Amendments for additional statistical analysis / Reasons in FSR for differing statistical methods
Reply to Thread
GL9 states that 'Reasons should be given if the statistical methods used
differed from those proposed in the study protocol'. Some interpret this to be 'differences' to the actual 'statistical method' used for a specific analyses (but the analysis remaining the same otherwise) and an amendment being required for any additional analysis (esp where it impacts the results and conclusion of the study). For this interpretation, different statistical methods applied need to be explained in the FSR and an amendment is not strictly required.
Others interpret the above GL9 statement as meaning that amendments are not required if additional (or different) statistical analysis are applied (regardless of whether they impact the conclusions) - as long as the FSR gives reasons. What's your take?
VICH GL09 does state in section 18.104.22.168 : Reasons should be given if the statistical methods used differed from those proposed in the study protocol.
When the statistical method used is a planned change it should be handled as per the definition of an amendment. (Section 1.30 a study protocol amendment is a written change or modification of the study protocol effected prior to the implementation of the protocol or execution of the changed or modified task.) As stated in section 3.2.4 in the guideline each study protocol amendment, should identify what has been changed or modified and the reasons for such change of modification.
When the statistical methods were changed and only noticed after the fact it will need to be handled as a deviation to the study protocol. As stated in section 1.31 of the guideline study protocol deviation should be recorded as a statement signed and dated by the investigator describing the deviation and the reason for it occurrence.
Section 22.214.171.124 require that a description of any protocol deviations and / or amendment and an assessment of their impact on the outcome of the study must be described in the FSR. It is recommended that the reason be provided here.
When additional statistical methods are used and the data included in the FSR it will need to be handled as above.