Group sequential designs using early outcomes in the R package gsearly: implementation and application in randomized clinical trials

Background : Group sequential clinical trials are widely used for the design of randomized clinical trials as they offer the possibility of stopping a study early for either treatment efficacy or futility. A common feature of many such designs is that the main study outcome is observed at a number of fixed occasions during participant follow-up with inferences being made on the final outcome. In settings where early outcomes are available, they offer the possibility of large gains in efficiency, due to the additional information they provide, compared to conventional group sequential designs based on the final outcome alone. Designing a study incorporating early outcomes currently requires users to undertake simulation studies at the planning stage that can be both time-consuming and costly. However, recently published R software package gsearly allows for the construction of designs in a routine manner, using analytic expressions that do not require users to undertake simulations. A brief description of the methodological approach and key assumptions are presented together with examples explaining implementation in R and code for user application. Methods: A model for longitudinal outcomes, with an assumed approximate multivariate normal distribution and correlation model for repeated outcomes, is described and used to motivate simple analytic expressions for the variance of (information on) the treatment effect at interim time-points for proposed recruitment models during data accrual in a clinical trial. Results: A worked example explains the basic relationships between model inputs and information accrual. Also, a detailed example of how a clinical trial in knee osteoarthritis might be planned is used to describe the use of the main functions within R package gsearly. Data can be simulated from a fitted gsearly model and used to explore potential design options more generally and monitor information accrual in a practical setting. Conclusion: R package gsearly provides an interface for developing and powering clinical trials using group sequential designs with early outcomes. It provides a wide range of options for building designs based on making some simple model assumptions or user inputs and tools for simulating data and monitoring information accrual, making the methods available for those without expert knowledge in this application area.