GT offered a constant survival benefit, for which PSM offered supporting proof, with Kaplan-Meier analysis showing significantly improved 30-day survival and a beneficial 90-day trend (Table 3; Fig. 1A and B; Supplementary Tables 5–5.1). The survival benefit of GT persisted in fungal and nonfungal disease and in MDR, and interaction terms were not significant (Supplementary Tables 7–9; Supplementary Fig. 1). In the HSCT subgroup, the total of 6 individuals survived to 30 days regardless of GT allocation, but numbers were insufficient to allow rigorous statistical testing (Supplementary Table 2).

Fig. 1.

Kaplan-Meier curves of 30- and 90-day survival of high-risk hematologic disorders. Kaplan-Meier survival curves comparing granulocyte transfusion (GT) versus control (no GT) among pediatric patients with high-risk hematologic disorders (n=48). (A) Thirtyday survival showed a significantly higher survival probability in the GT group versus the control group (log rank, P=0.018). (B) Ninety-day survival showed a consistent survival benefit favoring GT, although not significant (log rank, P=0.071).

5. Sensitivity and validation analyses

Multiple sensitivity analyses confirmed the strength of survival benefit of GT. A landmark analysis of 10 days demonstrated a significant trend in favor of GT but not significant at conventional levels of alpha (Supplementary Table 6; Supplementary Fig. 7), and no temporal bias was present between early and throughout subsequent treatment periods (Supplementary Fig. 2). Logistic regression adjusted for IPTW confirmed significant survival benefits for the entire cohort and for the high-risk subgroup, and E-value analysis demonstrated that vigorous unmeasured confounding would have to dominate to abolish the effect (Table 3; Supplementary Table 11; Supplementary Text 1). A doubly robust model of Cox regression using IPTW further established GT as an independent predictor of benefit for survival (Supplementary Table 12). Forest plots and receiver operating characteristic analysis demonstrated enduring benefit and resilient predictive performance (Fig. 2; Supplementary Fig. 8). Also, stratified analysis showed the benefit of GT persisted throughout and across neutropenia severity and by MDR status (Supplementary Table 10).

Fig. 2.

Forest plots of the effect of granulocyte transfusion (GT) on survival and 30-day mortality in the full cohort versus subgroup of high-risk hematologic disorders. (A) Cox proportional hazards models show the association between GT and overall survival. (B) Logistic regression models using propensity score-matched cohorts show the association between GT and 30-day mortality. For both panels, point estimates represent hazard ratios (HRs) or odds ratios (ORs), whereas horizontal bars indicate 95% confidence intervals. The vertical dashed line indicates the null effect (HR or OR=1.0).

6. Secondary outcomes

2) Clinical recovery outcomes

GT was associated with significantly faster clinical recovery compared with the control group. The median time to fever resolution was 9.5 (IQR, 5.75–19.0) days in the GT group versus 15.5 (IQR, 11.0–24.5) days in controls (P=0.003). Similarly, the median time to recovery from FN was shorter in the GT group (11.5 days; IQR, 8.0–19.5 days) compared with controls (19.5 days; IQR, 10.25–28.75 days) (P=0.050), and the median time to ANC recovery was 18 (IQR, 12.75–28.25) days in the GT group versus 23 (IQR, 15.25–46.5) days in controls (P=0.022) (Table 4).

Table 4.

Secondary clinical outcomes of GT versus control groups

These associations remained statistically significant in multivariate Cox regression analyses. GT was independently associated with faster fever resolution (adjusted HR, 2.24; 95% CI, 1.09–4.59; P=0.028), FN recovery (adjusted HR, 2.35; 95% CI, 1.17–4.73; P=0.017), and ANC recovery (adjusted HR, 2.10; 95% CI, 1.01–4.36; P=0.047) (Supplementary Table 20; Supplementary Fig. 9).

To evaluate the robustness of these findings to unmeasured confounding, E-values were calculated for each association. The point estimates ranged from 3.62 to 4.13, and the lower bounds of the confidence intervals ranged from 1.11 to 1.62 (Supplementary Table 21), suggesting that moderate unmeasured confounding would be required to explain away the observed effects. Kaplan-Meier curves for cumulative probabilities of clinical recovery are shown in Supplementary Fig. 10B–D.