Zelen randomisation remains unfamiliar to a large share of medical device clinical trial sponsors, even though the French National Authority for Health (HAS) has classified it, since 2021, among the recommended experimental designs alongside adaptive, cluster, and cross-over trials. The topic has just been highlighted by DeviceMed magazine (issue 5/2026, September 2026), which devoted an article to it written by the AFCROs Medical Devices working group.
First proposed in the late 1970s by biostatistician Marvin Zelen, Zelen randomisation reverses the classic principle of the randomised controlled trial: randomisation takes place before full informed consent is collected, rather than after, as in the conventional design. For medical device manufacturers facing methodological biases specific to their sector, understanding how Zelen randomisation works has become as much a strategic issue as a scientific one.
How Two-Stage Consent Zelen Randomisation Works
The most developed variant of Zelen randomisation, known as the double-consent design, unfolds in two stages. All eligible patients receive initial information and sign a first consent covering their participation in a follow-up study and data collection, without being told that an experimental arm exists or that randomisation is underway. Randomisation then takes place: the control group continues its usual care, exactly as presented from the outset, while the intervention group is informed of the experimental treatment and invited to give a second, specific consent. If they refuse, these patients receive the same follow-up as the control group but remain in the intervention group for the intention-to-treat analysis. This point is central to Zelen randomisation: refusals must be kept in the main analysis, because they reflect the real-world implementation of the device, even if this dilutes the observed effect.
Why Zelen Randomisation Is Particularly Suited to Medical Devices
A medical device trial runs into a problem that the classic randomised trial cannot solve: it is almost always impossible to create a fake device. For a drug, an active tablet can be matched against a placebo tablet that looks strictly identical, so neither patient nor clinician knows who is receiving what. For an implant, a surgical procedure, or a connected device worn daily, this “double blind” simply does not exist: in most cases, the patient knows whether they received the experimental device or not.
This lack of a placebo, in turn, creates a disappointment bias. In a classic randomised trial, the control-group patient is told from enrolment that an experimental arm exists, one they did not have access to. Knowing this, they may feel disappointed by their allocation, change their behaviour, report pain or quality of life differently, or even withdraw from the study before its end — a risk amplified by patient communities that are highly informed and active on social media, eager for innovative solutions. This bias matters even more given that many medical device trials rely precisely on subjective, patient-reported outcomes, in the absence of a well-established reference treatment.
This is exactly what Zelen randomisation neutralises: because the control group is never told an experimental arm exists, it cannot be disappointed by an allocation whose very possibility it is unaware of. The method does not remove the impossibility of blinding the device, but it cancels out the consequences on patients’ responses — which is why the HAS recommends it in particular for pragmatic medical device studies built around patient-reported outcomes.
A Method the Figures Show Is Still Marginal in France
Despite these strengths, the HAS itself acknowledges, in its methodological guide for the clinical development of medical devices (section 5.2, “Zelen or randomised consent design trial”, pp.18-21), that this design remains very little used in France. An exploration of the public Aggregate Analysis of ClinicalTrials.gov (AACT) database, carried out in May 2026, confirms this: of 132 trials worldwide identified as using Zelen randomisation or its modern variants, only 36 involve France, and just 8 of those since 2021. These figures likely understate the reality, since the method is not always specifically flagged as such on trial registries. On PubMed, the term “Zelen” appears in the title or abstract of only three French publications since 2021, none of them in the medical device field: Zelen randomisation therefore remains a very poorly documented method in the French-language literature, despite being an already-established practice at several centres abroad.
Anticipating the Methodological and Operational Requirements of Zelen Randomisation
Adopting Zelen randomisation calls for upfront preparation on two fronts. Methodologically, the refusal rate at the second consent stage must be built into the sample size calculation, or the study’s statistical power risks being reduced, and an intention-to-treat analysis paired with sensitivity analyses becomes essential. Operationally, confidentiality must be maintained throughout the trial: any sharing on social media or within patient associations could reveal the existence of the experimental arm to the control group. The waiver on pre-randomisation information must also be specifically justified and approved by the French Ethics Committee (Comité de Protection des Personnes), while traceability documenting the two consent stages is expected within the electronic data capture system. The AFCROs Medical Devices working group, which brings together clinical research organisations specialised in this field, regularly stresses investigator team training as a key success factor for this type of design.
A Concrete Opportunity for Medical Device Sponsors
Zelen randomisation is neither a methodological curiosity nor a device for sidestepping ethical requirements. It is a considered response to real biases and risks that medical device clinical investigations structurally encounter: disappointment bias, attrition risk, the impossibility of a placebo device, and subjective patient-reported primary endpoints. Recommended by the HAS in certain situations and already documented in the methodological literature dedicated to medical devices, Zelen randomisation represents, for sponsors and organisations specialising in medical device clinical investigation, a concrete opportunity to strengthen the quality of the data produced, provided the operational requirements are mastered from the protocol design stage onward.
Looking to secure the methodological choice for your next clinical investigation? Discover our clinical research expertise, or deepen your knowledge of good clinical practice for medical devices with our Module 1 training – GCP and ISO 14155 (delivered in French).
BluePharm supports medical device and biomaterial manufacturers in producing scientific and regulatory content on clinical evaluation methodologies, turning demanding trial designs such as Zelen randomisation into a competitive advantage rather than a methodological constraint.
