Is it appropriate to initiate a substantial amendment after the study completion which has been declared to the competent authorities, i.e. the status of the trial is more than 90 days after Last Patient Last Visit?
Example 1: Assuming that you want to change the primary efficacy analysis applying a model-based analysis instead of a nonparametric test in a double-blind randomized clinical trial with longitudinal data. One reason might be that you are faced with a higher proportion of study dropouts than expected during the planning stage. A modeling approach now seems more suitable to deal with these methodological challenges.
Example 2: Suppose you wanted to modify the main analysis of a multi-arm trial (e.g. placebo/low/high dosage) by replacing the closed testing procedure specified in the protocol with a hierarchical testing strategy (clinically justified).
Is it advisable to justify these changes only in the *Statistical Analysis Plan* which is finalized before breaking the blind?
What do you think?
This is a protocol modification but the point is, does this modification effects the safety of the patients in the trial? If it does than an amendment submission to the ethical committee and regulatory authority is mandatory. In this situation this is not valid. So, only an amendment submission to the local ethical committee is enough.
Dear Christine, two points to be considered: first, the assessment by the research ethics committee and the free and informed consent must be modified; then regarding adjustemnt in the statistical part hummmm ... then I have my doubts. The statistical should complement the study, but never modify it during or even after the trial. For this, application of a test of statistical power will guide the "n" to be included in the search.
Dear Christine, changing the primary analysis of the study can only be done prior to database lock in a randomized double-blind controlled study. In an open-label study, however, you might be accused of manipulating the study outcome by changing the primary analysis after the study was over! In any case you have to amend the protocol since the primary analysis is described there. Ethics committee aproval of that amendment appears required. In most cases, however, you would not need to adapt the patient information and informed consent form, which is good for you since the patients may be hard to track after they completed the study. Good luck!
Any such change must be done prior to code break (unblinding) of group allocation. As noted above, this is unlikely to have any impact on patient safety and as such could be done by way of amending the Statistical Analysis Plan (SAP) or by way of an amendment to the protocol (and the statistical analyses description therewith). There are many situations where this could be performed and reference to the E9 (E9 Statistical Principles for Clinical Trials which although these were published in 1998 are still the only "statistical guidelines" that i am aware of) may provide some further insights as to acceptable rationale for such a proposed changes. As an example if "imbalances" or unexpected variable differences were identified these could require a change to the analyses plan. More critical would be an actual change in the "primary endpoint" that is beyond the scope of current question but would raise significant issues without significant discussion with the relevant regulatory authorities.
Any changes to methods should be addended to the protocol, statistics included. In addition, the final manuscript (if planning on publishing) should note a change in SAP and the reason (i.e. increased drop-outs, lost to follow-up, decreased events...) Finally, blinding can be maintained throught the statistical analysis (a metholodogy that some departments are using to decrease bias). Simply analyzing group A vs. B and unblinding after final analysis. This may require an unblinded data person, but is very much achievable. I agree, changing statistical methods after unblinding may be viewed an manipulation of data (even if not intentional). That being said, post-hoc sensitivity analyses or reviewer suggested additional analyses are common without protocol changes because they do not change the primary analysis.
In short, you don´t need a protocol amendment, but a SAP revision with clear documentation that the change was implemented before unblinding.
data can be changed after close out before sending it to SAS, when SAS is done they put it in database and lock it
- Badges
- Science topic
- Similar topics
- Correction