Master Protocols and Platform Trials for Medical Devices: The Adoption Gap
Analysis of master protocol, platform, basket, and umbrella trial adoption in medical devices — with ClinicalTrials.gov data showing near-zero device uptake and when the design pays off.
The Most Important Trial-Design Innovation Has Essentially Bypassed Devices
Master protocols — basket, umbrella, and platform trial designs — let one overarching protocol answer multiple questions with shared infrastructure, shared control arms, and adaptive randomization. They have reshaped drug development, most visibly in oncology, where the NCI-MATCH trial ran 60+ treatment arms across tumor types and the RECOVERY trial generated definitive COVID-19 treatment evidence for dexamethasone in weeks rather than years.
But medical devices have almost none. An analysis of ClinicalTrials.gov data (July 2026 snapshot) finds that only 9 of 63,059 device interventional studies (0.014%) name a master, platform, basket, or umbrella design in the title — versus 211 of 454,529 across all interventional studies (0.046%). The device adoption rate is roughly threefold lower, and most of those nine device hits are borderline cell/gene therapy or research-methodology studies rather than regulated product platform trials.
The US FDA's only master-protocol guidance — finalized in March 2022 (Doc 2022-04398, Docket FDA-2018-D-2847) — is explicitly scoped to "Oncology Drugs and Biologics." Even the broader revised draft ("Master Protocols for Drug and Biological Product Development") remains drug- and biologic-only. CDRH has issued no device master-protocol guidance.
The obstacles are not statistical. They are structural: heterogeneous devices, operator-dependent outcomes, difficulty blinding, small patient populations, no regulatory framework, and a business model that typically produces one device per indication rather than many candidates. This guide explains the taxonomy, quantifies the adoption gap, identifies when a master-protocol design does pay off for a device, and addresses the IDE regulatory mechanics that any device master protocol must navigate.
What Is a Master Protocol, and How Do Basket, Umbrella, and Platform Trials Differ?
The canonical taxonomy comes from Woodcock and LaVange (NEJM 2017;377(1):62–70), who defined a master protocol as a single overarching protocol designed to evaluate multiple hypotheses — testing multiple therapies, multiple diseases, or both — within a unified infrastructure.
Three subtypes have emerged:
| Design | Structure | Key Advantage | Classic Example |
|---|---|---|---|
| Basket | One therapy tested across multiple diseases or conditions (typically grouped by a shared biomarker or target) | Tests a single intervention across indications simultaneously, accelerating signal detection | NCI-MATCH (60+ treatment arms across tumor types sharing specific molecular alterations) |
| Umbrella | Multiple therapies tested within one disease (patients stratified by subtype, biomarker, or risk profile) | Matches patients to the most relevant experimental arm within a single disease framework | Lung-MAP (squamous NSCLC with molecular stratification) |
| Platform | A perpetual trial framework where treatment arms enter and exit over time, sharing a common control arm and infrastructure | Shared concurrent control arm reduces sample size; perpetual enrollment eliminates startup delays for new arms; adaptive randomization optimizes allocation | RECOVERY (COVID-19), HEALEY ALS Platform Trial, I-SPY 2 (breast cancer) |
All three are "master" protocols, but the platform trial is the most radical: it is designed to persist indefinitely, with new experimental arms added as they become ready and existing arms dropped when evidence of superiority, futility, or harm accumulates. The control arm is shared across all experimental arms, which is the primary source of efficiency — each patient randomized to control serves as a comparator for every active arm running concurrently.
How a Master Protocol Differs From an Adaptive Trial
The terms are sometimes confused. An adaptive trial modifies parameters (sample size, randomization ratio, dosing) within a single experimental arm based on interim data. A master protocol manages multiple experimental arms or disease groups under a single overarching protocol. A platform trial is both adaptive (arms can be added or dropped based on interim analyses) and master (multiple arms share one protocol and control infrastructure). All platform trials are master protocols, but not all adaptive trials are master protocols.
