Online Program

Return to main conference page

All Times EDT

Friday, September 25
Fri, Sep 25, 11:45 AM - 12:45 PM
Virtual
Poster Session

PS05- Multiway Tipping Point Analyses in Longitudinal Clinical Trials with Missing Data (301074)

View Presentation

*Anders Gorst-Rasmussen, Novo Nordisk A/S 
Mads Jeppe Tarp-Johansen, Novo Nordisk A/S 

Keywords: missing data, multiple imputation, sensitivity analysis, tipping point analysis

When dealing with missing data in clinical trials, it is often convenient to work under a missing at random (MAR) assumption for the primary statistical model and rely on sensitivity analyses to address unverifiable missing data assumptions. One such sensitivity analysis, routinely requested by regulatory agencies, is the so-called tipping point analysis, in which the treatment effect is re-evaluated after adding a successively more extreme penalty to the predicted values among subjects with missing data. If the penalty needed to overturn the original conclusion is so extreme that it is considered clinically implausible then this indicates robustness to missing data assumptions.

One challenge with tipping point analyses is that it is not obvious how to best parametrise the penalty to be added to the predicted values in subjects with missing data. In a two-armed study, one might choose to penalise in the experimental arm only, yielding a computationally and conceptually convenient one-dimensional sensitivity analysis. On the other hand, it is arguably clinically more meaningful to apply arm-specific penalisation to both the experimental and comparator arm in order to obtain a two-way tipping point grid.

We describe a computationally efficient approach to evaluating a two-way tipping grid in longitudinal clinical trials when the underlying statistical model is ANCOVA with multiple imputation. Furthermore, we extend the idea of arm-specific penalisation to full subject-specific penalisation, providing a computational framework for perturbing a MAR model along the most conservative path towards the null hypothesis of no treatment effect. The methodology is illustrated using data from a trial in T2D.