The short version
- During an interim analysis, comparative results go to a data monitoring committee (DMC) and must not reach the people running the trial.
- Most interim problems are not statistical. They are access, data cut definitions and outputs that were never rehearsed on real data.
- Use this week to confirm the firewall, freeze the data cut rules and run the outputs once, blinded, end to end.
Why the firewall matters
ICH E9 explains why interim comparisons are handled with care: anyone who sees comparative results while a trial continues may, consciously or not, change how it is conducted, and the trial's integrity depends on keeping that knowledge to as few people as possible [1]. FDA's draft guidance on data monitoring committees, issued in 2024 to replace the 2006 final guidance once it is finalized, describes the committee's role and the separation of the statisticians who prepare its reports from the sponsor's study team [2].
For the study programming team, this translates into one rule: the study team prepares blinded programs and data; a separate unblinded team applies the randomization and produces the comparative outputs.
The checklist
Access
- Who holds the randomization codes, and who can read the unblinded datasets? Write down the names, not the roles.
- Are unblinded outputs stored where the study team has no access, including shared drives, email and the clinical data repository?
- Is there a log of who accessed what and when? Good clinical practice expects access controls proportionate to the risk, and an interim is high risk [3].
Data cut
- Is the cut-off defined as a date and a rule, for example "all visits up to and including the cut date, with data entered and cleaned by the transfer date"?
- Are external data, such as central labs or PK results, included up to the same cut, and do their transfers arrive in time?
- Has the statistician confirmed how partial data after the cut are handled?
Outputs
- Have the DMC outputs been run end to end on blinded or dummy-randomized data, with the same programs the unblinded team will use?
- Did someone outside programming read them for sense? A DMC meeting is the wrong place to discover a mislabelled column.
- Does the unblinded team have the run instructions in writing, with the expected number of outputs and checks?
Timing
- Count back from the DMC meeting: report delivery, unblinded run and QC, final transfer, last data cleaning. Which of these dates is closest to its limit?
- Who covers if a key programmer or statistician is unavailable that week?
A synthetic example
In the fictional ZEPHYR-101 study, the rehearsal on dummy-randomized data shows that the DMC safety tables take nine hours to run because one program re-derives the lab dataset. The unblinded statistician would have found out on the evening before the meeting. Moving the derivation into the analysis dataset build cuts the run to forty minutes and leaves a full day for QC.
Do this today
- Write down every person with access to randomization codes or unblinded outputs, and check it against the DMC charter.
- Confirm the data cut rule in writing with the statistician, including external data.
- Schedule an end-to-end rehearsal of the DMC outputs on blinded data before the real run.
Check yourself
One question
Show the answer
Answer 2. The point of the firewall is that the people running the trial cannot be influenced by comparative results. A separate unblinded team, with documented and limited access, protects the integrity of the trial; trusting people not to share is not a control.
Apply it
Turn it into a plan
The brief says what to check. The Navigator and the Handbook turn it into dated, owned actions for your milestone, in your browser, with no study data.
Sources
References
- ICH. E9 Statistical Principles for Clinical Trials. Step 4, 1998. www.ich.org ↗
- U.S. Food and Drug Administration. Use of Data Monitoring Committees in Clinical Trials. Draft guidance for industry; February 2024. www.fda.gov ↗
- ICH. E6(R3) Guideline for Good Clinical Practice. International Council for Harmonisation; 2025. www.ich.org ↗
Standards and guidance change. Confirm the version in force for your study, sponsor and region before relying on any reference. Examples are synthetic and illustrative. This brief is generalized professional guidance, not regulatory advice.
Next brief · Monday 10:00 ET · Standards Decoded