A sham procedure is a control-arm intervention in a clinical trial that mimics a real medical device procedure — anesthesia, incision, or implantation — without delivering the therapeutic component being tested. In a drug trial, a placebo is an inert pill: it carries essentially no physical risk of its own. In a device or surgical trial, the equivalent control almost always does carry real physical risk, because you cannot simulate a surgical or interventional procedure without exposing the participant to at least some of the same anesthesia, incision, or implantation risk as the active arm. That single difference — real physical risk with no prospect of therapeutic benefit to the person bearing it — is why sham procedure design is treated as a distinct methodological and ethical problem from ordinary drug placebo design, and why institutional review boards (IRBs) and research ethics committees (RECs) apply a heightened risk-benefit analysis to it.
What Counts as a Sham Procedure
A sham procedure reproduces the parts of an intervention a participant and, ideally, the outcome assessor can perceive — the setting, the sensations, the recovery experience — while omitting the step believed to produce the therapeutic effect. Documented examples include:
- Sham arthroscopic knee surgery: skin incisions and simulated instrument manipulation without the actual débridement or lavage.
- Sham deep brain stimulation (DBS): full electrode implantation with the device left unactivated, or a device that delivers stimulation at a sub-therapeutic or randomly withdrawn setting.
- Sham stereotactic neurosurgery: a burr hole drilled into the skull without penetrating the dura or delivering any biologic or cell-based product, as used in early intracerebral cell-transplant trials for Parkinson’s disease.
- Sham catheter-based procedures: catheter insertion and positioning without the therapeutic step (e.g., renal denervation, valve procedures) being performed.
Designs vary in how much of the real procedure they reproduce. A full sham replicates every observable step short of the therapeutic action (as in the knee arthroscopy example above). A partial or minimal sham stops earlier — for example, sedation and a scalp incision without a burr hole — trading some blinding fidelity for lower participant risk. That trade-off, fidelity of blinding versus magnitude of risk, is the central design tension covered throughout this guide.
Why Device Trials Need a Different Control Logic Than Drug Trials
The FDA’s guidance on device trial design, Design Considerations for Pivotal Clinical Investigations for Medical Devices (finalized November 7, 2013), notes plainly that placebo controls — sometimes called shams — for devices can be considerably harder to construct than a placebo pill, and that when a study is not properly masked, subconscious or unconscious influence on the patient or investigator can bias outcomes, an effect that is difficult to estimate and can be large. The guidance lays out the available control types for device studies — active control, no-treatment control, non-randomized concurrent (observational) control, historical control, patients as their own control, and sham/placebo control — and situates sham control as the option that best addresses expectation bias and the placebo response, at the cost of exposing the control group to procedural risk the drug-trial equivalent never carries.
This is compounded by the fact that many device and surgical interventions carry a documented placebo response of their own: patients who believe they received an active treatment frequently report symptomatic improvement regardless of whether the treatment did anything physiologically. Without a sham arm, a single-arm or open-label device study cannot separate a genuine treatment effect from this response — which is exactly the methodological argument for using one in the first place, and exactly what makes omitting one, or under-powering it, a design decision worth documenting.
The IRB/REC Risk-Benefit Analysis for Invasive Shams
Under the U.S. Common Rule (45 CFR 46), an IRB may only approve research where risks to subjects are minimized and are reasonable in relation to anticipated benefits — to the subject, if any, and the importance of the knowledge expected to result. A sham procedure arm sits awkwardly inside that test: by design, the participant assigned to it receives no prospect of direct therapeutic benefit, yet is exposed to real procedural risk purely so the trial as a whole can generate valid data. That means the “benefit” side of the calculation for a sham-arm participant is not personal but collective — the value of the knowledge the trial produces — which is a materially different justification than the one used for a genuinely low-risk research procedure, and boards are expected to scrutinize it accordingly rather than default to a minimal-risk determination.
In practice this means an IRB/REC reviewing an invasive sham-controlled design typically expects the study team to demonstrate, and document, several things before approval: that the scientific question cannot be answered without a masked control; that the sham has been engineered to expose participants to the least risk consistent with maintaining blinding (general anesthesia versus sedation, full incision versus a smaller one, drilling versus scalp incision alone); and that the informed consent process makes the no-benefit, real-risk nature of the sham arm unambiguous rather than allowing it to be obscured by the therapeutic framing of the trial as a whole. See CASRAI’s IRB/REC Approval Process guide for how this fits into the broader review pathway, and the Belmont Report entry for the underlying risk-benefit and beneficence principles the Common Rule’s test derives from.
A Published Ethical Framework: The Horng-Miller Criteria
The most widely cited framework for evaluating whether a sham procedure is ethically justifiable in a given trial comes from Sam Horng and Franklin G. Miller, “Ethical Framework for the Use of Sham Procedures in Clinical Trials,” published in Critical Care Medicine in 2003. It sets out six conditions, all of which the authors argue should be satisfied before a sham arm is ethically acceptable:
- There is a valuable, clinically relevant question the research is designed to answer.
- The sham (placebo) control is methodologically necessary to test the study hypothesis — not merely convenient.
- The risk of the sham control itself has been minimized as far as possible while preserving its methodological purpose.
- The risk of the sham control does not exceed a threshold of acceptable research risk.
- The risk of the sham control is justified by the value of the knowledge to be gained.
- The deception inherent in administering a sham is adequately disclosed and authorized during informed consent.
