A risk model for first cycle febrile neutropenia in cancer patients receiving systemic chemotherapy

Abstract

8122 Introduction: Myelosuppression remains the major dose-limiting toxicity of cancer chemotherapy. A prospective, nationwide study was undertaken to develop and validate a risk model for febrile neutropenia (FN) in cancer patients receiving chemotherapy. Methods: More than 2,300 patients with cancers of the breast, lung, colon and ovary or lymphoma initiating a new regimen have been prospectively registered at 137 randomly selected sites. Data on at least 1 cycle of chemotherapy are available on 2,106. A standard regimen has been defined in 1,464 patients permitting an estimation of planned dose intensity (PDI). A logistic regression model for cycle 1 FN in those with known PDI was developed and test performance characteristics estimated. Results: Of 222 patients experiencing FN events, 125 (56%) did so in cycle 1 with prophylactic growth factors reported in only 3.5%. Major independent clinical factors in the risk model include: anthracycline chemotherapy (OR=7.80;P=.001); immune disorders (OR=5.14;P=.014); pretreatment neutropenia (OR=4.00;P<.001); elevated bilirubin (OR=3.73;P=.009); ECOG PS ≥2 (OR=2.84;P=.010); stage 1–2 disease (OR=2.66;P=.009); pretreatment lymphopenia (OR=2.51;P=.001); elevated alkaline phosphatase (OR=2.15; P=.012); pretreatment thrombocytopenia (OR=1.5;P=.024), and anemia (OR=1.43;P=.069). The risk model was adjusted for practice, planned dose intensity, BSA and patient demographics. The risk model accounted for over 80% of the variability (R2=0.829) with a c-statistic of 0.874 [0.847,0.902](P<.0001). Model discrimination was optimized for an estimated risk ≥5% distinguishing high and low risk groups with average cycle 1 FN risks of 20% and 1%, respectively. The model was associated with a sensitivity[±95%CL] of 93%[86–96], specificity of 66%[63–68], predictive value + and - of 20%[17–24] and 99%[98–100], respectively and diagnostic odds ratio of 22[12–43]. Discussion: Based on excellent test performance characteristics, the risk model identified patients at increased risk for first cycle FN for possible targeted prophylaxis with filgrastim or pegfilgrastim. Validation of the model in phase 2 of this ongoing registry study is currently underway. Author Disclosure Employment or Leadership Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Amgen Amgen Amgen