Despite antibiotics, antifungals and haematopoietic growth factors, infections remain a major threat to neutropenic patients. To determine the role of granulocyte transfusions (GTs) in anti-infective therapy during neutropenia, GT administration was randomized in 74 adults with haematological or malignant diseases, febrile neutropenia and pulmonary or soft-tissue infiltrates after conventional or high-dose chemotherapy, a majority of them after allo-SCT (n=39). Neutrophil reconstitution was equal in the treatment and control arm. GT toxicity was minimal. The probability of 28-day survival after randomization was >80% in both groups, and no effect of GT on survival until day 100 could be detected in patients with fungal (n=55), bacterial or unknown infection (n=17) and various levels of neutropenia (ANC <500 vs >500 × 106/l). These findings can be attributed primarily to procedural obstacles, such as long delay from randomization to first GT, low cell content and slow sequence of GT, difficulties in randomizing a safe and potentially life-saving treatment in severely endangered individuals, and a large proportion of rapidly recovering patients in both arms. The requirement of another trial in a more specific patient population with daily transfusions of sufficient numbers of granulocytes to support or refute the empirically acknowledged benefits of GT is discussed.
Granulocyte transfusions (GTs) from G-CSF-stimulated donors have been shown to increase the absolute neutrophil count (ANC) before expected haematopoietic regeneration in neutropenic patients after chemotherapy or haematopoietic SCT.1, 2, 3, 4, 5 Thus, GT offers a therapeutic option as treatment adjunct along with antimicrobial agents and growth factors to improve the clinical outcome of patients with severe infections in neutropenia.6, 7, 8, 9, 10, 11, 12, 13, 14, 15 However, published studies rely on clinical observations of individual cases or series (highest evidence level II-2 or C), and no valid statistical comparison of this treatment method with control groups has been performed.16, 17 This prospective phase III study aims to compare the efficacy, safety and toxicity of GT in a randomized setting with standard antimicrobial treatment in 74 patients with febrile neutropenia and invasive, life-threatening infections, and to evaluate the possible alteration of incidence and severity of GVHD in 39 allografted patients due to allo-immunization.
Patients and methods
Between 1999 and 2005, seventy-four patients were randomized in 79 infectious episodes in neutropenia to receive antimicrobial treatment according to local standards (broad spectrum antibiotics, antifungals in case of suspected mycosis in a concordant fashion throughout the centres) either with or without administration of GT (3 times per week; Table 1). All patients received G-CSF (30–48 MIU, daily s.c.) continuously during neutropenia and also concomitantly with GT treatment. The mean age was 47 years (range, 14–75 years); the median follow-up time was 401 days. The study was approved by local ethic committees and conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. Informed consent was obtained from patients and volunteer donors.
(i) Haematological malignancy, acquired marrow aplasia or solid tumour, (ii) febrile neutropenia (ANC <0.5 × 106/l and anticipated duration of aplasia >5 more days) after conventional chemotherapy or haematopoietic SCT, and (iiia) pulmonary infiltrates (except exclusive bronchoalveolar lavage-confirmed detection of virus and interstitial pneumonia) or soft tissue infiltration (>5 cm diameter) or (iiib) the history of proven invasive fungal infection according to European Organisation for the Research and Treatment of Cancer and the Mycosis Study Group (EORTC/MSG)-criteria and anticipated duration of neutropenia >10 days. In accordance with an interim study board meeting, a minor proportion of patients was included and randomized in an infectious episode after high-dose chemotherapy already before reaching the nadir of aplasia (with falling levels of ANC, but still above 500 × 106/l; as indicated in ‘Results’ and Figure 1); these patients were excluded from the calculation of ANC increment and ANC reconstitution after GT.
Adult respiratory distress syndrome,
septic shock or
participation in another investigational drug study.
Survival on day 28 after randomization,
successful treatment of the infection, neutrophil increment, adverse effects, incidence of acute and chronic GVHD.
