Statistically-Guided Review of Safety Data in Clinical Trials
*Michael O'Connell, Spotfire, TIBCO Software, Inc 

Keywords: Drug Safety Visualization Forest Bagging Spotfire S+

Assessment of drug safety is a key aspect of drug development. However, clinical trial design is focused on the efficacy endpoints, and safety data analysis is typically an afterthought.

In this talk I present some statistical and graphical analyses of safety data that enable detailed clinical review of patient-level data. Statistical models considered include recursive partitioning, forests and penalized logistic regression. These models are fitted as S-PLUS® / R analyses with resulting graphical summaries being displayed in the interactive Spotfire environment and as publication report graphics.

The analysis, visualization and review can be done as interim analyses and in preparation for clinical study reports, DSMB and advisory board meetings. This provides an ongoing coherent and transparent picture of a drug’s safety profile for all stakeholders in the development process.