Evaluation of levofloxacine as antibiotic prophylaxis in MDS and AML patients treated with azacitidine: Results from the randomized open-label phase III azabac study Free

The incidence of febrile neutropenia in acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) patients is high. In AML patients receiving intensive chemotherapy, levofloxacine prophylaxis was associated with a reduction of the incidence of febrile neutropenia wich did not affect mortality in a meta-analysis (Owattanapanich et al, Hematology, 2019). In patients treated with azacitidine (AZA), who are typically treated in an outpatient setting, reducing the incidence of febrile neutropenia may be crucial.

We conducted a multicenter phase III open-label randomised (1:1) study to evaluate levofloxacine as antibacterial prophylaxis in AML and MDS patients treated with AZA. Patients aged over 18y treated with AZA in first line either for AML or for intermediate 2 or high-risk MDS according to IPSS were eligible. Associations of AZA with other treatments were allowed. Patients randomised in the experimental arm (A) received levofloxacine 500 mg daily during the first 3 cycles of AZA. Those in the reference arm (B) received no prophylaxis. The primary endpoint was the incidence of febrile neutropenia causing an hospitalisation and start of an other systemic antibiotic treatment during the first three cycles. One-year overall survival (OS) was assessed as secondary endpoint.

From July 2018 to July 2024, 59 patients (MDS: n=44, AML: n=15) were enrolled in the study. Twenty-nine patients were assigned to arm A and 30 patients to arm B. Median age was 75 [IQR: 70-80] and 61% (n=36) were males.

Among the 29 patients in the arm A, the median effective duration of the levofloxacine prophylaxis was 71 [IQR: 29-90] days, 65 [IQR:35-90] days for MDS patients and 77 [IQR: 18-87] days for AML patients. Toxicity was the reason for prophylaxis interruption in 7 patients (24%). Eight patients (13.3%) experienced at least one adverse event that was considered treatment-related, all of which were grade 1 or 2. Tendinitis was the most common (4 patients). Adverse events involved 50 patients (83.3%) (arm A 24/29, arm B 26/30).

Febrile events incidence did not significantly differ between the 2 arms (p=0.58), with 5 patients (17.2%) in arm A having experienced febrile events and 8 patients (26.7%) in arm B. The median time to febrile event was 31 days (range 12-49) among the 5 patients from the arm A, and 20.5 days (range 1-36) among the 8 patients from the arm B. No febrile events in arm A were grade 4-5 while 4 events (50%) in arm B were grade 4-5 (p=0.11). Among the 4 patients in arm B with grade 4-5 febrile events, 3 died due to this event, and 2 had probable pneumopathy. In a multivariate logistic regression analysis, risk factors for febrile events included low baseline neutrophils count (<0.5 G/L) (p=0.03), low baseline albumin level (<34g/dL) (p=0.01), and age over 75 (p=0.03).

Two over 5 febrile events in arm A (40%) and 3/8 febrile events in arm B (37.5%) had microbiological documentation.

Carbapenem and glycopeptide precription did not significantly differ between the 2 arms (carbapenem: 2/29 vs 1/30 patients, p=0.61 ; glycopeptide: 1/29 vs 2/30 patients, p=1).

Twelve patients in the experimental arm and 17 patients in the reference arm died during the study. Median OS was 12.2 months from day 1 of the first cycle of AZA. Cause of death was a febrile event in 3 patients (17.6%) in arm B and none of the patients in arm A, and the disease progression in 4 patients (33.3%) in arm A and 7 patients (41.2%) in arm B. OS did not significantly differ between the 2 arms (p=0.16). One-year survival rate was 57% (95%CI: 42%-79%) in arm A and 50% (95%CI: 35%-71%) in arm B.

Levofloxacine prophylaxis did not significantly reduced febrile events incidence or improve survival in MDS / AML patients treated with AZA. However, several deaths attributed to febrile events in the reference arm suggest a probable benefit of levofloxacine in this setting but the limited size of this trial does not enable us to confirm it. Larger studies would be useful to investigate this question.