Why the Protocol’s History Matters
Readers need to know not only what an observational study reported, but also what plans existed before the analysis was carried out and whether those plans changed. A dated protocol and a clear account of amendments make it easier to distinguish decisions made in advance from decisions made during or after analysis. That distinction helps readers assess the transparency of the research process; it does not, by itself, establish that the design is valid or that an estimated association is causal.
Registration and protocol reporting are relevant to propensity score and instrumental-variable studies as well as other comparative observational analyses. Registration is not universally required for every observational study, and registries differ in their purposes and fields. When a protocol, analysis plan, or registration record exists, report it precisely. When it does not, say so directly rather than leaving readers to infer that a plan was absent or unavailable.
The reporting task is to give readers enough information to locate the relevant materials, understand when the plans were recorded, identify the main planned outcomes and analyses, and follow any changes. This complements the broader reporting principles of STROBE for comparative observational studies, but it focuses specifically on the study’s documented plan and its history.
What to Report
For each protocol-related document or registration record, report its status and location. Useful details include the registry or repository name, registration identifier, direct access information, document title, version number, and date. If the record is public, provide a working link or a stable identifier. If a document is available only on request, explain how to request it and any restrictions. A statement such as “the protocol is available” is incomplete if readers cannot determine where it is available or which version is meant.
Also report the timing of registration in relation to study initiation. Give the registration date and, where relevant, the date study activities began, such as the start of data collection or analysis. Use dates that can be checked against records; do not describe a registration as prospective merely because it preceded publication. If registration occurred after the study began, label it as retrospective and state that timing plainly. If a registry record was created but the full protocol was written later, distinguish the dates and documents.
Summarize the outcomes and analyses that were specified in advance at a level that allows comparison with the report. Name the prespecified primary and secondary outcomes, if applicable, and identify the main planned analysis approaches. Do not imply that a brief registry entry is a complete analysis plan if it contains only a summary. If the SAP provides additional detail, identify it separately. The purpose is to reveal what was planned, not to repeat every later methodological detail in this section.
A plan can change for sound reasons: data availability may differ from expectations, a measurement may prove unusable, or a data-processing issue may require a revision. Report each consequential change, when it occurred, why it was made, and whether it was made before or after access to outcome data or results. State what the original plan said and what the analysis did instead. If a change affected only wording or formatting and not the study methods, it need not be treated as a substantive analytic amendment; describe the distinction where ambiguity is possible.
Identify the protocol, SAP, registry entry, and any amendments. Record each title, version, identifier, and access location.
Report when registration and documents were created, when the study began, and when any relevant changes were made.
Summarize prespecified outcomes and main analyses, then identify deviations that materially changed those plans.
State when materials are unavailable, incomplete, or retrospective, and avoid implying that registration certifies methodological quality.
Worked Examples
Worked Example: Reporting a Publicly Registered Protocol
Suppose a hypothetical research group compares two medication strategies using routinely collected records. The investigators finalized a protocol on 12 January 2025, registered the study on 20 January 2025, and began extracting study data on 3 February 2025. The registry provides a public identifier, and the full protocol and SAP are available in a repository with version numbers.
Check the timing. The registration date, 20 January, is earlier than the stated start of data extraction, 3 February. Thus it is accurate to report that registration preceded data extraction. It would not be accurate to claim, on these facts alone, that registration preceded every possible study activity; the report should name the milestone being compared.
Identify the records. The report should give the registry name and identifier, the registration date, and the repository location for the protocol and SAP. It should distinguish the protocol finalized on 12 January from the registry entry created on 20 January, rather than treating them as one document. It should also identify the versions available so readers can tell which materials were consulted.
Summarize the plans. The authors should state which outcomes and main analyses were prespecified, using the terms in the protocol or SAP. A concise report might say: “The protocol (version 1.0, finalized 12 January 2025) and SAP (version 1.0) are available at [repository location]. The study was registered on 20 January 2025 (identifier X). Registration preceded data extraction, which began on 3 February 2025. The protocol and SAP specify the outcomes and primary analyses reported here.” In a real report, the bracketed location and identifier must be replaced with usable details, and the named outcomes and analyses should be stated rather than left generic.
Conclusion. This account gives readers the dates and materials needed to inspect the plans. It does not treat a public registration as proof that the planned methods were adequate or that the analysis followed them without deviation.
Worked Example: Documenting a Change After Data Review
In a hypothetical comparison of two care pathways, the SAP dated 5 March 2025 specifies a primary outcome analysis using a particular recorded measure. On 18 June, after analysts discover that the measure is missing for many records in one data source, the team amends the plan to use a prespecified alternative measure. The team has already examined outcome data. The final report presents the alternative measure as its primary analysis.
A transparent account must include four facts: the original plan, the new plan, the date and reason for the change, and the timing relative to outcome-data access. For example: “The SAP dated 5 March 2025 specified [original measure] for the primary analysis. On 18 June 2025, after review of outcome data showed that this measure was frequently unavailable in one source, the team amended the plan to use [alternative measure]. The revised analysis was conducted after access to outcome data. The amendment and its date are documented in version 2.0 of the SAP.”