Why Are Master Protocols Everywhere in Oncology Drugs but Almost Absent in Devices?
The Quantitative Gap
The evidence for near-zero device adoption comes from two sources:
Park et al. (Trials 2019;20:572) conducted the first systematic landscape review of master protocols, identifying 83 studies: 49 basket trials, 18 umbrella trials, and 16 platform trials. Of these 83, 76 were in oncology and none evaluated a medical device as the primary intervention. The dominance of oncology was overwhelming.
ClinicalTrials.gov analysis (July 2026 snapshot) confirmed the gap at scale: among all 454,529 interventional studies in the database, 211 (0.046%) named a master, platform, basket, or umbrella design in the study title. Among the 63,059 device interventional studies (those with an intervention type of DEVICE), only 9 (0.014%) did so. The device rate is roughly three times lower than the overall rate. Inspecting the nine device hits reveals that most are cell/gene therapy studies where the device is a delivery system, or research-methodology studies examining master-protocol statistics, rather than regulated device product evaluations.
Six Structural Reasons Devices Are Different
1. Device Heterogeneity
Drug master protocols work because the same molecule can be tested across diseases (basket) or multiple molecules target the same disease pathway (umbrella/platform). Devices are inherently heterogeneous — they differ in design, materials, dimensions, surgical technique, and mechanism of action. Two knee implants from different manufacturers are not the same "intervention" in the way that two small-molecule kinase inhibitors are. This heterogeneity makes shared control arms and cross-arm statistical comparisons difficult.
2. Operator Dependence and Learning Curve
Device outcomes depend on surgeon skill, technique, and experience. The learning curve for a new implant or surgical instrument can take 30–50 cases before performance stabilizes. In a platform trial with multiple device arms, operator variability is an additional source of bias that does not exist in drug platform trials (where administration is standardized).
3. Blinding Difficulty
Many devices cannot be blinded. Implants, surgical instruments, external fixation devices, and diagnostic systems are physically distinct — patients and investigators can often tell which device is in use. Drug platform trials rely on blinding (placebo matching, double-dummy designs) to reduce bias in the shared control arm. Without blinding, a shared control arm in a device platform trial introduces assessment bias.
4. Smaller Patient Populations
Device indications are often niche. A specific implant for a specific orthopedic defect may have a few thousand eligible patients per year across the US. Platform trials achieve their statistical efficiency by enrolling large numbers of patients into a shared control arm — an advantage that disappears when the eligible population is too small to sustain multiple concurrent experimental arms.
5. No Regulatory Framework
FDA's Center for Drug Evaluation and Research (CDER) has a master-protocol guidance. FDA's Center for Devices and Radiological Health (CDRH) has none. Device clinical investigations run under Investigational Device Exemptions (IDE, 21 CFR Part 812), and the IDE regulations do not contemplate a multi-device master protocol. The absence of a regulatory pathway increases uncertainty and discourages sponsors from proposing the design.
6. Business Model
Pharmaceutical companies routinely develop multiple drug candidates for the same oncology indication, making a basket or platform design a natural tool for portfolio optimization. Medical device companies typically develop one device for a given indication, with iterative design improvements rather than parallel candidate development. There is rarely a portfolio of competing device candidates that would benefit from a shared platform.
When Does a Platform or Master Design Actually Pay Off for a Device?
Despite the structural obstacles, there are device niches where a master-protocol design can work. The decision framework centers on five conditions:
Condition 1: Multiple Near-Identical Candidates
The design pays off when multiple devices (or device configurations) target the same indication with similar mechanisms. In radiation oncology, for example, multiple MR-Linac configurations and treatment protocols target the same tumor types — a situation analogous to multiple drug candidates targeting the same pathway.
