Premium
Bayesian detection of potential risk using inference on blinded safety data
Author(s) -
Mukhopadhyay Saurabh,
Waterhouse Brian,
Hartford Alan
Publication year - 2018
Publication title -
pharmaceutical statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.421
H-Index - 38
eISSN - 1539-1612
pISSN - 1539-1604
DOI - 10.1002/pst.1898
Subject(s) - bayesian probability , clinical trial , safety monitoring , computer science , patient safety , statistical inference , bayesian inference , risk assessment , inference , data mining , medicine , risk analysis (engineering) , medical physics , artificial intelligence , statistics , computer security , bioinformatics , health care , mathematics , economics , biology , economic growth , pathology
SUMMARY Safety surveillance is a critical issue for ongoing clinical trials to actively identify and evaluate important safety information. With the new regulatory emphasis on aggregate review of safety, sponsors are faced with the challenge to develop systematic and sound quantitative methods to assess risk from blinded safety data during the pre‐approval period of a new therapy. To address this challenge, a novel statistical method is proposed to monitor and detect safety signals with data from blinded ongoing clinical trials, specifically for adverse events of special interest (AESI) when historical data are available to provide background rates. This new method is a two‐step Bayesian evaluation of safety signals composed of a screening analysis followed by a sensitivity analysis. This Bayesian modeling framework allows making inference on the relative risk in blinded ongoing clinical trials to detect any safety signal for AESI. The blinded safety teams can use this method to assess the signal and decide if any safety signals should be escalated for unblinded review.