×

Statistical issues in the evaluation of markers of HIV progression. (English) Zbl 0834.62105

Summary: Contemporary survival analytic praxis affords several methods for both appraising and extending the Cox proportional hazards model. With regard to assessment a variety of diagnostic plots based on martingale and related residuals are available. One extension, additive proportional hazards models, allows sums of univariate smooth functions of the covariates as a generalization of the Cox model linear predictor. A complementary method, tree-structured survival analysis, is adept at identifying homogeneous subgroups having distinct survival patterns.
We apply this array of techniques to an evaluation of markers of HIV disease progression. Issues concerning survival analysis from seroprevalent cohort data are described. Little used measures of partial dependence for censored data are invoked in this context. Results from the Western Australian HIV Cohort Study indicate that delayed-type hypersensitivity skin tests augment CD4 cell counts in predicting time- to-AIDS and time-to-death.

MSC:

62P10 Applications of statistics to biology and medical sciences; meta analysis
PDFBibTeX XMLCite
Full Text: DOI