Exclusion of viral infections (HIV1/2, HCV, CMV, HbsAg), stimulation with G-CSF (10–17 h (median: 12 h) before apheresis; 5 μg/kg/day=300–480 μg filgrastim (Neupogen, Amgen, Munich, Germany) or lenograstim (Granocyte, Chugai, Frankfurt, Germany)), leukapheresis with a continous-flow blood cell separator (Spectra, Cobe or CS3000plus, Baxter, Unterschleissheim, Germany) of 7.5 l of blood (range 6–8 l) over 135 min (range 117–180 min) and irradiation of GT (30 Gy) were performed as published.4, 18 GT was performed within 6 h of apheresis with premedication of prednisolone (1 mg/kg i.v.), antihistamines and antipyretics. The minimum recommended cell content per concentrate was 3 × 108 neutrophils per kg bodyweight.4 GTs were scheduled every second day. In many centres, no GT was performed during weekends for logistic reasons in the departments of transfusion medicine. Administration of GT was discontinued after leukocyte regeneration (that is, white blood cells stably >1.0 × 109/l 48 h after the previous GT).
Survival on day 28 after randomization was chosen as end point indicating successful anti-infectious therapy, because other studied parameters such as duration of fever, elevated C-reactive protein, arterial hypotension and other symptoms of severe infections proved too variable for statistical analyses. The probability to survive 28 days after randomization was estimated to be 60%. The increased survival probability after GT was predicted to be 80% (power). The risk to misinterpret the failure of GT (a-error) is 5%. Considering a drop-out proportion of 10%, the sample size calculation thus resulted in 2 × 80 patients. The comparison between different subgroups was done with the log-rank test. A P-value of <0.05 was considered significant.
Results and discussion
Ten centres participated in the trial; however, only five recruited patients (n=74). This corresponds to <50% of the expected sample size. The study was closed prematurely due to a dramatic decrease in the recruitment rate (from 15 in 2001 to 2 in 2005). Discussion during study committee meetings revealed the following main obstacles for participation: (i) patients' and physicians' refusal to randomize in a life-threatening situation, especially if a potentially life-saving strategy (GT) was available, (ii) lack of available donors, (iii) availability of new more effective antimicrobial drugs, for example, linezolid, carbopenems and new antifungal agents such as caspofungin, voriconazole and posaconazole.
Patient characteristics were comparable between the randomized cohorts (Table 1). Administration of GT was safe, toxicity minimal; only one reversible WHO-grade III pulmonary event (dyspnoea at rest) and one grade III nausea (treatment-requiring vomiting) were observed (Table 2). There was no difference in the incidence and severity of GVHD when intervention and control arm were compared, opposing the hypothesis that allo-immunization through third-party neutrophil surface epitopes might yield an increased rate of GVHD after GT. Most episodes included were definite or probable pulmonary or severe soft tissue fungal infections as defined by the international EORTC-MSG criteria (n=55), and all fulfilled the inclusion criteria. The minority suffered from a bacterial infection or a systemic inflammatory state with unknown pathogen (n=17; Table 1). These proportions differ from general infectious disease statistics on SCT/oncology wards (with predominant bacterial infections) due to the selection required by the GT study's inclusion criteria. The preceding interval from chemotherapy until the randomized infectious episode was clearly documented only in SCT recipients, where a range of 1–22 (median: 2) days passed from SCT date until the first GT administration, and the total duration of preceding episodes of neutropenia was highly variable.
