Table 3 Detail of Trial Data Reported in Modelling Papers
From: A systematic review of models to predict recruitment to multicentre clinical trials
Author | Used Real Life Data? | Where From/How | Was Prediction Useful/authors conclusions in relation to the model and data used |
|---|---|---|---|
Carter 2004 | No | Theoretical example only | Not applicable |
Carter 2005 | Yes | They used information from a multicentre RCT protocol under development to test and illustrate the three models | The authors conclude that the unconditional approach can yield results not consistent with trial assumptions and may endanger the successful completion of the trial. The models are sensitive to the estimated accrual rate and the conditional and poisson accrual estimation methods maybe useful to researchers designing a complex, multi-center RCT. |
Anisimov | Yes | Checked how a poisson-gamma model fits real recruitment data by analysing several tens of completed GSK trials in different therapeutic areas. | Authors conclude that for a sufficiently large number of centres (N > 20) the poisson-gamma model is in good agreement with real data and can serve as a basic recruitment model. |
Moussa | No | Theoretical analysis only | Not applicable |
Williford | Yes | Analysed weekly patient intake data from 9 multihospital clinical trials coordinated by the Veterans Administration Coordinating Centre between 1975 and 1982. Used the data to review how mean patient intake varies over time within a trial, and whether the assumption of equal patient intake rates throughout a trial is appropriate. Used trial data to compare fit of poisson model to a negative-binomial model | Authors conclude: 1. poisson distribution did not always provide an acceptable fit, nor was the assumption of equal intake rates over three time periods in a trial acceptable. 2. There is some evidence that the negative binomial distribution might provide an acceptable fit for trials with more than 200 patients. However, this distribution does not lend itself to any pre-study prediction unless the distribution of changes in patient intake rates can be estimated and modelled |
Gajewski | Yes | To illustrate the proposed model the authors use data from the Kansas University DHA outcome study which is a single centre trial. They take enrolment dates of the first 41 patients recruited and use these to predict the time needed to recruit the remaining 309 patients. | Unclear, the authors don't discuss how long the trial actually took to recruit all patients and compare that to the time predicted from the model. They use the data to illustrate the model only. |
Abbas | No | Hypothetical examples presented | Not applicable |
Haidich | Yes | A retrospective analysis of database of all 782 clinical studies launched by the AIDS Clinical Trials Group between Oct 1986 and Nov 1999 to identify factors which affect recruitment (use simple regression and multivariate first-order autoregressive model). Model not applied to other trials. | Authors conclude modelling enrolment rates may be used to comprehend long-term patterns and to perform future strategic planning. |