Condition 2: A Stable Standard-of-Care Control
A shared control arm requires a well-defined, stable standard-of-care comparator that is unlikely to change during the trial. If the control arm shifts (new device approved, practice change), the cross-arm comparisons are compromised. Drug oncology trials have historically used well-defined chemotherapy regimens as controls; device fields with a stable comparator (conventional external-beam radiation, standard wound care) are better candidates.
Condition 3: Large Enough Patient Population
The statistical efficiency of shared control arms only materializes with sufficient enrollment. Rule of thumb: if the eligible population cannot support at least 2–3 concurrent experimental arms plus a shared control arm with adequate statistical power, the platform overhead exceeds the benefit.
Condition 4: Objective or Standardizable Outcomes
Outcomes that can be measured objectively — tumor response on imaging, wound closure, physiological parameters — are less susceptible to operator and assessment bias than subjective outcomes like pain scores or surgeon-assessed function. Platform trials work best with objective endpoints.
Condition 5: Multi-Site Infrastructure
Platform trials require coordinating multiple sites, central randomization, adaptive analytics, and data-sharing infrastructure. Academic consortia (ECOG-ACRIN, NRG Oncology) and registries (Society of Thoracic Surgeons, TVT Registry) can provide this infrastructure in specific device fields.
Real Device Examples
MR-Linac radiation oncology trials: The SMART trial (n≈1,000 planned), CONFIRM, and JUMP all evaluate MR-guided adaptive radiotherapy using MR-Linac systems. While not formally structured as a single platform trial, they share infrastructure, patient populations, and endpoint definitions — and Gilbert et al. (JNCI 2024;116(8):1220–1229) identified 12 radiotherapy master protocols (4 platform, 3 umbrella, 5 basket). Radiation oncology is the device niche closest to platform-trial adoption because the "device" (the treatment protocol and planning system) is standardizable, the control arm (conventional radiation or best supportive care) is stable, and academic consortia provide the multi-site infrastructure.
Diabetic foot ulcer modified-platform trial (NCT07086443): This trial tests multiple cellular and/or tissue-based products (CAMPs) with a common standard-of-care control arm for diabetic foot ulcers and chronic wounds. The "modified platform" design allows new wound-care products to enter the trial as arms without restarting enrollment. This is one of the few non-oncology device-adjacent platform trials.
HEALEY ALS Platform Trial: While primarily evaluating drug interventions, the HEALEY platform demonstrated the efficiency model: approximately 33% reduction in per-regimen sample size and approximately 66% fewer participants randomized to placebo compared to running each drug arm as a standalone trial. These efficiency gains translate directly to any device platform trial with similar structure.
How Is a Device Master Protocol Regulated Under the IDE?
The IDE Framework (21 CFR Part 812)
Device clinical investigations of significant-risk devices in the US require an Investigational Device Exemption (IDE) approved by both the FDA and an Institutional Review Board (IRB). The IDE regulations (21 CFR Part 812) govern study design, informed consent, monitoring, adverse-event reporting, and device accountability.
The Multi-Device IDE Question
The central regulatory question for a device master protocol is: can multiple investigational devices be tested under one IDE? FDA has issued no guidance on this. The IDE regulations were written for a single investigational device evaluated in a single study protocol. A master protocol testing multiple devices raises several unanswered questions:
- IDE scope: Does each device arm need its own IDE, or can one IDE cover the master protocol with device-specific supplements or amendments?
- Significant-risk determination: Each device arm may have a different risk profile, potentially requiring separate significant-risk vs. nonsignificant-risk determinations.
- IRB review: The IRB must review the overarching protocol and each device arm. Adding a new arm to a platform trial requires an amendment — but is this a minor amendment (faster review) or a new protocol (full review)?
- Adverse-event reporting: Adverse events must be attributed to the correct device arm. The shared control arm complicates attribution when a safety signal emerges.
- Device accountability: 21 CFR 812.140 requires detailed records of device disposition. A master protocol with multiple devices requires parallel accountability systems.
