G-17

ON ESTIMATION OF THE PROPORTION OF TREATMENT BENEFIT WITH A CONTINUOUS RESPONSE WITHOUT DICHOTOMIZATION

Y. Tsong, CDER, FDA, Silver Spring, MD

Walter and Irwig (2001, SIM) proposed a definition of the number needed to treat (NNT) based on normally distributed response variables without dichotomizing the data and proposed an estimate using method of moment. The NNT index is defined by inverting the proportion of benefit of the test treatment adjusted for the reference treatment. However, the definition is defined in an asymmetric way. In this presentation, the symmetric definition is proposed. In addition, the moment method proposed by Walter and Irwig underestimates the variance of the estimate. It is illustrated with a comparison with the results derived from the bootstrap method.


2006 FDA Science Forum | FDA Chapter, Sigma Xi | CFSAN | FDA
Last updated on 2006-MAR-27 by frf