The risk of death during a serious infectious episode as defined by the inclusion criteria decreased during the study period due to the introduction of new antimicrobial agents. Nevertheless, if GT had caused an accelerated neutrophil reconstitution, further therapeutic improvement of the underlying infection could have been expected. In practice, no significant difference in clinical outcome was found whether patients received GT or not (Table 2). The probability of survival on day 28 was 84 and 82% in the GT and the control arm, respectively (intention to treat). No difference in the incidence and causes of death could be identified between the two cohorts. The responses to GT in patients with fungal infections versus bacterial infections might differ due to the underlying immunologic mechanisms of the diseases. Thus, to identify a possible difference in the anti-infective treatment due to GT administration in these subgroups, we analysed patients with proven or probable fungal infections (n=55 out of 72 episodes, 28 of whom received GT) versus those with bacterial or unidentified pathogens (n=17 episodes, 10 of whom with GT). Similar to the whole cohort results, the survival on day 28 and until day 100 was independent of GT administration in both subgroups (Table 2). A minority of patients was randomized at a time when their neutrophil counts were still above 500 × 106/l because of a qualifying serious invasive infection and rapidly falling ANC with an expected long duration of aplasia (n=27 of 72 episodes). We did not retrospectively exclude these patients, as an interim study committee meeting decided that these subjects were eligible for evaluation within this study because of comparable inclusion criteria. However, we performed a subanalysis to exclude a potential bias. All calculations of survival on days 28 and 100 in patients with (i) any, (ii) fungal or (iii) bacterial or unknown infections yielded similar results with no detectable effect of GT treatment independent of whether patients had <500 × 106/l ANC at the time point of randomization or more (Table 2), suggesting that neutrophil substitution therapy within this study was ineffective even in patients with the most severe neutropenia. Thus, this study failed to identify a target population of patients with a significant benefit from GT administration. Notably, some important observations were made in this trial that might limit the value of any conclusion drawn from the following findings.
First, the ANC reconstitution curves of the treatment and the control arm were congruent (cumulative incidence of patients with ANC>500/μl per group; Figure 1), indicating that GT administered at the used schedule and dose were not sufficient to accelerate a lasting peripheral blood neutrophil reconstitution. Directly after GT, we detected an increase of ANC at 1, 8 and 24 h after GT (median: 480, 636 and 493 × 106/l above pretransfusion levels, respectively; data not shown) with few exceptionally high increments (maximum: 7200, 5700 and 5700 × 106/l neutrophils at 1, 8 and 24 h after GT, respectively). However, these high increments were randomly distributed among recipients and episodes and only rarely correlated to the cell dose in the transfusions (that is, upper-quartile ANC increments did not correspond to the neutrophil-richest concentrates and vice versa; data not shown). Similarly, the variability of GT cell content (median neutrophils 6.6 × 108/kg/concentrate, range 1.2–16, indicating that about 16% of GT contained less than the minimum recommended neutrophil dose of 3 × 108 per kg bodyweight) was evenly distributed between the GT recipients/episodes and thus did not cause a detectably different outcome of individuals. In contrast to previous observations, the response of GT in patients in whom GT were given early (<4 days) was equal to that of GT given ≥4 days after infection diagnosis and randomization (range 1–14 days, median=4). Thus, there was neither survival benefit from a short duration between randomization and GT administration, nor any disadvantage from a prolonged interval (see below). Furthermore, the relatively long median time lag from randomization to first GT of 4 days and the administration schedule (thrice a week instead of daily) did not allow rapid and durable restoration of the neutrophil count. Studies in severely life-threatening infectious episodes in children and adults clearly showed a correlation between transfused cell number, early GT administration and effect on ANC increment and survival.4, 5, 14, 19 Some collection facilities could not process GT during weekends. Yet, prepared concentrates were administered to the corresponding patients, even if the cell content or timing were not satisfactory. Additionally, there was no difference in the neutrophil response to the first or subsequent GT in any single recipient, suggesting that antineutrophil immunization of the recipient did not occur or was not clinically relevant. Concomitant cytokine therapy in addition to G-CSF to enhance the function of transfused neutrophils, such as recombinant IFN-g1b, has shown promising results,20, 21, 22 but detection or pharmacological stimulation of neutrophil function were not part of this study, which relied on neutrophil counts and clinical performance parameters. Peripheral blood counts might be an inadequate surrogate marker to assess the efficacy of GT because neutrophil tissue infiltration at the site of infection precedes the peripheral blood ANC increment. Mucosal neutrophil counts obtained by oral lavage have been shown to correlate better to tissue granulocyte concentrations than ANC,23 but they were not available within this study.
