HEMATOPOIETIC STEM CELL TRANSPLANTATION OUTCOMES- LOGISTIC REGRESSION MODEL DEVELOPMENT
Author(s)
Hatfield MD1, Kramer MA2, Johnson ML1
1University of Houston, Houston, TX, USA, 2The University of Texas MD Anderson Cancer Center, Houston, TX, USA
OBJECTIVES: The objective was to determine the most effective logistic regression models in terms of explaining the greatest amount of variance regarding four outcomes: graft versus host disease (GVHD), liver toxicity, neurotoxicity, and mortality, among a cohort of patients undergoing hematopoietic stem cell transplantation. METHODS: Busulfan is used in combination with fludarabine or clofarabine as part of an effective chemotherapy based myeloablative preparative regimen for patients undergoing stem cell transplantation. Pharmacokinetic data regarding patient busulfan clearance was used in the analysis, since dosing is very sensitive. Other clinically relevant covariates included: age, gender, race, primary cancer diagnosis, type of transplant (autologous or allogeneic), and prior transplant history. Descriptive statistics and logistic regression analyses were performed to assess the effect of these variables on each of four outcomes: GVHD, liver toxicity, neurotoxicity, and mortality. Hosmer and Lemeshow goodness-of-fit tests and c-statistics were used to optimize the models. Only aggregate level information was reported. Statistical significance was set at 0.05. RESULTS: Data on a cohort of 752 patients undergoing hematopoietic stem cell transplantation were collected. Most patients were: 46.7±15.8 years old, male (59.8%), Caucasian (68.6%), with acute myeloid leukemia (28.9%), underwent an allogeneic transplant (67.0%), and had not received a prior transplant (93.5%). Controlling for the covariates listed, the models resulted in Hosmer and Lemeshow goodness-of-fit test statistics (chi-square, number of degrees of freedom, p-value) for each dependent variable as follows: GVHD (2.57, 5, 0.77); liver toxicity (5.17, 8, 0.74); neurotoxicity (8.45, 8, 0.39); and mortality (5.95, 8, 0.65). The c-statistics for each model were: GVHD (0.87); liver toxicity (0.74); neurotoxicity (0.75); and mortality (0.72). CONCLUSIONS: The logistic regression models used were effective in determining the outcomes of GVHD, liver toxicity, neurotoxicity, and mortality, among a cohort of patients undergoing hematopoietic stem cell transplantation.
Conference/Value in Health Info
2014-05, ISPOR 2014, Palais des Congres de Montreal
Value in Health, Vol. 17, No. 3 (May 2014)
Code
PRM59
Topic
Methodological & Statistical Research
Topic Subcategory
Modeling and simulation
Disease
Oncology