Zelen Design

The Zelen design, also known as randomized consent design or prerandomization design, is a clinical trial methodology in which participants are randomized to treatment groups before being asked for informed consent. In the original single consent version, only participants assigned to the experimental intervention are approached for consent; those randomized to the control group receive standard care without being informed of the trial. The design was proposed by biostatistician Marvin Zelen in 1979 to make it easier for physicians and patients to take part in randomized trials, because many physicians were reluctant to ask patients to accept random assignment; it has since also been used to limit contamination and disappointment in the control groups of pragmatic trials.

The Zelen design addresses a persistent challenge in clinical trials: when participants learn they have been randomized to a control group, some seek out the experimental treatment independently or are disappointed and change their behavior. Both responses can dilute or distort the comparison between groups. By randomizing before consent, the Zelen design preserves a control group that reflects real world standard care. It does not remove refusals in the experimental group: participants who decline the experimental intervention are still analyzed in the group they were randomized to, which dilutes the estimated effect. The design is particularly relevant for trials of screening programs, behavioral interventions, and public health measures where awareness of randomization would alter participant behavior. It raises important ethical considerations that systematic reviewers must understand when appraising trial quality.

In the single consent Zelen design, all eligible participants are randomized. Those assigned to the experimental group are informed about the trial and asked for consent. Those who consent receive the experimental intervention. Those who decline receive standard care but remain in the experimental group for analysis (intention to treat). Participants in the control group receive standard care and are not contacted about the trial. Because they are not asked for consent, the design needs approval from a research ethics committee; in the United States, an IRB may waive or alter consent under the Common Rule only when a set of conditions are all met, including that the research involves no more than minimal risk and could not practicably be carried out without the waiver (45 CFR 46.116(f)).

In the double consent variant, both groups are told about the treatment to which they have been assigned and asked to consent to it; those who decline can receive the alternative treatment or usual care instead. This reduces the ethical concerns of the single consent version, but participants who decline and switch still count in their randomized group, so they dilute the intention to treat comparison, and knowing the allocation before consenting can influence who agrees to the treatment.

Analysis follows the intention to treat principle, comparing outcomes by original randomization assignment regardless of actual treatment received. This keeps the comparison protected by randomization, but it estimates the effect of offering the intervention; the more participants decline, the more the effect is diluted and the larger the sample needed.

Use the Zelen design when conventional randomization would cause substantial contamination or disappointment in the control group, when the experimental intervention is noninvasive or widely available (such as screening or education programs), and when the research question concerns effectiveness under real world conditions rather than efficacy under ideal conditions. It is most appropriate for pragmatic trials where the control condition is existing standard practice.

Applying the Zelen design when the experimental intervention carries significant risk is ethically problematic, because studying control participants without consent is generally justified only when the research adds no more than minimal risk, and participants offered the experimental treatment are approached only after chance has chosen it for them. Failing to use intention to treat analysis defeats the purpose of the design and introduces the same biases it was created to avoid. Ignoring the share of participants who will decline their allocated treatment, which dilutes the intention to treat estimate and requires a larger sample size, leads to underpowered studies.

A public health team evaluates a community based diabetes screening program. Using a single consent Zelen design, 10,000 eligible residents are randomized: 5,000 to receive screening invitations and 5,000 to continue with usual care. Only the screening group is contacted; the ethics committee approved collecting the usual care group's outcomes from health records without contacting them. Of the 5,000 invited, 3,200 consent and are screened. Analysis compares diabetes related outcomes across all 10,000 participants by original assignment, providing an unbiased estimate of the population level effect of offering screening. Because 1,800 of the 5,000 invited residents (36%) declined and were still analyzed in the screening group, this estimate is diluted relative to the effect of actually being screened.