External validation of a risk model of febrile neutropenia occurrence in patients with non-Hodgkin lymphoma.

Febrile neutropenia (FN) is a common and serious complication of chemotherapy treatment. Clinical risk models may help to identify patients at high risk of FN but must undergo external validation before implementation in medical practice. Therefore, this study externally validated previously published clinical models of FN occurrence during chemotherapy in 240 patients with non-Hodgkin lymphoma by using an independent observational dataset (n = 1829). The models demonstrated predictive ability, and validation criteria for predicting any cycle of FN were partially met but a larger than expected decrease in performance was noted (area under the receiver operating characteristic curve was 0.71 in the validation dataset and 0.83 in the training dataset). Age, weight, baseline white blood cell count and planned chemotherapy parameters were confirmed to predict FN risk. Chemotherapy dose reductions, dose delays and colony-stimulating factor use were confirmed as risk modifiers during treatment. Further work is needed to improve the predictive ability of FN risk models.