Add Time Dimension to the Prediction: a Multi-Stage, Sequential Prediction Model

• Activity Number: 587 - Risk Modeling
• Type: Contributed
• Date/Time: Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Risk Analysis
• Author(s): Zheng Zhang*
• Companies: Brown University
• Keywords: Risk prediction; ROC curve; AUC; predictive biomarker; predictive accuracy; external validation

Abstract

In ACRIN 6698/I-SPY II, a prospective, multicenter trial, serial MRI imaging has been used to monitor the treatment response and predict pathological complete response among neoadjuvant breast cancer patients. Up to four DWI-MR scans were done at pre-, early-, mid- and post-treatment time points on 242 participants. Changes in the ADC value from pre-treatment at each of the three later time points were evaluated for predictive accuracy by the ROC curves and the corresponding AUC values. We propose to build a multi-stage, sequential prediction model by incorporating the serial imaging data to achieve the best prediction accuracy. Separate models were built for tumors with different molecular subtype. Model performances were validated externally.