Prospective external validation of an updated algorithm to quantify risk of febrile neutropenia in cancer patients after a cycle of chemotherapy.

Purpose: Febrile neutropenia resulting from chemotherapy is a significant cause of morbidity and mortality in cancer patients. We had previously published the associates of the risk of febrile neutropenia, and this study now extends and modifies the previous model as well as tests its external validity.

Methods: We have recruited documented febrile neutropenia cases with solid tumors, in addition to a selected control group of cancer patients from one institution treated between 2015 and 2019. We then united our sample with our previously published original derivation group, to modify and update our previous model by logistic regression analysis. Additionally, consecutive cancer patients from 5 institutions were recruited in 2020 to test external validity of the resultant algorithm.

Results: A total of 4075 cycles of chemotherapy in 1282 cases were recruited in the updated, new model derivation group, and a total of 8 variables were selected for the updated algorithm. In the new external validation group, 653 cycles of chemotherapy in 624 patients were analyzed, to indicate that after cycles without prophylactic granulocyte colony-stimulating factor (GCSF) usage, the algorithm yielded a sensitivity value of 91%, specificity of 40%, and an area under curve (AUC) figure of 0.78, when a risk cutoff threshold value of ≥ 0.20 is chosen. This algorithm is now embedded in a web application for free clinical use.

Conclusion: Our algorithm identifies and quantifies the risk of febrile neutropenia in cancer patients. Further studies are required to improve this model with additional predictors.