What If Your Control or Comparison Group Does Not Work as Planned?

What If the Group You Planned to Compare Against Is No Longer the Comparison You Intended?

A study can recruit enough participants, collect its planned outcomes, and still develop a serious design problem: the control or comparison group does not function as intended.

Participants assigned to a control condition may gain access to parts of the intervention. Usual care may differ substantially across sites. A comparison group may receive an alternative program that is more intensive than anticipated. Participants may switch conditions, providers may not deliver the comparison consistently, or an external change may alter what the control condition actually means.

When this happens, the problem is not merely that the protocol was imperfectly followed. The comparison itself helps define what effect or difference the study can estimate. You therefore need to establish what each group actually experienced and whether the resulting contrast can still answer the original research question.

Do Not Assume the Original Comparison Still Exists

The Comparison Group Is Part of the Question Your Study Answers

A Control Group Is Not Simply the Group That Did Not Receive the Main Intervention

The meaning of an intervention effect depends partly on what it is being compared with. “Intervention versus no treatment,” “intervention versus usual care,” “intervention versus placebo,” and “intervention versus another active intervention” are different comparisons and may answer different questions.

The same principle extends beyond randomized trials. Observational studies may compare exposed and unexposed participants, people receiving different naturally occurring treatments, different settings, different cohorts, or groups defined by another relevant characteristic. Whatever the design, the validity of the comparison depends on what those groups actually represent.

This is why a problem with the comparison group cannot be evaluated simply by asking whether the group still exists. You need to ask whether it still provides the contrast required by the research question.

Identify What “Did Not Work” Before Deciding What to Do

Comparison-group problems can arise in several ways, and they do not have identical consequences.

The diagnosis should come before the remedy. “The control group failed” is too broad to determine an appropriate response.

Distinguish Contamination From Non-Adherence

Two problems that may look similar can arise through different mechanisms.

There can also be poor fidelity at the provider or study level, meaning that the intervention or comparator itself was not delivered according to protocol. These phenomena can coexist, but distinguishing them helps identify where the design broke down.

CONSORT and SPIRIT guidance distinguishes adherence by participants from fidelity in implementation and emphasizes documenting what was actually delivered. This is important because the interpretation of the findings depends on the difference that truly existed between conditions, not merely the labels originally assigned to them.

“Usual Care” Can Change During the Study

Usual care is particularly easy to treat as though it were a stable, self-explanatory control condition. It often is not.

Clinical or institutional practices can vary among sites, providers, and participants, and they can change while a study is underway. New policies, technologies, professional guidance, or local practices may alter what participants in the comparison group receive.

CONSORT 2025 specifically notes that usual care may vary substantially across sites, patients, and the duration of a trial. Researchers should therefore establish what was actually provided rather than assuming that the phrase “usual care” adequately describes the comparator.

This issue is not confined to clinical research. An educational study comparing a new learning intervention with “regular teaching,” for example, needs to know what regular teaching consisted of. If instructors in the comparison group independently adopt similar techniques, the planned contrast may narrow considerably.

A Weakened Contrast Can Change the Meaning of the Result

Suppose an intervention and control group produce very similar outcomes. One interpretation is that the intervention provides little additional benefit over the comparator. But suppose the control group received many elements of the intervention through contamination. The observed difference then represents a different contrast from the one originally planned.

This does not authorize researchers to dismiss an inconvenient result. Rather, it means that fidelity, adherence, contamination, concomitant interventions, and actual comparator delivery become relevant to interpretation.

In some studies, contamination may bias the observed difference toward a smaller contrast. In others, the direction and magnitude of the effect may be less predictable. Avoid assuming that every comparison-group problem necessarily makes the intervention look weaker.

Do Not Automatically Remove Participants Who Did Not Follow Their Assigned Condition

In randomized trials, excluding participants after randomization because they did not adhere to their assigned intervention can compromise the benefits of randomization and introduce bias. CONSORT guidance therefore emphasizes reporting participant flow, deviations, adherence, and the numbers included in analyses.

Alternative analyses based on treatment received or adherence may sometimes answer useful secondary questions, but they do not automatically replace the analysis appropriate to the original randomized comparison. The appropriate approach depends on the estimand, study design, reasons for deviation, and prespecified analysis strategy.

Ask Whether the Problem Can Be Corrected Prospectively

If the study is still underway, some causes may be addressable. Providers may need clarification or retraining. Procedures separating study conditions may need strengthening. Comparator delivery may need better monitoring. Participants may need clearer instructions.

Any corrective action should itself be evaluated carefully. A modification intended to restore the original comparison can introduce a new difference between participants studied before and after the correction. If the change affects approved procedures or substantive features of the protocol, determine whether ethics approval, the protocol, or preregistration needs to be revisited before implementation.

Sometimes the Comparison Problem Reveals a Faulty Design Assumption

A comparison group may be functioning exactly as implemented while failing conceptually. Perhaps researchers assumed that usual care would not contain a particular intervention component, but it routinely does. Perhaps they assumed participants in different groups would have no opportunity to interact, yet the research setting makes separation unrealistic.

In that situation, the issue is not merely poor implementation. An assumption underlying the design may have been incorrect. You then need to assess what happens when an important research-design assumption turns out to be wrong.

Report the Comparator as Delivered, Not Only as Planned

Readers need to know what the study actually compared. CONSORT 2025 calls for reporting intervention and comparator delivery as they were actually administered, including adherence and fidelity where appropriate.

This means that the methods can describe the planned conditions, but the results and relevant methodological reporting should make deviations visible. If a comparison condition varied across sites, substantial contamination occurred, or many participants did not receive the assigned condition, that information may be central to understanding the result.

When the departure is consequential, consider whether it constitutes a methodological deviation that needs explicit reporting.

A Comparison Group Begins Receiving Elements of the Intervention

The presence of contamination does not tell the researchers what the result “would have been” without contamination. That counterfactual cannot simply be reconstructed by assertion. Instead, the observed comparison must be interpreted in light of the actual exposure and the study's analytical framework.

Common Mistakes When the Comparison Condition Breaks Down

Reconstruct the Actual Comparison Before Interpreting the Findings

When the control or comparison group stops behaving as planned, separate the implementation problem from its inferential consequence. First establish what happened. Then determine what comparison remains.

Sometimes a damaged comparison can be accommodated analytically and interpreted cautiously. Sometimes it fundamentally changes what the study can establish. If restoring the comparison would require a substantial methodological change, consider whether changing the research design after data collection has started is justified rather than disguising the redesign as a minor correction.

When a Control or Comparison Group Goes Off Plan, Check These Points

Questions About Control and Comparison Group Problems

The Group Labels Matter Less Than What the Groups Actually Experienced

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.