01 · The Question
What Should You Record When a Study Does Not Proceed Exactly as Planned?
A research problem appears during data collection. The team discusses it, agrees on a solution, updates a procedure, and continues. Months later, someone asks why the change was made.
The answer is scattered across meeting notes, emails, revised files, comments in a spreadsheet, and several people's memories. Nobody is entirely sure which participants were studied before the change, which protocol version was current, or whether the decision occurred before or after the team examined the relevant outcomes.
This is why documentation matters. The purpose is not bureaucratic completeness for its own sake. A useful record should allow someone to reconstruct what was originally planned, what actually changed, why the decision was made, when it occurred, what information informed it, who authorized it, and what consequences it had for the study.
03 · What You Need to Know
A Good Change Record Preserves Both the Original Study and the Revised One
Do Not Replace the Original Plan With the Revised Version
One of the most important documentation principles is simple: preserve history.
If a protocol changes from Version 1.0 to Version 2.0, Version 1.0 should not disappear. If a questionnaire is modified, preserve the earlier questionnaire. If the statistical analysis plan changes, retain the original plan. If eligibility criteria are broadened, keep a record of the criteria that applied before the amendment.
SPIRIT 2025 explicitly treats the trial protocol as a living document and recommends a transparent audit trail documenting protocol versions, dates, and changes. Its explanation and elaboration guidance recommends sequentially identifying and dating protocol versions and listing changes relative to the previous version together with their rationale.
The principle applies well beyond randomized trials. Without earlier versions, it becomes difficult to distinguish what was planned from what emerged during study conduct.
A Useful Change Log Answers More Than “What Changed?”
Simply recording “eligibility criteria revised” is rarely enough. A useful methodological record captures the context and consequences of the decision.
| Record |
What to capture |
Why it matters |
| Original plan |
The procedure, criterion, outcome, analysis, or other feature before modification |
Establishes the prespecified baseline |
| Problem identified |
What occurred or what new information prompted reconsideration |
Explains why adaptation was considered |
| Decision |
The exact modification selected |
Defines what changed |
| Rationale |
The methodological, ethical, operational, or external reason for the decision |
Allows the change to be evaluated rather than merely observed |
| Decision date |
When the team formally decided to make the change |
Establishes chronology relative to data collection and analysis |
| Implementation date |
When the revised procedure actually began |
Identifies which participants or observations fall before and after the change |
| Information available |
What study data or results the decision makers had seen |
Helps assess whether the decision could have been influenced by emerging outcomes |
| Decision authority |
Who proposed, reviewed, approved, or authorized the change |
Clarifies governance and responsibility |
| Affected study elements |
Participants, sites, instruments, outcomes, analyses, data, or procedures affected |
Supports later analysis and reporting |
| Formal actions |
Ethics amendments, registry updates, protocol versions, sponsor notifications, or other required actions |
Links the methodological decision with applicable oversight |
Not every minor operational adjustment requires an elaborate dossier. The level of documentation should be proportionate to the consequence of the change and the formal requirements governing the study.
Record the Decision Date and the Implementation Date Separately
These dates are often different and both can matter.
A team might identify a problem on 3 March, decide on a modification on 10 March, receive ethics approval on 25 March, and implement the revised procedure on 27 March. Collapsing these events into “changed in March” loses information needed to reconstruct the study.
The distinction becomes particularly important when determining which participants were studied under which version or what outcome information was available when the decision was made.
Record What Researchers Knew When They Made the Decision
Chronology is especially important for decisions that could be influenced by accumulating results.
Suppose the primary analytical method changes midway through a study. A future reader may reasonably want to know whether the decision occurred before anyone examined outcome differences or after the team knew which analysis produced a more favorable result.
The change log should therefore record the relevant information state where feasible. Examples include whether data were still blinded, whether only recruitment totals were available, whether outcome distributions had been examined, or whether comparative results had already been generated.
This does not require documenting every thought anyone had. It requires enough information to establish the methodological chronology of consequential decisions.
Separate the Reason for the Change From the Evidence Supporting the Decision
A rationale such as “recruitment was poor” may be true but incomplete. How poor? Relative to what expectation? What evidence showed that the problem was unlikely to resolve without intervention?
Where practical, link the decision to the evidence that motivated it: recruitment reports, device-error logs, missing-data patterns, documented changes in external policy, participant feedback, fidelity assessments, or other relevant information.
This makes the rationale auditable and reduces the temptation to reconstruct a cleaner explanation after the outcome is known.
Document Alternatives Considered When the Choice Is Consequential
For major changes, it can be useful to record realistic alternatives considered and why they were rejected.
