Modelling and Predicting Patient Recruitment in Multi-centre Clinical Trials

One of the main concerns in multi-centre clinical trials is how to enrol an adequate number of patients during a specific period of time. Accordingly, the sponsors are keen to minimise the recruitment time for cost effectiveness purposes. This research tended to concentrate on forecasting the patients’ accrual time for the pre-arranged number of sample size by simulating an on-going trial. The method was to model the data from the recruitment frequency domain and apply the estimations derived from the frequency domain to predict the time domain. Whereas previous papers did not concentrate on variations of recruiting over centres, this research assumed that patient arrivals followed the Poisson process and let the parameter of the process vary as a Gamma distribution. Consequently, the Poisson-gamma mixed distribution was confirmed as the promising model of the frequency domain. Then with the help of the relationship between the Poisson process and the exponential distribution, accrual time was predicted assuming that the waiting time between patients followed the Gamma-exponential distribution. As the result of the project, a trial was simulated based on the estimated values derived from completed trials. The first part of the prediction estimated the expected average number of patients per centre per month in an on-going trial. The second part, predicted the length of time (in months) to enrol specific number of patients in the simulated trial.