8.3 The Decision Log
What the decision log is for
The pre-specification documents—protocol, SAP, DSMB charter, SSR charter—record the decisions that were made before the trial began. The decision log records the decisions made during the trial: the unexpected events that required responses, the governance decisions about protocol deviations, the DSMB recommendations and the sponsor’s actions in response, the regulatory interactions that affected the trial’s conduct, the protocol amendments and their rationales.
The decision log is the audit trail for the trial’s conduct. It answers the question that every regulatory reviewer, every DSMB member, and every independent statistician will eventually ask: when a decision was made, what information was available, who made it, and why? Without the decision log, these questions can only be answered from memory, from email chains, and from meeting records that may be incomplete or selective. With it, the decisions are contemporaneously documented in a single record that is designed to answer exactly these questions.
The decision log is not a medical record. It is not a regulatory submission document. It is a governance record—the management record of the decisions that shaped the trial’s conduct between the moment the first patient was enrolled and the moment the database was locked. Its audience is anyone who later needs to understand whether the trial was conducted as designed and whether deviations from the design were handled according to pre-specified governance rules.
What the decision log must contain
The decision log has four categories of entry, each corresponding to a different type of decision made during the trial.
Protocol amendment entries. Every protocol amendment—every change to the pre-specified design—must be entered in the decision log at the time the amendment is implemented. The entry must record: what was changed, what the rationale for the change was, whether the change was driven by operational necessity or by information generated by the trial (and if the latter, what information), whether the change was reviewed by the regulatory agency before implementation, and whether the change required re-consent of enrolled patients. The entry must be specific enough that a reviewer can determine whether the amendment was pre-planned or responsive to trial data.
DSMB interaction entries. After every DSMB interim meeting—open session, closed session, or executive session—the decision log must record: what the DSMB reviewed (without revealing the arm-specific interim data), what recommendation the DSMB made, what information the recommendation was communicated to the sponsor, and what action the sponsor took in response. When the sponsor’s action differs from the DSMB’s recommendation—when the sponsor continues a trial against a non-binding futility recommendation, for example—the entry must record the sponsor’s documented rationale for the divergence.
Regulatory interaction entries. Every interaction with the regulatory agency that affects the trial’s conduct—a Type B or Type C meeting response, a clinical hold and its resolution, a complete response letter’s requests—must be entered in the decision log. The entry must record the agency’s specific request or guidance, the sponsor’s response, and the timeline of implementation. Regulatory interactions that are not recorded in the decision log cannot be traced when the regulatory reviewer asks, during the submission review, when a specific change to the trial’s conduct was made and why.
Operational deviation entries. Protocol deviations—departures from the pre-specified protocol that were not amendments—must be entered in the decision log when they are identified. The entry must classify the deviation (major or minor, by the pre-specified classification criteria), record the action taken in response, and note whether the deviation affects the primary analysis. Major deviations—those that affect the primary endpoint, the eligibility of the patient, or the blinding—require a more detailed entry that includes the assessment of the deviation’s potential impact on the primary analysis and the pre-specified handling of deviating patients.
When entries must be made
The timing of decision log entries is as important as their content. An entry that is made days after the decision, from reconstruction, is not the same as an entry made contemporaneously. The contemporaneous entry reflects what was known at the time of the decision; the reconstructed entry reflects what was known afterward, including the decision’s consequences.
The discipline for decision log entries is: the entry is made within a specified time window after the decision—24 to 72 hours for most decisions, immediately for DSMB recommendations and regulatory interactions. The time window must be specified in the decision log’s operating procedures, which are part of the trial’s governance documents. A decision log with no operating procedures—no specification of who enters what, by when, in what format—is a document that will be completed retrospectively, and its retrospective nature will be apparent to the reviewer who examines the entry timestamps.
For DSMB interactions, the entry timing is particularly critical. The DSMB closed session recommendation—the arm-specific recommendation that the DSMB makes after reviewing the unblinded interim data—must be recorded in the decision log immediately after the DSMB meeting, before the sponsor has time to consult internal advisors about the recommendation and respond. A recommendation entry made days after the meeting, after the sponsor has formulated its response, is an entry that reflects the sponsor’s considered response rather than the DSMB’s immediate recommendation.
