Should You Preregister the Primary Outcome of Your Study?

Should You Identify Your Primary Outcome Before Collecting Data?

Suppose your study measures student achievement, motivation, engagement, and satisfaction. After analyzing the data, you discover that only motivation shows a statistically significant difference between groups. Can you present motivation as the study's primary outcome, even though achievement originally motivated the research?

This is one reason researchers distinguish primary outcomes from secondary and exploratory outcomes. When several measures are available, the decision about which result matters most can influence the apparent success of a study.

Preregistering the primary outcome establishes that decision before relevant results are known. The challenge is specifying it precisely enough to prevent selective reporting without unnecessarily restricting legitimate research.

Yes, for Confirmatory Studies With a Defined Primary Outcome

How to Preregister a Primary Outcome Properly

What Exactly Is a Primary Outcome?

A primary outcome is the outcome designated as the principal basis for evaluating a study's central research objective. In an intervention study, it is often the variable used to assess whether the intervention produced its intended effect. In other confirmatory designs, it may be the principal dependent variable or endpoint used to evaluate a specified association or prediction.

The primary outcome is not necessarily the most interesting finding, the variable with the largest observed effect, or the measure producing the smallest p-value. Its priority comes from the research question and the prospective study plan.

These classifications concern the intended evidential role of each measure. A secondary outcome can be scientifically important, and an exploratory finding may justify substantial further investigation.

Why Does Preregistering the Primary Outcome Matter?

Researchers often collect more information than they can reasonably emphasize in one publication. This creates opportunities for selective outcome reporting, particularly when choices about which variables to highlight are influenced by the results.

Imagine a study measuring five educational outcomes. If the researchers choose the most statistically favorable outcome after analysis and describe it as the original primary endpoint, readers may overestimate how strongly the evidence supports the stated hypothesis.

Prospective outcome specification helps distinguish the planned evaluation from additional findings. It also supports decisions about sample size, measurement quality, and statistical analysis.

Research on outcome-reporting discrepancies, including the work of Chan and colleagues comparing trial protocols with publications, has documented that outcomes specified in protocols are not always reported consistently in final articles. Such evidence supports the value of maintaining an accessible prospective record, although preregistration alone cannot guarantee complete reporting.

What Information Should Be Prespecified About the Outcome?

Naming the construct is not enough. A statement such as "the primary outcome is learning achievement" may leave several consequential decisions unresolved.

A reproducible specification should ordinarily address the following:

Where applicable, also specify the outcome's direction, aggregation across repeated measurements, treatment of incomplete scale responses, and the population or observations included in the primary analysis.

These details should align with the broader prespecified research design and analysis decisions.

Should You Preregister the Measurement Time Point?

Yes, when measurement timing affects which result is considered primary.

A study might assess achievement immediately after an intervention, four weeks later, and at the end of a semester. These measurements address related but distinct questions. Immediate performance may differ from retention over time.

Without a specified primary time point, researchers could choose whichever assessment produces the strongest result.

Identify the intended assessment occasion or time window and explain how deviations from that window will be handled. If a longitudinal trajectory rather than one occasion is the primary outcome, specify the relevant longitudinal estimand or summary measure.

Can a Study Have More Than One Primary Outcome?

Yes. Multiple primary outcomes may be justified when a study addresses several equally important dimensions of an intervention or phenomenon.

For example, an educational intervention might aim to improve both conceptual understanding and practical performance. Neither measure necessarily captures the full objective.

However, multiple primary outcomes introduce an important inferential question: what combination of findings will support the study's principal claim?

Researchers should clarify whether success requires improvement on every primary outcome, improvement on at least one, or another prespecified criterion. When multiple statistical tests are used, an appropriate approach to multiplicity may be necessary depending on the intended claims.

Methods such as Holm's procedure or other familywise error-rate or false discovery-rate controls may be relevant, but the choice should reflect the hypothesis family and inferential objective. There is no universally appropriate adjustment for every study.

Should the Primary Outcome Be Chosen Before the Sample Size Is Calculated?

For many confirmatory quantitative studies, the primary outcome should be established before finalizing the sample-size justification.

The required sample may depend on the expected variability, event rate, clinically or practically meaningful difference, or precision requirements associated with that outcome.

A sample-size calculation based on one variable may not provide adequate precision or statistical power for a different primary variable.

Accordingly, sample size and stopping decisions should be coordinated with the outcome definition and planned analysis.

Can You Change the Primary Outcome After Preregistration?

A primary outcome can sometimes be changed for a legitimate scientific or operational reason. For example, an instrument might become unavailable, a serious scoring defect might be identified, or new evidence might establish that the planned measure is unsuitable.

The implications depend strongly on when the decision occurs and what researchers already know.

A change made before collecting relevant outcome data is generally less vulnerable to outcome-driven selection than a change made after inspecting differences between study groups.

When changing an outcome, preserve the original registration, document the reason and timing, and explain the implications for the confirmatory claim. The principles for changing preregistered decisions apply directly.

In regulated clinical research, changes may additionally require registry updates, protocol amendments, ethics review, or other approvals. A transparent preregistration amendment does not replace those obligations.

What If the Primary Outcome Produces a Null Result?

A null or statistically nonsignificant result does not justify silently replacing the primary outcome with a favorable secondary result.

The planned primary analysis should generally be reported, subject to legitimate methodological limitations. Secondary findings can also be reported, but their evidential role should remain distinguishable.

A nonsignificant result does not necessarily demonstrate that an effect is absent. Interpretation depends on the estimated effect, uncertainty, measurement quality, study design, and whether the data are informative about effects of practical importance.

Do Qualitative and Exploratory Studies Need a Primary Outcome?

Not necessarily. The language of primary outcomes is particularly associated with confirmatory and intervention-oriented research.

Qualitative studies may instead organize inquiry around research questions, phenomena of interest, sampling principles, and analytic commitments. Some mixed-methods studies may have a primary quantitative endpoint alongside qualitative objectives.

Researchers should not impose an artificial primary outcome merely to complete a registration template. The appropriate approach is to specify the methodological commitments that genuinely structure the investigation, consistent with qualitative preregistration requirements.

Choosing a Primary Outcome for an Educational Intervention

This example illustrates how advance specification protects the intended interpretation. It does not imply that the selected assessment is valid merely because it was preregistered. The researchers would still need evidence supporting the instrument's suitability, scoring, and interpretation for the intended population.

Common Misunderstandings About Primary Outcome Preregistration

How to Decide What Your Primary Outcome Should Be

Begin with the study's central objective rather than the available measurement instruments. Ask which outcome would provide the most direct and credible evidence concerning that objective.

Next, evaluate whether the proposed measure captures the intended construct, whether its scoring is defensible, and whether the chosen time point reflects the effect you actually want to study.

Where feasible, seek methodological and measurement expertise before registration. A well-chosen primary outcome should be substantively meaningful, measurable with adequate quality, and compatible with the proposed analysis and sample-size justification.

Before Preregistering Your Primary Outcome

Frequently Asked Questions About Preregistering Primary Outcomes

Choose the Primary Outcome Before Knowing Which Result Looks Best

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.