# Randomised is useful. How was the assignment hidden?

Draft • Research checked 4 October 2026 • AI-assisted DomDNA editorial content. No independent clinical review or publication is claimed. General education, not individual medical advice.

A study says participants were randomly assigned. That is a useful starting point, but it leaves an important question unanswered: could the person enrolling the next participant discover which group would come next? The assignment sequence and the way it was protected are separate parts of the design.

For a reader comparing research about exercise classes, nutrition programmes or other health interventions, this distinction helps explain why an impressive baseline table is not the whole story. You want to understand how the groups were created before asking what happened afterwards.

## Before anybody joins a group

Randomisation uses an unpredictable assignment process. Allocation concealment protects that process while participants enter the study. [Cochrane's risk-of-bias guidance](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08) addresses both because knowledge of the next assignment can influence who gets enrolled. Alternating between groups is predictable, even if the resulting groups look balanced.

Concealment does not require that nobody ever learns the assignment. A person attending an exercise class will usually know they are attending it. What matters here is whether that knowledge was available early enough to influence entry into the comparison.

The [CONSORT 2025 statement](https://www.bmj.com/content/389/bmj-2024-081123) supports transparent reporting of the assignment methods and participant flow. Seeing its name in a paper is not a certificate that the trial was conducted well. It gives you a place to look for the relevant explanation.

## A recruitment desk example

Imagine a hypothetical study comparing two appointment-reminder formats. The enrolment assistant has a list showing that the next person will receive the more intensive reminder. A participant with a history of missing appointments arrives. The assistant might enrol them now, or decide to wait for another participant, depending on what the assistant hopes the study will show.

Nobody needs to invent a participant or alter an outcome for the comparison to become less trustworthy. An ordinary decision about enrolment can change the kinds of people entering each group.

Now imagine the assistant checks eligibility and records consent before a separate system releases the assignment. The assistant cannot use the upcoming reminder type to decide who enters that slot. This protects one part of the comparison. It does not solve every possible problem in the study.

These are invented scenarios, not accounts of real trials. Their purpose is to show why a hidden sequence matters even when the intervention itself cannot be hidden later.

## Balance is an observation, not the randomisation method

A baseline table describes the groups the study actually produced. It cannot reveal every unmeasured characteristic, and random assignment can produce some differences by chance, particularly in small samples. Two near-identical columns therefore do not prove that the assignment was protected.

Conversely, a visible age difference does not automatically prove that randomisation failed. The relevant question is whether the reported process was credible, not whether every entry happened to match.

The [official trial-design reporting item](https://www.consort-spirit.org/item9-trialdesign) also asks authors to describe the design and allocation ratio. A two-to-one ratio may be deliberate. It does not tell you whether assignment was random, concealed or appropriate for the research question.

## Reading the methods without becoming a trial auditor

Look for a short account of who generated the sequence, how assignments were released and who enrolled participants. If the paper uses an unexplained phrase such as “sealed envelopes”, there may be more information in a protocol or supplement. The phrase alone leaves practical questions about predictability and access.

You can record uncertainty without accusing the researchers of wrongdoing. “Concealment was not explained in the text I could access” is a defensible note. “The trial was manipulated” requires evidence you do not have.

A pragmatic reading habit is to write two separate sentences in your research notes: how assignment was generated, and how it stayed unavailable until enrolment was complete. If both sentences collapse into “they say randomised”, you have found a reporting gap worth remembering.

This is especially useful when a promotional summary treats randomisation as the final word on quality. It is one design strength, with implementation details behind it. Keep that strength in proportion to the outcomes, follow-up and analysis rather than letting a single label decide the entire interpretation.

For an educational starting point, the [DomDNA lifestyle quiz](https://domdna.com/quiz) records lifestyle answers, not DNA analysis or a clinical diagnosis.