A New Risk Score for Invasive Aspergillosis in Patients with Haematological Malignancies.

Invasive fungal infections, particularly invasive aspergillosis (IA), are frequent complications in immunocompromised patients. The identification of risk factors for IA is crucial to select patients who will benefit from different antifungal treatment. Retrospectively, we studied 1,194 hospitalizations registered at our Institution between March 2005 and December 2007, involving 618 patients affected by haematological malignancies (acute leukemia 34%, lymphoma 34%, multiple myeloma 24%, chronic myeloid leukemia 5%, severe aplastic anemia 1.5%, chronic lymphocytic leukemia 1.5%), with the purpose of developing a risk score able to stratify patients in categories at different risk for IA. Every febrile episode and every hospitalization without febrile episodes were recorded as single “phase”, for a total number of 1,409 “phases”. In 53% of the phases, the haematological malignancy was fully active (first presentation or resistant disease), and the main treatments administrated were: induction chemotherapy (18%), consolidation chemotherapy (20%), salvage chemotherapy (15%), HSCT in 33% (autologous: 21%; allogeneic: 12%). In 13% of the hospitalizations, patients did not receive chemotherapy but supportive cares. Seventeen variables were assessed as risk factors for IA according to literature data and to local clinical practice (age, smoking habit, type of job, previous IA, prolonged steroids treatment, diabetes, type of haematological malignancy, disease staus, type of treatment, prolonged neutropenia and lymphocytopenia, acute or chronic GvHD, CMV infection, mucositis, presence or absence of HEPA filter, proximity of building sites to the Haematology Unit) receiving a score ranging from 0 to 2. Summing the weighted points assigned to each variable, a total score (TSCORE08) was calculated for each phase and allowed to empirically identify 3 groups of patients at low (1 to 8), intermediate (8.5 to 11.5), and high (12 to 17) risk for IA, and was significantly higher in patients with IA (p < 0.001). Nine variables resulted to correlated significantly with IA in univariate analysis and were entered into the logistic regression analysis and 4 of them (type of job, lymphocytopenia, disease status, prolonged severe neutropenia) maintained their prognostic significance, allowing to derive, for each phase, a new score (BoSCORE08), which was also significantly higher in patients with IA (p < 0.001). The Areas Under the Curve derived from the TSCORE08 and the BoSCORE08 were equivalent, but the BoSCORE08 significantly reduced the number of variables taken into consideration. Adopting 5 as cut-off value for BoSCORE08, the model showed a high Negative Predictive value, a sensitivity of 0.950 (38/40) and a specificity of 0.545 (623/1,369), with a diagnostic Odds Ratio of around 22, which indicates that the risk of experiencing an IA in a patient with high BoSCORE08 was around 22 times higher than that of a patient with a low BoSCORE08. In conclusion, BoSCORE08 may contribute to optimize the management of patients at high risk for IA, improving the outcome, reducing the toxicity and the costs related to the use of antifungal agents in patients with haematological malignancies.