- Posted By: Susan Dutton
An excellent talk by Andy Fisher from the MHRA highlighted Statistical Issues in MHRA Inspections related to data management and statistics.
Key points:
It is a legal requirement to follow the protocol and this includes the statistical analysis section and any details on interim analyses. Any changes to this including timings of interim analyses will need to be updated in an amended protocol.
All statistical documents, including randomisation and blinding schedules and any analyses are essential documents and therefore need to be part of the Trial Master File. The confidential documents should only be available to the relevant personnel and access restrictions should be tested and the tests recorded.
MHRA Inspections are interested in looking at the processes, if there are SOPs in place, are they being followed and is there documentary evidence that they are being followed.
The aim is to be able to reconstruct the trial from the SOPs.
The Inspectors are not statistical assessors, although they can request a statistical assessor to attend an inspection. This is most likely to occur in the case of serious breaches or suspected fraud.
The Inspectors are interested in:
• Appropriate training
• Statistical input into the protocol
• Quality control of the sample size calculation – and documentation
• The statistical analysis plan (either within the protocol or as a separate document)
• Randomisation: who created, stored and has access to this confidential information. Is there an audit trail?
• Analysis programming should follow a “good programming guide” :
o Are they fit for purpose?
o Does merging occur appropriately?
o Has the data extraction been validated? (maybe an IT SOP)
o Programs should be validated and the validation stored
o Use version control for interim as well as final analyses. Store any created log files and if programs are rerun, ensure they don’t overwrite previous versions
o Hard coding in final analyses should be minimised – detail how it will be dealt with in advance (either in Data management plan of SAP) and where it will be reported.
o A list of programs together with which data sets they are dealing with and their function can make it easier to inspect.
• Evidence of documentation of protocol deviations and compliance with study protocol and how is the information transferred to the statistician
• Suspected fraud
• Reconstruction of trial: Can the route from raw data/database through statistical programming to output to published results be reconstructed?
• Audit trails: data integrity – who, when, why - evidence of QC of audit trails required.
Andy’s final comment was that generally statisticians will be doing most of this as part of their day-to-day work in the trials, but may not be documenting it. And it is the documentation that is often poor and needs to be improved.
Many of the recent Inspection Findings relating to Data Management and Statistics should be obviated by following the OCTRU SOPs. However, the OCTRU Hub team will be reviewing the relevant SOPs to ensure that the SOPs are accurate and cover what is needed. The OCTRU Hub team will then start carrying out audits to ensure that they are being carried out satisfactorily. Training will be provided if required.