Artificial Neural Networks: A Novel Modeling Technique for Outcome

Artificial Neural Networks: A Novel Modeling Technique for Outcome

Prediction Following Idiotypic Immunotherapy in Ovarian Cancer

SA McQuarrie, AJB McEwan, C Ediss, TR Sykes and AA Noujaim,

Faculty of Pharmacy, University of Alberta, Edmonton, Canada

OVAREX(TM) is a recently developed product based on a monoclonal antibody (MAb-B43.13) known to recognize CA 125, a tumor antigen associated with epithelial ovarian cancer. This project utilizes OVAREX(TM) in a Phase II/III study designed to study the effectiveness of this MAb as an idiotypic immunotherapeutic agent. An artificial neural network was used to assess the impact of many potentially affiliated factors to model patient outcome following sequential iv injections. Relevant serum-based pharmacokinetic data, HAMA response, blood profiles, CA

125 levels, tumor burden, previous therapeutic interventions and patient response were used as input parameters to the model. This model used feed-forward, back-propagation fitting techniques with 1-2 hidden layers to provide weighting factors for each of the inputs. Preliminary results suggest that for 10-20 input parameters, patients can be successfully categorized based on tumor response or length of survival following diagnosis. With the additional data available at the completion of this study, it is anticipated that the weight associated with each input parameter will expose those factors that may impact treatment response to OVAREX(TM) idiotypic immunotherapy.

Steve McQuarrie, PhD

Assoc. Professor

Faculty of Pharmacy and Pharmaceutical Sciences

Download this Application Summary in Word97 format.
If you don't have Word97....Download the Word Viewer.