A Prospective Risk Model for Neutropenic Complications in Patients with Malignant Lymphoma.

Background: Myelosuppression including severe and febrile neutropenia continues to represent a major cause of dose-limiting toxicity of cancer chemotherapy. Neutropenic complications in cycle 1 have been shown to frequently lead to reduced dose intensity or addition of a myeloid growth factor. A prospective, nationwide study was undertaken to develop and validate risk models for first cycle neutropenic events associated with cancer chemotherapy.

Methods: Patients with malignant lymphoma initiating a new chemotherapy regimen have been prospectively registered at 115 randomly selected practice sites. Data on at least one cycle of chemotherapy were available on 357 patients including 56 with Hodgkin’s disease and 301 with non-Hodgkin’s lymphoma. A logistic regression model for first cycle neutropenic events based on pretreatment characteristics was developed and predictive test performance characteristics examined.

Results: Severe or febrile neutropenia occurred in 81 (22.7%) of patients in cycle 1. Two-thirds of patients with one or more neutropenic events experienced their initial event in cycle 1. Significant independent pretreatment predictive factors for first cycle neutropenic complications were: history of renal disease (OR=33.15, P=.011) or recent infection (OR=12.41, P=.035), Caucasian (OR=4.13, P=.053), use of an anthracycline-based chemotherapy regimen (OR=6.9, P<.001), baseline absolute neutrophil count (OR=0.95, P<.001) and lymphocyte count (OR=0.95, P<.001), anemia (OR=2.14, P=.043), elevated lactate dehydrogenase (OR=2.0, P=.040) and elevated bilirubin (OR=2.25, P=.051). Model fit was excellent (P<.001), R²= 0.503 and c-statistic = 0.769 [95% CL: .71–.83, P<.0001]. Individual predicted risk of cycle 1 events based on the model ranged from 0 to 96% with mean and median probabilities of 0.23 and 0.21, respectively. Two-thirds of patients were classified as high risk with mean risk scores in high and low risk subjects of 32.0% and 5.6%, respectively. Model test performance characteristics [±95% CL] included: sensitivity: 92% [84–96]; specificity: 40% [34–46]; likelihood ratio positive: 1.53 [1.36–1.73]; likelihood ratio negative: 0.19 [0.09–0.42]; positive predictive value: 32% [26–38]; negative predictive value: 94% [88–97] and diagnostic odds ratio: 7.92 [3.32–18.91].

Discussion: The risk model identified lymphoma patients at increased risk for first cycle neutropenic complications using common clinical parameters. Validation of the model in a separate population of patients is in progress.