All Times EDT
Keywords: Artificial intelligence, anchor-based approach, composite endpoint, continuation logit model, gene therapy, patient reported outcome, rare disease, supervised machine learning
To address various limitations of the multi-domain responder index (MDRI), a patient centric PAOC composite ordinal response endpoint (CORE) by which each patient’s treatment outcome is classified as positive response, no response, or negative response via an anchoring approach by a PRO endpoint. Due both various structural possibilities of utilizing multiple endpoints and large number of potential cut-off values for each structure for defining a CORE, a supervised machine learning methodology is used for a training set arising from blinded accumulating data of a randomized trial. The derived CORE is then validated with later data and incorporated into the statistical analysis plan (SAP) prior to the database lock. The methodology and process expand the existing anchor-based approach for defining responder variable based on a single clinical endpoint. A continuation logit model is used to combine inference of comparisons of both positive and negative response between treatment groups to substantial increase the statistical power.