Modelling, prediction and adaptive adjustment of recruitment in multicentre trials

Abstract

This paper is focused on statistical modelling, prediction and adaptive adjustment of patient recruitment in multicentre clinical trials. We consider a recruitment model, where patients arrive at different centres according to Poisson processes, with recruitment rates viewed as a sample from a gamma distribution. A statistical analysis of completed studies is provided and properties of a few types of parameter estimators are investigated analytically and using simulation. The model has been validated using many real completed trials. A statistical technique for predictive recruitment modelling for ongoing trials is developed. It allows the prediction of the remaining recruitment time together with confidence intervals using current enrolment information, and also provision of an adaptive adjustment of recruitment by calculating the number of additional centres required to accomplish a study up to a certain deadline with a pre-specified probability. Results are illustrated for different recruitment scenarios.

Discussion on the paper “Real-Time Prediction of Clinical Trial Enrollment and Event Counts: A Review”, by DF Heitjan, Z Ge, and GS Ying

“…On this way we are naturally coming to using a so-called Poisson-gamma enrollment model (P-G model) developed in [1][2][3][4][5]. This model assumes that the patients arrive at clinical centers according to delayed doubly stochastic Poisson processes where the variation in rates between different centers is modelled using a gamma distribution.…”

“…In the framework of P-G model [1][2][3][4][5], the enrollment processes at different levels are modelled as non-homogeneous Poisson processes with time-dependent and in general random rates, which are governed by the processes of center's opening and closing as well as individual center's data. Together with modeling enrollment at the start-up stage, P-G model can be efficiently applied to an interim prediction.…”

“…One of the essential features is the opportunity to evaluate the interim adaptive adjustment (if enrollment is going slower as expected, evaluate the number of new centers needed to be added with the purpose to complete enrollment in time with a given confidence). In addition, this technique has several other features that are available only in this framework, e.g., predicting center/country performance, number of "empty" centers, creating optimal enrollment design [1][2][3][4][5].…”

“…3.3 [11] that "Mijoule et al [15] proposed replacing the gamma with a Pareto mixture." Actually in [15] the authors investigated further properties of P-G model, compared them with the Pareto-Poisson model using real datasets from [2], and also investigated the feasibility of the model. In final conclusions the authors "recommend the use of the Poisson-Gamma, which is easier to handle", and also recommend using a uniform distribution for centers initiation when the opening dates of the centers are not known precisely, which is proposed in [3,4].…”

“…The technique accounts for the events that may happen for patients already at risk in the trial and for events that may happen for patients that will be recruited in the future. The enrollment is modelled using P-G model [1][2][3][4][5], and the process of events is modelled on the top of enrollment. For multiple events (recurrence, death and lost-to-follow-up) the event process is described by finite Markov models.…”