Protocol Deviations: What Actually Counts and How to Report Them

A gap between your registered protocol and your final reported methods is not automatically a problem -- reviews legitimately evolve as real-world circumstances emerge during conduct -- but knowing what genuinely counts as a reportable deviation, and disclosing it properly, is what separates defensible methodological adaptation from an undisclosed change that undermines your review's credibility.

Why deviations happen even in well-planned reviews

Search strategies sometimes need adjustment once initial results reveal an unanticipated volume or an unexpected gap in coverage. Eligibility criteria occasionally need clarification once screening surfaces genuine edge cases the original protocol didn't anticipate. Planned analyses sometimes become infeasible once actual data availability across included studies becomes clear. None of these are inherently problems -- they're normal features of conducting real research under genuine uncertainty.

What distinguishes a defensible deviation from a problematic one

A defensible deviation is disclosed explicitly, explained with clear reasoning, and doesn't appear to have been made specifically to produce a more favorable result. A problematic deviation is either undisclosed entirely, discovered only by a reader comparing your registered protocol against your reported methods, or appears timed and reasoned in a way suggesting it was made after seeing preliminary results specifically to improve how those results look.

Common categories of legitimate deviation

Adding a database to your search after discovering your initial search strategy missed a relevant source is a defensible, disclosable deviation, provided you document when and why the addition occurred. Clarifying an ambiguous eligibility criterion once screening reveals a genuine edge case your original wording didn't anticipate is defensible, provided the clarification is documented and applied consistently going forward. Switching from a planned meta-analysis to narrative synthesis after discovering your included studies are too heterogeneous to pool meaningfully is defensible, provided this decision is explained with reference to your actual heterogeneity findings rather than left unexplained.

What generally is not defensible

Changing your primary outcome after seeing that a secondary outcome shows a more favorable or more statistically significant result is a serious, difficult-to-defend deviation, since it closely resembles outcome switching, a well-documented form of reporting bias. Silently narrowing your eligibility criteria after screening reveals an unexpectedly large volume of records, without documenting this change and its rationale, undermines the transparency registration is specifically meant to provide.

Using PROSPERO's amendment feature properly

PROSPERO allows registered protocols to be formally amended, and using this feature actively, rather than treating your initial registration as fixed and then quietly deviating from it in your final manuscript, is the correct way to handle a genuine, well-justified mid-review change. Each amendment should include a brief, dated explanation of what changed and why.

Reporting deviations in your final manuscript

Beyond the registry's own amendment log, your methods section should explicitly note any material differences between your originally registered protocol and your final conducted methodology, with brief reasoning for each. This might be a short dedicated paragraph or a supplementary table listing each specific change, its timing, and its justification, depending on how many deviations occurred and how substantial they were.

How reviewers and readers actually check for this

An experienced methods reviewer will frequently pull up your registered protocol directly and compare it against your reported methods section, specifically looking for undisclosed discrepancies. This comparison is often one of the first, most mechanical checks performed, precisely because it's a relatively quick way to assess whether a review's reported methodology can be trusted as an accurate account of what was actually done.

A practical habit throughout your review

Keeping a running, dated log of any decision that departs even slightly from your original registered protocol, updated in real time as your review progresses rather than reconstructed from memory at the writing stage, makes honest, complete deviation reporting considerably easier and more accurate than trying to recall every methodological adjustment only once you're drafting your final manuscript months later.

Amendments versus corrections

It's worth distinguishing a genuine protocol amendment, reflecting a deliberate methodological adjustment made during conduct, from a simple correction of a typographical or clerical error in the original registration that doesn't reflect any actual change in how the review was conducted. Both should be documented, but they carry different implications for how a reader should interpret the change, and conflating the two in your reporting can create unnecessary confusion about whether your methodology genuinely evolved or a minor administrative error was simply fixed.

Deviations discovered only at the writing stage

If you discover, only while drafting your final manuscript, that your actual conduct diverged from your registered protocol in some way you hadn't been actively tracking, disclosing this honestly at that point, even though it's later than ideal, is still considerably better practice than omitting it simply because it wasn't caught and logged in real time during the review's conduct. This honest, if imperfect, disclosure is still considerably more defensible than a reader discovering the same discrepancy independently, without any acknowledgment from the authors at all. Readers and reviewers alike generally respond far more favorably to honest, if late, disclosure than to silence, which is worth remembering if you're ever tempted to simply leave a discovered discrepancy unaddressed rather than disclosing it directly. This principle applies as much to a small, seemingly minor discrepancy as to a more substantial one, since readers generally extend considerably more trust to authors who disclose proactively than to those whose omissions are later discovered independently. Cultivating this instinct toward proactive disclosure, rather than hoping a discrepancy goes unnoticed, is ultimately what earns a research team lasting credibility within its field over a longer career. This reputation for transparency, once established, tends to make future collaborators, editors, and reviewers considerably more receptive to a research team's subsequent work as well, strengthening a research team's standing within its field over the course of an entire career, reinforcing why honest disclosure is best treated as a habit rather than a case-by-case judgment made under pressure, particularly in the difficult early moments of a project when staying quiet still feels like the easier, safer path.