- Posted By: Susan Dutton
Members from OCTRU attended this biannual event on the 22 April hosted by the Pragmatic Clinical Trials Unit (PCTU) in London.
The PCTU have carried out complex intervention trials with complex designs including early feasibility and pilot studies leading to more valuable trial designs. One key highlight was the mention of the “simsam” sample size add-on to STATA which can be used to create sample size for complex designs using simulation, developed by Richard Hooper at QMUL: http://webspace.qmul.ac.uk/rlhooper/simsam/.
Further points to note from the day were:
SAP guideline and checklist.
Discussion was undertaken of the minimum set of items to be included in the Statistical Analysis Plan checklist and guideline to finalise the wording around the items developed following 2 Delphi surveys and a consensus meeting. The final version of the checklist and Explanation and Elaboration document are currently being drafted and should be published later this year.
Validation of Statistical Packages – the experience from the Glasgow Clinical Trials Unit and how IT can support statisticians.
Statistical Programming Validation including version control of statistical programming using a freely downloadable package called SUBVERSION, together with a user interface called TORTOISE. This allows all previous versions to be stored with those actually used for reports to be tagged and unchangeable. The Statistics team are exploring the possibilities of this and will feedback their experiences.
Re-randomisation designs - where participants can be randomised into a study more than once.
This could potentially be useful in diseases where there are multiple episodes but it is difficult to recruit, with the potential to shorten the length of trials and therefore save on costs. The sample size in this case would relate to the number of observations rather than the number of participants. This method would only provide unbiased estimates of the treatment effect and correct type-I errors under certain conditions: patients are only eligible for re-randomisation when the follow-up period from the previous randomisation is complete; randomisation for the same patient is performed independently and the treatment effect is constant across randomisations. Analysis, either ‘Unadjusted’ (ignoring patient) or ‘adjusted’ (including patient) is likely to have at least the equivalent power to the standard two-arm parallel group trial under certain conditions, such as no systematic differences between patients who are re-randomised vs those who are not, and no systematic difference between randomisation periods). Research in this area is ongoing, with the first publication likely to be in print soon: Kahan BC, Forbes AF, Dore CJ, Morris TP. A re-randomisation design for clinical trials.
OCTRU will continue to attend these meetings and keep to up to date on the issues highlighted above.