The presentation will start with a discussion of the statistical considerations in the selection of the study population including the issue of homogeneity versus heterogeneity, risk of the outcome and implications for sample size and power.
The intention-to-treat principle will be contrasted with per-protocol analyses and the intent-to-treat design and analysis will be described. The impact of missing data on the analyses will be described and it will be shown that Last Observation Carried Forward (LOCF) is fallacious.
The basic principles of organizing a data and safety monitoring board (DSMB) are then described along with a non-technical description of the statistical principles of interim monitoring for effectiveness, safety and/or futility.
John M. Lachin, Sc.D.
Research Professor of Biostatistics and of Epidemiology, and of Statistics
The George Washington University Biostatistics Center