Why the Clinical Study Report Must Describe What Actually Happened
A clinical study report (CSR) should let a reader reconstruct the adaptive design as it was carried out, not merely find the design that was intended. In an adaptive trial, the final participant count, treatment allocation, set of active arms, or timing of an analysis may depend on decisions made during the study. Reporting only the original plan can therefore leave a reader unable to tell which design produced the reported results.
The report should distinguish what was prespecified, what decision was made, and what was implemented. It should also explain when the decision was based on information available, who or what body made it, and why the action was taken. The earlier material on recording adaptation decisions and their audit trails addresses the contemporaneous record; the CSR should give a coherent account that is consistent with that record and with the final analyses.
A clear account is not a claim that every action was valid simply because it was documented. The report should provide enough information for a reviewer to assess whether the decision followed the applicable rules, whether the implementation matched the decision, and what consequences may matter for interpretation. In particular, an adaptation can affect the amount or structure of the data, the population exposed to a treatment, or the assumptions behind an analysis. Claims about error control should be consistent with the design and analysis actually used, not just the design originally proposed.
A Reconciliation That Readers Can Follow
For each adaptation, a reader should be able to follow a chain from plan to execution: the rule and its source, the information cutoff, the decision, the action taken, and the resulting trial conduct. Dates and participant counts make that chain auditable in the CSR. If a planned branch was not selected, say so where relevant; if a change was considered but rejected, do not describe it as though it were implemented.
The following information is often useful to report, adapted to the trial and the decision:
- The planned rule, its location in the protocol or analysis plan, and the version in force at the time.
- The data cutoff or information used for the decision, the date of review, and the date the action took effect.
- The decision authority and whether the decision makers had access to unblinded treatment comparisons.
- The actual action, including its scope, effective date, and effect on enrollment, randomization, follow-up, or analysis.
- The rationale, stated in relation to the prespecified criterion and the information available when the decision was made.
- Any difference between the plan, decision, and implementation, with a concise account of its potential consequences.
A useful summary table can make the reconciliation visible without replacing the narrative needed to explain important details.
| Reporting element | Question the CSR should answer |
|---|---|
| Prespecified plan | What action was permitted or required, and under which rule? |
| Information and timing | What information was available, at what cutoff, and when was it reviewed? |
| Decision and rationale | What was decided, by whom or under what authority, and why? |
| Implementation | What actually changed, for whom, and from what date? |
| Interpretation | What aspects of the design, analysis, or results may be affected? |
Worked Example: Reporting a Prespecified Sample-Size Increase
Suppose a hypothetical two-arm trial prespecified an interim assessment after 120 participants had completed the primary outcome. The protocol allowed an increase from 240 to a maximum of 360 participants if a blinded estimate of the outcome standard deviation exceeded 17 units. The study team received a blinded estimate of 18 units at the interim cutoff. The authorized decision was made on 12 March to increase the target to 320, and enrollment under the new target began on 18 March.
A report should state the planned criterion and the actual result against it: the blinded estimate was 18 units, which exceeded the threshold of 17 units. It should distinguish the decision date (12 March) from the effective date (18 March) and identify which sample-size target applied before and after implementation. It should also state who authorized the action and whether the decision makers remained blinded, consistent with the trial records.
The increase was 80 participants over the original target. Relative to the original target, this was \(80/240=0.3333\), or about 33.3%. The report should say that the target rose from 240 to 320—not that it rose “by 33.3 participants” or that 320 participants were added. It should then report actual enrollment, for example, 160 participants per arm and 320 in total if that is what occurred.
Finally, the report should identify which prespecified analysis and error-control approach applies to the implemented design and describe any analysis changes, if any. The fact that the increase met a stated rule is not by itself evidence that every statistical property is preserved; that claim must be supported by the design and analysis documentation. If the final enrollment stopped at 314 rather than 320, the CSR would report 314 as the actual number and explain the shortfall rather than presenting the target as the achieved sample size.
Explaining the Rationale Without Rewriting History
A rationale should be contemporaneous in substance: it should explain the decision using the information and rule available at the time. Later knowledge of treatment outcomes must not be presented as though it motivated an earlier decision. Distinguish the trigger for the action from the broader scientific reason for having that rule. For example, an observed nuisance-parameter estimate may have triggered a sample-size rule, while the design rationale for the rule was to protect the planned precision under plausible variation.
A report should also avoid implying that an adaptation was selected because it “worked” when that conclusion depends on results observed after the decision. If the action was not prespecified, or if the available documentation does not establish that it followed the rule, describe the facts plainly. The report is a record of conduct and evidence, not a justification constructed from the final outcome.
Worked Example: Reporting a Prespecified Arm-Dropping Decision
Consider a hypothetical three-arm trial with two experimental doses and a control. The plan allowed the lower-dose arm to stop for futility at an interim review if its prespecified predictive probability of meeting the trial’s success criterion was below 10%. At the data cutoff, the independent review committee’s calculation was 7%. The committee recommended stopping that arm, and the sponsor implemented the recommendation four days later. At the cutoff, 45 participants had been assigned to the lower-dose arm, 46 to the higher-dose arm, and 45 to control.
