Abstract P3-15-08: A multicenter observational study to investigate the relationship between physician-assessed febrile neutropenia (FN) risk and prediction model FN or SN (severe neutropenia) risk in patients with breast cancer

Abstract

Abstract Background: A prediction model was previously developed to estimate the risk of SN or FN during the first cycle of chemotherapy, taking into account both the myelotoxicity of the chemotherapy regimen and the interplay of specific tumor and patient characteristics (Lyman, 2011). This model may potentially be useful for predicting which patients are at greatest risk for developing neutropenic complications, particularly among those patients receiving intermediate FN risk chemotherapy regimens, where the influence of patient characteristics becomes a critical consideration. To assess the clinical utility of this model, physicians assessed FN risk in patients with non-myeloid malignancies, and then physician-assessed FN risk was compared to prediction model risk. Methods: This was a prospective, multicenter, observational study (124 community-based oncologists, 944 patients). Analysis of a breast cancer subgroup (93 oncologists, 364 patients) is reported here. Patients were eligible if they were: ≥18 years old, newly diagnosed, and candidates for initiating a new course of chemotherapy using an NCCN intermediate (10-20%) FN risk chemotherapy regimen. Oncologists entered clinical data about the patient into the model; they also made a clinical prediction of FN risk. Oncologists were blinded to both the data elements collected by the model and the risk predicted by the model. Data were only collected until the chemotherapy order was written; no outcome data were collected. The primary objective was to investigate the relationship between physician-assessed FN risk and prediction model SN or FN risk. As an exploratory endpoint, physicians were asked to estimate FN risk on the same set of four hypothetical case studies with varying FN risk factors. The correlation between risk probability scores was estimated as well as a 95% confidence interval accounting for intra-physician correlation. A smooth spline curve was fit to show the average relationship between physician-assessed and prediction model risk probability scores. Results: Most patients were planning to receive TC (54%) and most had stage I-II disease (71%). Median (min, max) age was 58 years (23, 83). Physician-assessed FN risk correlated weakly with prediction model FN or SN risk: correlation 0.166 (95% CI: 0.027, 0.298). There was wide variability among all 124 physicians in their assessment of FN risk for the four case studies (Q1, Q3 case study 1: 20%, 40%; case study 2: 10%, 18%; case study 3: 20%, 40%; case study 4: 25%, 60%). Conclusions: This study suggests that community oncologists would benefit from the use of an automated system for assessing FN risk among those receiving intermediate risk chemotherapy regimens. Such a system would help oncologists identify the patients who would benefit most from clinical intervention and help improve practice efficiency and quality of care. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P3-15-08.