Comparison of Treatments with Multiple Outcomes
Treatment comparisons in clinical studies often involve several endpoints, particularly those related to quality of life of patients suffering from diseases like cancer and arthritis. We review traditional statistical methods for this problem and describe some new approaches. One approach is based on a new formulation of the null hypothesis that incorporates the essential univariate and multivariate features of the treatment effects. Another approach is based on assigning benefit scores to different regions of the toxicity-efficacy outcome space. A third approach involves patient thresholds for tolerating different treatments. Bootstrap methods are used to circumvent the analytic and computational complexities of the new approaches. We illustrate these approaches using data from patients with rheumatoid arthritis. In this setting, quality of life involves trade-offs between efficacy and toxicity of treatments.
KeywordsNull Hypothesis Rheumatoid Arthritis Patient Error Probability Multivariate Normal Distribution Multiple Outcome
Unable to display preview. Download preview PDF.
- Cook, R.J. and Farewell, V.T. (1996). Multiplicity considerations in the design and analysis of clinical trials. Journal of the Royal Statistical Society Series A 159, 93–110.Google Scholar
- Efron, B. and Tibshirani, R.J. (1993). An Introduction to the Bootstrap. New York: Chapman and Hall.Google Scholar
- Kudo, A. (1963). A multivariate analogue of the one-sided test. Biometrika50, 403–418.Google Scholar
- Perlman, M.D. and Wu, L. (1999). The emperor’s new tests (with discussion). Statistical Science 14, 355–381.Google Scholar