The CSR should report the criterion, the value compared with it, the committee’s role, the decision date, and the implementation date. It should state that \(7\%<10\%\), so the stated futility condition was met. The actual enrollment at the cutoff was \(45+46+45=136\) participants. The report should then give the subsequent participant flow by arm, making clear whether any participants entered the lower-dose arm during the four-day interval and how they were handled.
The rationale is that the prespecified futility boundary was crossed—not that the dose was proven ineffective. Stopping an arm on this basis does not establish that its true treatment effect is zero. The CSR should describe how the arm’s participants contribute to the analyses and whether the planned estimand and analysis remain applicable. It should also explain any implications of the arm’s early closure for the amount of information available, without treating smaller enrollment as evidence of a particular treatment effect.
Timing, Departures, and Their Interpretive Consequences
Timing is part of the design description. The planned information fraction or review point and the actual cutoff should be stated separately. A delay in review, an altered data cutoff, or implementation after a decision may matter if it changes the information available, permits additional enrollment under a previous allocation, or affects masking. A date alone is not enough when the number of participants or outcome events at the cutoff is also relevant.
When conduct departs from the plan, describe what differed and what is known about its consequences. Do not silently relabel an unplanned action as a prespecified adaptation. Conversely, do not imply that every operational difference invalidates the trial: whether it affects interpretation or error control depends on the specific change, the data available, and the analysis. The report should state whether the primary analysis followed the planned method, identify any alternative analyses used to assess the impact, and avoid claiming that a concern is resolved unless the evidence supports that conclusion. Detailed documentation of deviations from prespecified rules is a separate subject; here the essential point is that the CSR must not obscure a departure relevant to understanding the results.
Worked Example: Reporting a Delayed Interim Review
Suppose a hypothetical protocol planned a review after 100 of 200 participants had completed the primary outcome. Data cleaning delayed the usable cutoff by two weeks. The review ultimately used 108 completed outcomes, and the decision was to continue enrollment without change. The planned information fraction was \(100/200=0.500\), or 50.0%; the actual fraction was \(108/200=0.540\), or 54.0%. The difference was \(54.0\%-50.0\%=4.0\) percentage points.
The CSR should report both the planned and actual cutoff, the two-week delay, its documented operational reason, the number of completed outcomes available, and the decision made. It should also explain whether the additional information was available to anyone involved in trial conduct or treatment assignment, and whether the prespecified decision rule was applied as written. If the rule or analysis was affected by the later cutoff, the report should identify that issue and explain its possible implications for interpretation and error control. If no change was made, say so explicitly; “reviewed” does not mean “adapted.”
Justifying a Complete Account
A concise way to check a report is to reconcile each adaptation across three states: planned, decided, and implemented. These states can differ. A rule may allow an action, the decision authority may choose it, and operational implementation may begin later or differ in scope. Because the trial’s observed data arise from what was implemented, reporting only the plan or decision leaves a gap between the design description and the study results.
State the relevant rule and the governing document version, so the reader knows what was authorized at the time.
Give the cutoff, review date, decision authority, and information access relevant to understanding the decision.
Report the action, effective date, and resulting participant or study conduct, rather than assuming the decision and implementation were identical.
Connect the change to analyses and interpretation, and state what cannot be concluded from the adaptation alone.
This reconciliation matters because the interpretation of an estimate depends on the design that generated the data. For example, a changed sample size affects the amount of information; a changed allocation can affect the numbers contributing to each treatment comparison; and a delayed review can affect which data informed a decision. None of these facts alone determines whether the final inference is valid. The statistical justification must match the actual conduct, and limitations should remain visible.
Common Reporting Pitfalls
- Reporting the target as the achieved enrollment. A target of 320 does not mean 320 participants enrolled. Give the target and actual count separately.
- Collapsing decision and implementation dates. State when the decision was made and when the change took effect, especially if conduct continued in between.
- Calling a considered action an implemented adaptation. Report actions considered but not taken as such, if material to understanding the decision process.
- Using final outcomes to explain an earlier rationale. Distinguish the information available at the decision from evidence observed later.
- Equating a rule-following action with guaranteed validity. Explain how the analysis and error-control argument apply to the design actually used; do not infer validity from documentation alone.
- Omitting a departure because the outcome was unchanged. A departure can matter to interpretation even when the decision or headline result is the same.
Check Your Understanding
Answer each question, then compare your reasoning with the brief solutions.
- A trial planned to enroll 300 participants, increased its target to 360, and enrolled 352. Which numbers should the CSR report as the original target, revised target, and actual enrollment?
- A prespecified futility threshold was 15%, and the interim value was 12%. What can the report conclude about the rule, and what should it avoid claiming about the treatment?
- A committee made a decision on 4 May, but the change took effect on 9 May. Why should both dates be reported?
- A planned review was based on 80 completed outcomes, but the actual review used 84. Name two details the CSR should report beyond simply stating that the review was delayed.
Solutions: (1) Report 300 as the original target, 360 as the revised target, and 352 as actual enrollment; do not substitute a target for the achieved count. (2) Since \(12\%<15\%\), the stated condition was met. This does not prove the treatment has no effect. (3) The dates distinguish the authorized decision from the start of implementation and reveal whether conduct continued under the prior design in the intervening period. (4) Report the actual number of completed outcomes and the actual cutoff or review date; also explain the reason for the delay and whether the rule was applied as written or its implications for the analysis.