That wording does not hide the change by presenting the alternative measure as if it had always been planned. It also does not claim that the change was improper: the reason may be defensible, while the post-data-access timing remains important context for readers evaluating the analysis. If the amendment was not documented at the time, the report should say that it was not formally recorded then rather than assigning it a misleading earlier date.
Worked Example: Reporting No Public Registration
A hypothetical team conducts a retrospective comparison using an existing administrative database. It did not register the study and has no dated protocol or SAP. The investigators did prepare a methods outline after completing the analysis, but it does not document plans made before analysis.
The report should not call the later outline a prespecified protocol. A suitable statement is: “This study was not registered. No protocol or statistical analysis plan dated before analysis is available. A methods document describing the analyses is available at [location]; it was prepared after the analyses were completed and is not a record of prespecified plans.”
This wording distinguishes absence of a public registration from absence of a contemporaneous plan, and distinguishes both from the later methods document. If a protocol existed but cannot be shared, the authors should state that it exists, explain the access restriction, and give its date and version if known. Precise disclosure lets readers assess the documentation that is actually available without implying that registration was mandatory or that the study cannot be evaluated.
Why the Timeline Is Evidence About the Decision Process
The value of a dated plan can be understood by comparing two possible accounts of the same analysis. In the first, the outcome and analysis are specified before outcome data are examined. In the second, the analysis is selected after results are visible, but the report describes it as though it had been planned. The numerical analysis could be identical in both accounts; what differs is the evidence about when the choice was made. A dated record and a clear amendment history help readers distinguish these situations.
This reasoning does not mean that every post-initiation change is a flaw. A planned method may become infeasible, and a documented revision may be more appropriate than mechanically following a plan that no longer fits the data. The key reporting requirement is to preserve the sequence: what was planned, what was learned, what changed, and when. Without that sequence, readers cannot tell whether a result reflects a prespecified analysis, a justified adaptation, or an analysis chosen after observing results.
Common Pitfalls and Checks
- Giving an identifier without access details. A registry number alone may not locate the protocol or later versions. Provide the registry name and a usable link or stable identifier, and separately identify repository-hosted documents.
- Calling retrospective registration prospective. Compare the registration date with a clearly stated study milestone. If registration followed study initiation, describe it as retrospective, even if it preceded publication.
- Reporting only that the analysis followed the protocol. Readers need to know which outcomes and main analyses were prespecified and whether material changes occurred. A general assurance does not show what was planned.
- Omitting the reason or timing of an amendment. “The analysis plan was updated” does not say what changed, when, or whether outcome data had been reviewed. Supply those details where known.
- Treating registration as a quality certificate. A registry entry records information; it does not certify that a study is unbiased, that all assumptions hold, or that the analysis is correct.
- Using a later document to imply earlier planning. Label documents by their actual dates and versions. If a document was prepared after analysis, say so.
A practical consistency check is to compare the registry record, protocol, SAP, amendment history, and final report side by side. Differences in dates, outcome labels, or analysis descriptions should either be reconciled or explained. This check can reveal an unreported amendment, a version mix-up, or an overly broad claim that no changes occurred.
How Documentation Options Differ
| Record type | What it can show | Important limitation |
|---|---|---|
| Public study registry | A dated record of selected study information and, sometimes, updates | May be brief and may not include the full protocol or detailed SAP |
| Full protocol or SAP in a repository | Detailed planned methods, outcomes, analyses, and version history | Its date and version must be clear; a document created later cannot establish earlier plans |
| Journal supplement or protocol publication | Accessible methods documentation linked to a publication | Publication timing may be later than study initiation; amendments may be elsewhere |
| Internal or restricted document | May preserve a contemporaneous plan when public materials are unavailable | Readers may not be able to inspect it; report its existence, date, access limits, and version if possible |
These options can be used together. For example, a registry may provide a public timestamp while a repository supplies the detailed SAP. None removes the need to describe the timing of study initiation and material changes. A registration system also cannot replace clear reporting in the study report: readers should not have to reconstruct the protocol history from scattered files.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A study began data extraction on 10 April and was registered on 2 May. How should the timing of registration be described?
- A registry entry lists an outcome, but the detailed SAP lists an additional primary analysis. What should the report identify?
- An analysis plan changes after the team reviews outcome data because a planned measure is unavailable. Name the minimum details needed to report the change transparently.
- A retrospective study has no registry entry, but the authors wrote a methods document after analysis. Can that document be described as a prespecified plan? Explain.
- Why does a public registration not establish that the study’s methods are valid?
Solutions. (1) State that registration was retrospective relative to data extraction: it occurred on 2 May, after extraction began on 10 April. Do not call it prospective on the basis that it occurred before publication. (2) Identify the registry entry and SAP separately, state which analysis each records, and make clear that the additional analysis appears in the SAP; do not imply the registry contains the full plan. (3) Report the original measure and planned analysis, the replacement and revised analysis, the date and reason for the change, and that outcome data had already been reviewed. Identify the amendment document and version if available. (4) No. State that the document was prepared after analysis and is not evidence of a prespecified plan. (5) Registration records selected information and its timing; it does not verify the design, data quality, assumptions, implementation, or correctness of the analysis.