Secondly, another explanation for the better-than-estimated outcome of both cohorts and the failure of this study to detect an improved survival after granulocyte support for infections in neutropenia could be a ‘healthy cohort’ bias due to inclusion of patients not requiring GT and thus not expected to profit from this measure. Although all patients were severely ill as defined by meeting the inclusion criteria, it remained unclear why many patients recovered from their infectious episodes after only one or two GT (Table 2). Of note, 17 of 39 GT-receiving patients (44%) received only 1–2 GTs before neutrophil recovery and thus discontinuation of GT treatment (median number: 3 GTs, range 1–13; median duration of GT administration: 6 days, range 1–27), suggesting that the potentially detectable effect of granulocyte support in a majority of subjects was minimal. It appears unlikely that the administration of only 1–2 GTs represented the key to rapidly resolve a life-threatening infection justifying immediate discontinuation of GT treatment. Potentially, a lower level of neutropenia at randomization than 500 × 106/l ANC (for example, at ANC<200 × 106/l) as inclusion criterion would have facilitated detection of an improved clinical outcome after GT treatment. However, the ANC level for inclusion of patients was not intended to represent the nadir of neutropenia but to allow an early intervention in life-threatening infections even before the patients underwent the deepest phase of anticipated aplasia. Six patients, who did not receive GT although randomized to the treatment arm, and subjects with prolonged time lag (>7 days, n=7) from randomization to their first administration of GT, performed equally with respect to survival from severe infections, such as those who received GT within 7 days after randomization.
Thirdly, it was noted that numerous GTs were performed in the observation period without randomization in centres that planned to participate but did not randomize any patients. Three patients received six episodes of GT because of their deteriorating clinical state although randomized to the control arm, indicating that the existing empirical evidence for the benefit of this therapeutic measure was sufficient for the responsible physicians (Table 2), as if these subjects were not eligible for randomization according to the uncertainty principle.24 However, because it is not statistically correct to exclude patients retrospectively from a randomized study, we calculated the effect of GT for all identified subgroups of patients (intention-to-treat, per-protocol, ANC<500 × 106/l at randomization, fungal or bacterial infections and so on) to rule out a bias in the interpretation of results.
In summary, these data fail to confirm or refute the benefit of GT treatment because (i) the patient population in which this randomized clinical study was performed recovered faster than expected even without GT, and (ii) the transfused cumulative granulocyte doses and GT interval density in this trial were inefficient according to previous own and published experience.4, 10, 14, 16, 17, 19, 25, 26, 27 This study raises issues about the ethical feasibility of randomization of a safe and potentially life-saving treatment with level II-2 (or C) evidence of efficacy, points out future technical and clinical challenges, as well as recommendations for a still missing valid randomized controlled trial to obtain level I (A) evidence of the efficacy of GT treatment:
The observed delay and the slow sequence of GT administration could be circumvented in a future study by randomizing patients and identifying potential family donors already before administration of chemotherapy and scheduling daily aphereses once GT treatment is commenced.
To detect a difference in the clinical outcome with or without GT therapy, a selection of a homogenous cohort with even higher risk to succumb to a life-threatening infection during neutropenia, such as patients with refractory AMLs with prolonged preceding phases of neutropenia and ANC<200 × 106/l, appears to be essential.
In a future trial, all generated and handed out GT should be reported by the departments of transfusion medicine to avoid an undetected bypass of GT for not-included recipients during the study period, even if randomization is neglected by the patient or physician.
We thank Drs U Pötschger and S Karlhuber for their expertise and help in statistics and data documentation/processing, respectively. Furthermore, we thank Amgen, Schering and Chugai Pharma for travel grants. CP, AW, HN and HE designed the study; MGS, AW, RM, AB, GS and WG collected data; MGS, CP and HE analysed and interpreted the data and drafted the manuscript. All authors revised the manuscript critically for intellectual content and approved of its final version.