This study aimed to evaluate the efficacy of amikacine and ceftazidime as an empirical antibiotic therapy for neutropenic patients affected by haematological neoplasms and to investigate the presence of prognostic features suggesting a poor outcome with this antibiotic combination at the onset of infection. This could allow the identification of subgroups of patients with a low rate of response to amikacin/ceftazidime therapy; in these patients different initial empirical therapy may be indicated. The study population comprised 166 severely neutropenic (absolute neutrophil count below 500/Ixl) oncohaematological patients with fever or clinical signs of infection. Multivariate analysis confirmed four negative prognostic factors: 3 or more days of hospitalization at the onset of an infectious episode, a diagnosis of acute myeloid leukaemia, a haematological disease status different from complete remission, the presence of pneumonia. Depending on how many factors are present, cases can be stratified into three groups, of significantly different prognosis: favourable (0 or 1 factor) 76% success; intermediate (2 factors) 52% success; unfavourable (3 or 4 factors) 19% success. At the onset of an infectious episode a subgroup of patients with a very low response rate to empirical amikacin/ceftazidime antibiotic therapy is identifiable, for whom a different therapy is indicated. Because of the high rate of proven or probable fungal infections in this group, the immediate administration of a systemic antifungal therapy, in addition to antibacterial agents, could be considered in these high-risk patients. Studies should be specifically addressed to evaluating a stratification of empirical antibiotic therapy according to risk factors present at the onset of infection.