Suppose a recruitment problem could be addressed by extending the recruitment period, adding sites, broadening eligibility, or accepting a smaller sample. Recording why one option was chosen can clarify the methodological trade-off.
This does not mean every research decision needs minutes worthy of a parliamentary committee. The point is to preserve enough reasoning for consequential adaptations that the decision can later be understood on its own terms.
Version the Documents That Actually Changed
A central change log is valuable, but it does not replace version control for the affected study documents.
If the protocol changes, create a new protocol version. If the questionnaire changes, identify the revised instrument version. If the statistical analysis plan changes, preserve the earlier plan and date the revision. If participant information or consent materials change, use the document-control procedures required by the relevant institution or ethics body.
SPIRIT 2025 recommends version identifiers, dates, and explicit descriptions of changes to facilitate trial conduct, review, and oversight. This reduces the risk that different sites or team members unknowingly use different procedures.
Make It Possible to Identify Which Data Were Generated Under Which Version
If a methodological change occurs during data collection, your dataset or study records may need enough information to distinguish observations generated before and after the change.
For example, if a questionnaire changes on 1 June, you should be able to determine which participants completed the earlier version and which completed the revised one. If a device is replaced, record which device or version generated each relevant observation where appropriate. If eligibility criteria change, preserve the rule under which each participant entered the study.
This information can become essential when assessing comparability, conducting sensitivity analyses, or explaining heterogeneity later.
Do Not Alter Raw Data to Make Them Conform to the Revised Plan
A methodological change can alter how data should be interpreted or analyzed, but it should not rewrite the observations that were actually collected.
Good Clinical Practice principles emphasize traceable records and data changes. More generally, preserve original observations and use documented, reproducible transformations or derived variables rather than silently replacing source values.
Watch Out
Never overwrite earlier data, instrument versions, protocol files, or analysis plans merely to make the study appear internally consistent after a change. Consistency created by erasing history is not methodological consistency.
Link the Change Log to Ethics and Other Formal Actions
If a modification requires ethics approval, a protocol amendment, registry update, sponsor authorization, or another formal action, record the relevant reference, date, and status.
For example, a change entry might state that the decision was proposed on one date, submitted to the ethics committee on another, approved later, and implemented only after approval. This separates the scientific decision from authorization to put it into practice.
If you are uncertain which formal actions apply, determine whether the change requires revisiting ethics approval, the protocol, or preregistration before implementation.
Preregistration Should Preserve the Original Commitment
Preregistration is valuable partly because it establishes when a research decision was documented. The Center for Open Science describes registrations as time-stamped records and notes that preregistration is normally submitted before data collection or analysis.
OSF also emphasizes that preregistration is not intended to lock researchers permanently into a plan. Significant changes can be documented, and deviations can be acknowledged when results are reported. The key is that later changes should not obscure what was originally prespecified.
If the platform supports updates or linked registrations, use its current procedures. Do not alter the historical record in a way that removes the chronology.
Internal Documentation and Publication Reporting Are Not the Same Thing
Your internal change log should generally be more detailed than the eventual manuscript. It exists to preserve the methodological history from which publication reporting can later be constructed.
The final article should communicate deviations that readers need in order to understand the design, assess potential bias, distinguish prespecified from subsequent decisions, and interpret the findings. It does not necessarily need every operational detail recorded internally.
When preparing the manuscript, use the change log to determine which methodological deviations need to be reported and where they belong.
Documentation Should Begin When the Problem Appears, Not When the Paper Is Written
Retrospective reconstruction is vulnerable to forgotten details and hindsight. Once the study outcome is known, reasons that were uncertain at the time can also begin to look more obvious than they actually were.
Record consequential problems and decisions contemporaneously. A short dated entry created when a decision occurs can be more methodologically valuable than a polished paragraph reconstructed eighteen months later.
Your future self will also appreciate this. Few research experiences are quite as humbling as discovering that “final_protocol_REVISED_v2_reallyfinal.docx” was apparently not, in fact, final.
07 · A Quick Checklist
What to Record for Every Consequential Unplanned Change
For each important research change, record:
The original procedure, criterion, instrument, outcome, analysis, or other plan before the change.
The problem or new information that prompted reconsideration.
The exact change that was proposed and ultimately adopted.
The methodological, ethical, operational, or external rationale for the decision.
The date the problem was identified, the decision date, and the implementation date where these differ.
What relevant study data or results were available to the decision makers at the time.
Who proposed, reviewed, and authorized the change.
Which participants, sites, observations, instruments, datasets, or analyses are affected.
Any ethics approval, protocol amendment, registry update, sponsor action, or other formal authorization associated with the change.
The expected implications for data comparability, analysis, interpretation, and eventual reporting.
