Evaluating treatment efficacy using AUC modeling.
Access RightsNo access - Contact email@example.com
Van Zyl, Johanna S. 1991-
MetadataShow full item record
We have proposed the use of the area under the receiver operating curve (AUC) to address two problems related to determining the efficacy of a treatment based on clinical data. The proposed methods are developed for the scenario when standard parametric assumptions do not hold. Dodd and Pepe (2003) developed a semi-parametric AUC regression model that enables one to adjust the AUC for covariates. Since the Mann-Whitney statistic is a nonparametric unbiased estimate of the AUC (Bamber, 1975), Buros et al. (0) utilized the AUC regression model and the relationship between the Mann-Whitney statistic and AUC to develop a covariate adjusted Jonckheere Terpstra (JT) test. Buros et al. (2017) developed a nonparametric multiple comparison procedure for tests where the alternative hypothesis is monotone as is often the case in a dose response study. Chapter Two develops a solution to a multiple comparison problem in a dose-response study when each dosage group is compared with a zero-dose or placebo control. The proposed methods are extensions of Buros et al. (0) development of AUC regression for the Jonckheere Terpstra test. In Chapter Three we present a procedure to make an indirect comparison between treatments B and C when each have been compared to a common control, A. Our approach combines the use of the AUC as the measure of clinical effectiveness and a nonparametric Bayesian mixtures of finite Polya trees (MFPT) model to estimate the AUC (Branscum et al., 2008, 2015). Chapter Four contains a discussion of some of the topics and issues associated with using the MFPT model in the network meta-analysis described in Chapter Three. In Chapter Five we present a practical application of a meta-analysis in synthesizing outcomes in Tuberculous Pericarditis (TBP).