Who maintains the decision log
The decision log must be maintained by a person or team that is independent of the sponsor’s analysis function—the team that will eventually analyze the primary endpoint data. This independence requirement mirrors the independence requirements established throughout this book for interim data access: the people who maintain the decision log must not be in a position to use the log to shape the primary analysis.
In practice, the decision log is most commonly maintained by the clinical operations team, with entries made by the study manager or the project manager for each category of decision. The clinical operations team typically does not have access to unblinded interim efficacy data—they see overall enrollment and safety data, but not the arm-specific primary endpoint comparison—so their independence from the primary analysis is maintained.
The analysis team’s role in the decision log is limited to reviewing entries that affect the primary analysis—major protocol deviations, endpoint adjudication clarifications, adaptive rule implementations—and confirming that the entries accurately reflect the decisions made. The analysis team does not write the entries; they review them for completeness and accuracy in matters that bear on the analysis.
The DSMB’s interaction with the decision log is also limited. The DSMB meeting minutes—which may contain arm-specific interim data—are not part of the public decision log. The entry in the public decision log records only the DSMB’s recommendation and the sponsor’s response, not the data on which the recommendation was based. The DSMB’s closed session minutes are maintained separately, available to the regulatory agency under confidential review but not part of the standard decision log.
The decision log and the clinical study report
At the end of the trial, the decision log becomes the primary source document for the clinical study report’s account of protocol deviations, amendments, and governance events. The clinical study report—the regulatory submission document that describes the trial’s conduct—must account for every major deviation, every amendment, and every DSMB recommendation. Without the decision log, this account is reconstructed from disparate sources; with it, the account is extracted from a single contemporaneous record.
The decision log also provides the evidence that the primary analysis was conducted according to the pre-specified plan. When the clinical study report states that the primary analysis was pre-specified in the SAP and conducted without modification, the decision log supports this statement by showing that no entries record a change to the analysis plan after the SAP was finalized. When the clinical study report states that the sponsor followed the DSMB’s recommendations at each interim, the decision log supports this by showing the recommendation and the sponsor’s response for each interim.
A clinical study report that cannot be supported by a contemporaneous decision log is a report that rests on reconstruction. The regulatory reviewer who examines the submission will ask for the source documents supporting the report’s account of the trial’s conduct. If the source documents are the decision log, the support is direct. If the source documents are email chains, meeting minutes from various committees, and declarations from the study team, the support is indirect and the reviewer will need to reconcile disparate sources to assess whether the trial was conducted as described.
The decision log as a governance discipline
The discipline of maintaining the decision log creates a specific kind of accountability that is otherwise absent from the trial governance system. When every decision is entered within 72 hours, by a specified party, using a specified format, the people making the decisions know that their decisions will be recorded—promptly, in the terms defined in the operating procedures, without the opportunity to frame the decision in a more favorable light after its consequences are known.
This accountability structure changes the decisions made during the trial. Decisions that would be embarrassing in the decision log—decisions that reflect commercial pressure on the DSMB, decisions to continue a trial past a futility boundary without a documented rationale, decisions to change a protocol element in response to an interim trend that was not supposed to reach the sponsor—are decisions that the decision log’s accountability structure discourages. The decision log does not prevent these decisions, but it makes them more visible and more consequential.
The goal of this visibility is not punishment. It is honesty: the trial’s conduct should be defensible from the inside—every decision traceable to a documented basis, every deviation handled according to pre-specified rules—not just explainable from the outside after the fact. The decision log is the mechanism by which the inside and the outside of the trial’s conduct are made the same.
References: ICH E6(R2) Good Clinical Practice (2016); FDA Guidance for Industry, Oversight of Clinical Investigations—A Risk-Based Approach to Monitoring (2013); Herson, Data and Safety Monitoring Committees in Clinical Trials (2009); Ellenberg, Fleming, and DeMets, Data Monitoring Committees in Clinical Trials: A Practical Perspective (2002).