You can now register for
this event. Registration fees are as follows:
\n- \;Members of PS
I \;= Free of charge
\n- \;Non-Members of PSI \;= £\
;20+VAT
\nTo register for the session\, please \;click here.
To assess the strength of clinical study findings\, reg
ulatory authorities often request tipping point analyses. However what is
a tipping point analysis\, and how is one performed? What are the differen
t approaches for continuous\, binary and time to event data? Kevin and Jua
n present some practical examples.
\n
\nTo view the flyer for th
is event\, please click here.
\n
Speaker \n | \n Biography \n | \n Abstract \n |
\n
| \n Juan is a statistician b y training\, by experience and by passion. He studied Maths and Stats and did his PhD at the University of Valencia (Spain)\, and worked as a resear ch fellow at Imperial College London. He has worked as a statistician in v arious roles in Public Offices\, Academia and Industry in Spain\, Germany and the UK. He's currently a GSK fellow and Statistics Director at the Sta tistics and Data Science Innovation Hub\, based in London. His current mai n interests include the use of Bayesian thinking for better quantitative d ecision making in drug development\, estimands and their estimation in the presence of missing data and the analysis of data collected with digital wearable technologies. \n\; \n | \n In this presentation I will illustr ate one way of implementing a tipping point analysis (TPA) for a time-to-e vent endpoint to assess the robustness of results to the censoring-at- random assumption. The method is based on multiple imputation and it assumes the hazard rate post-censoring changes. A grid of values reflectin g such changes is considered to vary the hazard rate post-censoring indepe ndently for each treatment arm. The (experimental\, control) pairs of post -censoring hazard rate changes form the TPA scenarios. Within each of the TPA scenarios\, participants who are censored are imputed a time to the ev ent of interest and are administratively censored if the imputed time exce eds the length of the follow-up. Results are combined across imputed datas ets using Rubin&rsquo\;s rules. Finally\, the plausibility of the scenario s where the results tip is discussed. \n |
\n
| \n
Kevin has 13 years of industry experience of designing and imple menting study design of clinical trials in the areas of both general medic ine and oncology. Kevin previously worked in J&\;J (late phase oncology ) and Novartis (late phase immunology) and now is working in AZ as Statist ical Science Associate Director for early phase oncology. Kevin has 14 aca demic journal publications and has special research interest in practical use of estimand and approaches of dealing with missing data in late phase trials. He presented &ldquo\;The Application of Tipping Point Analysis in Clinical Trials&rdquo\; in 2018 JSM meeting. \n\;\n | \n Practical use of Tipping Point Analysis in regul atory submissions of clinical trials \nThe tipp ing point analysis (TPA) approach has gained popularity recently as an app roach for performing the sensitivity analysis under the missing not at ran dom (MNAR) assumption. This presentation will review why TPA gets popular in clinical trial submissions\, its implementation for binary endpoints fo r time-independent imputation and time-dependent imputation\, general proc edure of TPA implementation for continuous endpoints\, and Interpretation of TPA result based on clinical input. The presentation then shows six rea l examples of FDA statistical review of submitted NDA/BLAs (all in public domain) that use TPA as sensitivity analysis to illustrate the practical u se of this method in real trials and important consideration points from t he regulatory perspective. \n |
\;
\n\n