131 – Topics in Clinical Trials
Exploring Heterogeneity of Treatment Response: Assumptions, Logic, Algorithm, Computations
Lev S. Sverdlov
Redmond Analytics, LLC
Two views on population heterogeneity are discussed: heterogeneity of treatment effect and heterogeneity of response of the population to the treatment. Categories of sensitivity to treatment and capacity of spontaneous recovery are essential for defining treatment response. Our approach seeks to attribute population treatment response down to the individual level, rather than to proportions of the population. The population is split into four strata by binary treatment and outcome status. Aggregations of individuals sharing identical sets of descriptive variables are systematically identified. Hypotheses about variables affecting outcomes are formed by logical elimination through systematic comparison of individuals within and across strata and aggregations. This leads to a description of the heterogeneity of the treatment response, where phenomenologically identical outcome in different subsets of individuals can be related to distinct subsets of conditions. Results on the probability of random occurrence of relevant aggregates are derived using population parameters. The method‘s assumptions, logic, computational aspects, areas of applicability and limitations are discussed.