370 – Contributed Oral Poster Presentations: ENAR
Tipping Point Analyses: A Case Study
Emilea Norris
PPD
Graham Carron
PPD
The purpose behind a tipping point analysis (TPA) is to find the point ("tipping point") at which the objectives of an underlying analysis are no longer met. TPAs have gained popularity in recent years as a tool to assess sensitivity to assumptions about missing data without additional modeling for a variety of analysis types. However, little guidance exists regarding methodologies for implementing a TPA. Various methods exist for a TPA with an underlying time-to-event analysis through which incremental adjustments of the censored observations achieve the tipping point, such as observed time of the censored event (earliest, latest, or random). As a result of recent experience with regulators, this presentation will illustrate TPAs and associated output based on actual data with a time-to-event endpoint. The impact of conversion order will be evaluated, as will the distributions of the different censored observations. Visual outputs summarizing different implementation methods will be included, along with interpretations of the overall results and comparison of TPA performance against the performance of another well-established approach.