Further resources
Roche resources
Introductory video by Greg Spaniolo and Kaspar Rufibach on R&D academy nucleus
PTS handbooks: One-stop shop for all PTS resources for all functions at Roche. To get more background on the concept and how it is applied in different parts of the organization it is recommended to have a look at these handbooks.
ReImagined TPP guidance: Portal to TPP guidance after substantial overhaul in 2021.
Shiny app to compute DDCP: Not developed by successR. We strongly recommend to verify computations from the shiny app using the
bpppackage and store the code for reproducibility.Shiny app to compute predictive probabilities: Neither developed by successR nor Unicycle. We strongly recommend to verify computations from the shiny app using the
bpppackage and store the code for reproducibility.Project Providentia: Project Providentia is developing Bayesian Decision Support Tools that can assist members of both PD Clinical and Regulatory to create transparent, consistent estimates for the Probability of Success for either clinical trials or regulatory submissions. The methods developed here are intended to be used after DDCP computations, i.e. DDCP is an input for these methods. The idea is to formalized the informal process of up- / or downgrading DDCP to finally arrive at PTS as described in the above PTS handbooks.
Recent informative publications
Kunzmann et al. (2021) or an earlier version on arxiv: A comprehensive review of the literature around DDCP. Figure 1 provides the connection between all the different quantities that fly around in this space.
Hampson et al. (2022): Technical paper by Novartis colleagues outlining their very comprehensive approach to try to quantify virtually all aspects of PTS computation. See here for a summary.
Hampson et al. (2022): Summary paper for a lay audience of the above paper by Novartis colleagues. See here for a summary.
Kundu et al. (2021): Paper by Boehringer-Ingelheim colleagues providing formulas for various DDCP scenarios for continuous, binary, and time-to-event endpoints and one or two samples. Accompanying R package is LongCART.