Risk Factors Associated with Occurrence of Febrile Neutropenia in NHL Patients in the First Cycle of Chemotherapy - A Prospective Multivariate Model.

Current guidelines highlight non-Hodgkin lymphoma (NHL) patients receiving standard chemotherapy regimens as having a significant risk of febrile neutropenia (FN). However, limited data are available regarding the contribution of additional risk factors to FN occurrence. The majority of initial FN events occur in the first cycle of chemotherapy. We present a logistic regression model describing cycle 1 FN occurrence in NHL. Covariates were chosen based on their clinical and statistical significance, tested at the 5% significance level. Data were obtained from the INC-EU Prospective Observational European Neutropenia Study, which was conducted to assess the incidence and predictors of neutropenia, FN and reduced chemotherapy administration for breast cancer and lymphoma patients in European practices. 240 NHL patients were recruited from 37 centres in Belgium, France, Germany, Spain, and the UK. Ann Arbor stages were distributed: I 18%; II 26%; III 17%; IV 40%. Chemotherapy regimens received were 3-weekly CHOP-like (74%), 2-weekly CHOP-like (17%), ACVBP-like (4%) and other regimens (5%). Prophylactic use of colony stimulating factor (CSF) in the first cycle was 33%. Overall incidence of FN in cycle 1 was 9%. In the multivariate model of cycle 1 FN occurrence, the final set of covariates included low baseline albumin; high baseline alkaline phosphatase; glomerular filtration rate (GFR); height; recent infection; previous chemotherapy; planned cyclophosphamide, cytarabine and etoposide doses and prophylactic CSF. Missing categories were added for albumin and alkaline phosphatase to avoid loss of observations. Clinically relevant components of the model are shown in the table. The area under the receiver operating characteristic curve was 0.89. When the optimal probability cut-off was used to predict FN occurrence, test characteristics were 81% sensitivity and 80% specificity; corresponding to positive and negative predictive values of 29% and 98%, respectively. GFR is inversely related to age and height is related to weight. Replacing GFR and height with age and weight slightly affected predictive ability but did not affect overall model stability. Prophylactic CSF use suggested a significant protective effect. Low baseline albumin, high alkaline phosphatase, low GFR or high age, low height or weight, recent infection, previous chemotherapy, and high planned cyclophosphamide dose, may be important clinical signals of an increased risk of first cycle FN occurrence. This study was supported by an unrestricted educational grant from Amgen, Europe.