Select Matching Controls for Treated Cohort

• Activity Number: 178
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xin Chen*+ and Zhenyu Jia and Dan Mercola
• Companies: University of California at Irvine and University of California at Irvine and University of California at Irvine
• Keywords: classification ; clinical

Abstract

Control cohort was missing in a study of an adjuvant multimodality therapy for prostate cancer. We hope to select matching untreated patients from another study to complete the survival comparison and justify the efficacy of the therapy. This is a modified statistical classification problem where traditional classification methods such as logistic regression and support vector machine do not apply since only one group is specified in the training set. We proposed an unsupervised clustering method which involves the calculation of the weighted Euclidean distance based on clinical variables of interest, such as age, pre-OP PSA, Gleason score and etc. The control patients that are close to the treated cohort in terms of Euclidean distance will be selected. The method was verified by analyzing two simulated datasets, and further demonstrated using prostate cancer patient data.