Prediction Region as a Quality Measure for Dose-Response Curves

• Activity Number: 254
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Quality and Productivity
• Author(s): Zhibao Mi and Nan Song and Zhewen Fan*+ and Kui Shen and Joseph Collins and Gong Tang
• Companies: Precision Therapeutics, Inc. and VA Cooperative Studies Program and Precision Therapeutics, Inc. and VA Cooperative Studies Program and University of Pittsburgh
• Keywords: Dose-response curves ; Prediction region ; Quality control

Abstract

Even though dose-response curves have been widely used as efficacy readouts in the life sciences, there is still interest in improving quality control methods for dose-response curves that do not belong to a parametric family. We propose constructing a prediction region based on a group of standard or previously accumulated dose-response curves as a quality control estimate of future generated dose-response curves with a predetermined level of probability. Lack of curve fitted parameters, such as LD50, the sample means, and the variances and covariances of the responses at various doses are used to build prediction regions using a multivariate technique. The prediction regions were built and tested using dose response data from human tumor cell growth curves inhibited by a panel of chemotherapy agents. If a testing dose response curve falls out the 95% prediction region, the curve is con