This framework has become a reference point in the neurosurgical and orthopedic literature specifically because sham-controlled trials for those specialties reintroduced the ethical debate that drug-placebo trials had, by comparison, already settled decades earlier. It is not a regulatory requirement in itself, but IRBs and REC reviewers frequently reason through some version of these same six questions when they evaluate a sham-controlled device protocol, whether or not they cite the paper by name.
Design Approaches That Reduce Sham-Arm Risk
Because the Horng-Miller framework and the Common Rule risk-benefit test both push toward minimizing sham-arm risk without destroying blinding, several design patterns have become common in device trials that use, or consider using, a sham control:
- Minimal/partial sham: reproducing only the steps necessary to maintain the participant’s and, where feasible, the assessor’s blinding — e.g., sedation and scalp incision without a full burr hole, or catheter positioning without device deployment.
- Delayed-start (crossover) design: sham-arm participants are offered the active intervention after the primary blinded endpoint has been assessed, converting the sham exposure from a permanent denial of treatment into a temporary one. This was the design used for the Freed et al. Parkinson’s fetal-tissue transplant trial discussed below.
- Randomized withdrawal / sub-therapeutic activation: for implanted devices such as DBS systems, all participants receive the implant, and the “sham” condition is achieved by leaving the device off or set below a therapeutic threshold for the blinded portion of the trial, rather than omitting implantation itself — which shifts the sham risk toward the shared implantation procedure rather than adding a second, unique risk exposure.
- Blinded outcome assessors even when subjects cannot be fully blinded: when the risk of a true full sham is judged unacceptable, some trials blind only the assessor evaluating outcomes, accepting a weaker but non-zero reduction in expectation bias.
Documented Precedents
Two trials are frequently cited in the methodological and bioethics literature on device/procedure shams:
Moseley et al., “A Controlled Trial of Arthroscopic Surgery for Osteoarthritis of the Knee” (New England Journal of Medicine, 2002) randomized 180 patients to arthroscopic débridement, arthroscopic lavage, or a placebo procedure consisting of skin incisions and a simulated débridement without insertion of the arthroscope. At two years, outcomes in the placebo group were statistically indistinguishable from both active-treatment groups — a result that materially changed clinical practice around arthroscopic surgery for knee osteoarthritis and is now a standard teaching example for why device/procedure trials need masked controls at all.
Freed et al., fetal-tissue transplantation trial for Parkinson’s disease (New England Journal of Medicine, 2001) randomized 40 patients to either fetal dopamine-neuron transplantation or a sham procedure involving general anesthesia, a stereotactic frame fitted to the skull, an MRI scan, and a burr hole drilled partway into the skull without penetrating the dura. Participants randomized to sham were offered the option of receiving the active transplant after the trial’s primary endpoint had been assessed. The trial is one of the most extensively debated cases in research ethics precisely because it combined general anesthesia and a cranial surgical procedure with zero prospect of direct benefit to the sham group, and it remains the touchstone case cited in most subsequent sham-surgery ethics literature, including the framework above.
Informed Consent for Sham-Controlled Device Trials
Because a sham arm involves both real risk and, in most designs, deliberate non-disclosure of which arm the participant is in during the blinded phase, the consent process carries extra weight. IRB/REC reviewers typically expect the consent document and discussion to make several things explicit and separable from the general trial description: that random assignment could result in a procedure with anesthesia and/or incision risk but no active therapeutic component; the specific steps the sham procedure involves, described concretely rather than folded into general trial language; whether and when unblinding or crossover to active treatment will occur; and that participants understand this is not simply “a less effective version” of the active arm but a control with no expected direct clinical benefit. This is also where clinical equipoise becomes relevant as a threshold condition — a trial that includes a real-risk sham arm needs genuine uncertainty about the intervention’s benefit within the expert medical community to justify randomizing anyone away from it in the first place.
Frequently Asked Questions
Is a sham procedure the same thing as a placebo?
They serve the same masking function — a control condition indistinguishable to the participant from the active intervention — but a drug placebo is inert and essentially risk-free, while a device/surgical sham reproduces some or all of a real invasive procedure and therefore carries real physical risk. That risk difference is why sham procedures are evaluated under a heightened risk-benefit standard rather than treated as automatically acceptable the way an inert pill placebo generally is.
Can an IRB approve a sham arm that offers no possible medical benefit?
Yes, under the Common Rule’s risk-benefit test the “benefit” justifying a sham arm’s risk does not have to be a direct clinical benefit to that participant — it can be the value of the knowledge the overall trial produces, provided the risk to the sham-arm participant has been minimized and the sham is methodologically necessary to answer the research question. This is a materially different justification than for most other research procedures, and IRBs are expected to document that distinction explicitly rather than apply a standard minimal-risk analysis.
Does the FDA require a sham control for device trials?
No. FDA’s 2013 pivotal-trial design guidance discusses sham/placebo control as one of several acceptable control types (alongside active control, no-treatment control, historical control, and patients as their own control) and describes when it is methodologically preferable, but the choice of control type is a design decision the sponsor justifies within a specific study, reviewed by FDA and the IRB together, not a blanket requirement.
What is a “minimal sham” and why would a trial use one?
A minimal or partial sham reproduces only enough of the real procedure to preserve blinding — for example, sedation and a scalp incision without drilling into the skull — rather than replicating every step of the active procedure. Trials use this approach specifically to reduce the physical risk borne by sham-arm participants while still controlling for the placebo response, trading some blinding fidelity for a meaningfully lower risk profile.
