Central Statistical Monitoring at ISCB Conference
Every year, the International Society for Clinical Biostatistics (ISCB) holds its annual conference covering all essential industry questions. This year, the conference took place in Vigo, Spain.
During the event, invited lecturers from the world’s top universities (Harvard University, University Of Cambridge, University of Oxford, Stanford University, University of Tokyo, Kyoto University, and others) gave their presentations and training sessions in academic biostatistics. Some lectures and workshops were held by representatives of the pharmaceutical industry: Novartis Pharma AG, Bayer AG, F. Hoffmann-La Roche LTD, Glaxo Smith Kline Pharmaceuticals, CYTEL INC, Costello Medical Consulting LTD, Taiho Pharmaceutical. CO, LTD, etc.
It is worth noting that most of the presented approaches can be applied in practice, such as the methodology used to deliver statistically unbiased estimators of treatment response using SAS, R and Stata (follow the link to learn more). The participants have also considered a number of solutions for sample size determination in adaptive design, repeated measurements, etc. Many interesting and relevant solutions for R package were demonstrated.
Data MATRIX team gave a poster presentation on central statistical monitoring (CSM) approach and prospects of its application in risk-based monitoring (RBM) activities. This method is designed for binary data (i.e. events, where 0 — ‘no’, 1 — ‘yes’) collected over a specified time period. In this approach, patients are considered as vectors elements equal to 0 or 1. With that, the resultant vector of overlapping/matching is defined for each study site, which describes the ‘uniformity’ of patients. The normalized resultant vectors of overlapping/matching are then used for standard classification problem.
A similar question was addressed in another presentation (Tomoyoshi Hatayama Biostatistics Department A2 Healthcare Corporation, Department of Biostatistics Graduate School of Medicine Yokohama City University, Japan). Instead of abstract events, adverse events (absence or presence) were used, and a completely different approach was taken to provide a solution (Bayesian statistical methods).
What makes this question particularly challenging is that there are no specific guidelines on application of CSM methods. Moreover, this problem is discussed not only at CSM and RBM focused conferences.