03 · What You Need to Know
Methodological Flexibility Is Not the Same as Methodological Convenience
Changing a Study Is Not Automatically a Methodological Failure
A protocol is a plan created under uncertainty. Researchers cannot always anticipate how recruitment, measurement, implementation, or external circumstances will unfold. Discovering a genuine problem after a study begins does not create an obligation to preserve that problem merely because it was prespecified.
Formal methodological frameworks recognize this reality. SPIRIT 2025 treats the trial protocol as a living document that may require formal amendments during a study. Adaptive trial methodology goes further by allowing specified features of a study to change in response to accumulating information under prospectively defined rules.
The existence of legitimate adaptation does not mean that every mid-study modification is equally defensible. The methodological question concerns how the decision was generated and what it does to the study.
Begin With the Problem, Not the Change You Want to Make
A defensible adaptation starts with a clearly identified problem. “We want to change the sample” is a proposed solution. “The accessible population is substantially smaller than the feasibility estimates used in planning” identifies the underlying problem.
Likewise, “we want to replace the questionnaire” is less informative than “participants consistently interpret a central item differently from the intended construct, making the resulting scores difficult to interpret.”
Separating problem from solution helps prevent researchers from adopting the first convenient modification without considering less disruptive alternatives.
Problem State what is preventing the original design from functioning as intended.
Consequence Identify what happens to validity, feasibility, participant protections, or inference if nothing changes.
Options Consider realistic responses, including continuing unchanged when that remains defensible.
Adaptation Select the option that best preserves the scientific purpose while addressing the problem.
The Revised Design Must Still Produce Evidence Worth Interpreting
An adaptation is not justified merely because it allows data collection to continue. The revised design must still have a credible scientific purpose.
Ask whether the modified study can answer the original research question. If not, can it answer a narrower or revised question that remains worthwhile? What population does the evidence now concern? What comparison remains? What outcome is actually being measured? What kind of inference can still be supported?
If the adaptation solves an operational problem while destroying the study's ability to answer a meaningful question, it has preserved the project rather than the research.
Prospectively Planned Adaptation Is Methodologically Different From Improvisation
Some designs specify in advance that particular features may change. FDA guidance on adaptive clinical trials describes adaptive designs as those allowing prospectively planned modifications to one or more aspects of a design based on accumulating data. The modification rules and relevant statistical procedures should be prespecified so that the validity and integrity of the study can be protected.
Prospectively planned adaptation
The features that may change, information used, decision rules, timing, and relevant safeguards are specified before the adaptation is triggered.
Unplanned adaptation
A change becomes necessary because an unexpected methodological, ethical, operational, or external problem emerges during the study.
An unplanned adaptation is not automatically invalid. It simply lacks some protections provided by prospective planning and therefore requires particularly clear reasoning and documentation.
Ask What Information Influenced the Decision
Timing alone does not tell you whether a change is problematic. What researchers knew when they made the decision can be more important.
Suppose recruitment is far below expectations, and the team decides to add another appropriate site without examining comparative outcome data. That is different from examining results, discovering that a particular subgroup produces an unfavorable effect estimate, and then narrowing eligibility so future recruitment favors another subgroup.
The second decision is outcome-dependent in a way that threatens the independence between the study plan and the evidence used to evaluate it.
Watch Out
The most difficult adaptations to defend are often those selected after researchers know which methodological choice produces the most desirable result. Changing outcomes, exclusion rules, subgroups, models, or populations after examining results requires particular scrutiny and should never be disguised as though it had been planned from the outset.
Prefer the Least Disruptive Change That Actually Solves the Problem
Methodological adaptation should generally be proportional to the problem. If recruitment is slow because appointment times are inaccessible, changing scheduling may be preferable to broadening the target population. If assessors are administering an instrument inconsistently, retraining may be preferable to replacing the measure.
This does not mean that smaller changes are always better. A minor adjustment that fails to address a fundamental problem merely postpones the decision. The relevant principle is to avoid changing more of the study than the problem requires while still solving the actual problem.
Assess What Happens to Data Collected Before the Change
A defensible adaptation must account for the evidence already generated. If a questionnaire changes, are scores from the two versions comparable? If eligibility broadens, do early and later participants belong to a coherent population? If a comparison condition changes, can observations from both periods contribute to the same contrast?
Sometimes all data remain usable. Sometimes the change requires indicators for study phase, stratified or sensitivity analyses, or a more cautious interpretation. In other circumstances, observations generated under the original and revised designs may not be meaningfully combinable for the intended purpose.
Do not decide that the adaptation is defensible only by looking forward. Its relationship with existing data is part of the methodological judgment.
Preserve the Original Research Question Unless You Intentionally Revise It
One useful test is to compare the revised design with the original research question before rewriting either one. If the revised design still answers essentially the same question, the adaptation may preserve the study's scientific identity.
If it answers only part of the question, the scope of the claim may need to narrow. If it answers a materially different question, acknowledge that change rather than altering the wording retrospectively to make the new study look inevitable.
When the proposed solution changes the central evidence or inference, consider whether changing the research method has effectively created a different study.
Methodological Justification and Ethics Approval Are Separate Tests
A change can be methodologically sensible and still require approval before implementation. Conversely, approval of an amendment does not establish that the resulting design provides strong evidence.
If an adaptation affects participants, recruitment, eligibility, consent, risks, interventions, measurements, privacy, or other approved activities, check the applicable ethics or institutional requirements. Changes can also affect protocols, registries, preregistrations, statistical analysis plans, sponsor requirements, or funder commitments.
Before implementing a substantive modification, determine which research records or approvals need to be revisited.
Transparency Does Not Make Every Adaptation Methodologically Sound
Disclosure is necessary for many consequential changes, but transparency alone does not rescue a weak decision.
A researcher could transparently report that participants were excluded after outcomes were examined because their inclusion weakened the desired result. The disclosure would be preferable to concealment, but the methodological decision would still be problematic.
Defensibility requires both a sound reason for the change and transparent reporting of what happened.
Several Defensible Adaptations Can Still Accumulate Into a Weak Study
Researchers often evaluate modifications individually: extending recruitment seems reasonable, replacing a malfunctioning measure seems reasonable, relaxing one criterion seems reasonable, and changing one analysis seems reasonable.
The cumulative study may nevertheless end up far from the original design. Each compromise consumes some of the study's ability to support its intended inference.
Periodically evaluate the revised study as a whole rather than asking only whether the latest amendment can be justified in isolation. This becomes especially important when considering how much methodological compromise the study can tolerate.
Document the Decision While It Is Being Made
A defensible adaptation should leave a contemporaneous record. Preserve the original plan and record the problem, alternatives considered, selected modification, rationale, decision date, implementation date, information available to decision makers, required approvals, and anticipated consequences.
Documentation is particularly important when the decision occurs after data collection begins because it helps establish whether researchers had access to information capable of influencing the choice.
Use a structured process to document unplanned research changes rather than reconstructing the reasoning only when the manuscript is written.