Contrast With Drug INDs
FDA's master-protocol guidance for oncology drugs explicitly addresses how multiple experimental arms can operate under a single IND, with cross-referencing, shared control data, and coordinated safety reporting. CDRH has no equivalent document. Until CDRH issues device-specific master-protocol guidance, sponsors must negotiate the regulatory structure on a case-by-case basis through pre-submission meetings (Pre-Subs).
Practical Recommendation
Any device company considering a master-protocol design should:
- Request a Pre-Sub meeting with CDRH specifically to discuss the master-protocol structure, IDE scope, and multi-device arm management
- Propose a specific IDE structure (single IDE with device-arm supplements, or multiple IDEs with a shared protocol and statistical analysis plan)
- Define the adaptive rules (how arms are added, how futility and superiority are determined, how the shared control arm is managed)
- Establish a Data Safety Monitoring Board (DSMB) or Independent Data Monitoring Committee (IDMC) with device-specific expertise
What Efficiency Gains Have Real Platform Trials Achieved?
| Trial | Disease | Type | Key Efficiency Metric |
|---|---|---|---|
| HEALEY ALS | Amyotrophic lateral sclerosis | Platform | ~33% per-regimen sample-size reduction; ~66% fewer placebo participants |
| PLATO | Rectal cancer (radiotherapy) | Umbrella | ~20% cost savings (~GBP 450,000) vs. three separate trials |
| RECOVERY | COVID-19 | Platform | 47,000+ participants across 20+ treatments; definitive dexamethasone evidence in weeks |
| I-SPY 2 | Breast cancer | Platform | 10+ experimental arms "graduated" to Phase III with ~60% fewer patients than traditional Phase II |
| NCI-MATCH | Multiple cancers (biomarker-selected) | Basket | 60+ treatment arms across tumor types under one protocol |
The efficiency gains are real and substantial. The shared control arm is the primary driver: in a platform trial with five experimental arms and one shared control, each control-arm patient serves as a comparator for all five experimental arms simultaneously. This can reduce total enrollment by 30–50% compared to running five separate two-arm trials, with proportional reductions in cost, time, and patient burden.
Cost and Time Implications for Devices
For a hypothetical device platform trial with three device arms and one shared control arm:
- Without platform design: Three separate two-arm RCTs, each needing (for example) 200 patients (100 device + 100 control) = 600 total patients, three separate IRB approvals, three IDE submissions, three site startup processes.
- With platform design: One three-arm-plus-control trial needing approximately 400–450 total patients (100 per device arm + 100–150 shared control, with some adaptive optimization), one IRB approval (with amendments for each new arm), one IDE (with supplements), one site startup process.
The platform design could save approximately 25–35% of patient enrollment and significantly reduce site startup duplication, monitoring overhead, and regulatory interaction costs.
The Path Forward: How Devices Could Adopt Master Protocols
Pragmatic Registry-Embedded Designs
The most realistic near-term path for device master protocols may be registry-embedded prospective studies. Large device registries — the TVT Registry (transcatheter valve therapies), the STS National Database (cardiac surgery), the AJRR (American Joint Replacement Registry) — already capture real-world device outcomes at scale. Embedding prospective, randomized device arms within these registry infrastructures could create a de facto platform-trial capability without the full burden of a standalone master protocol.
The NESTcc (National Evaluation System for health Technology Coordinating Center) is developing infrastructure for coordinated device evidence generation across multiple data sources. While NESTcc focuses primarily on real-world evidence, its coordinating capabilities could support prospective multi-arm studies.
FDA CDRH Real-World Evidence Pathway
FDA CDRH's final guidance on real-world evidence (August 2017) and its ongoing Total Product Life Cycle (TPLC) initiative provide a framework for using registry data and electronic health records in device regulatory decisions. A device master protocol that leverages this infrastructure — using registry-embedded randomization and real-world data capture — could be more feasible and more attractive to CDRH than a traditional standalone platform trial.
What a Device Master-Protocol Guidance Would Need to Address
If CDRH were to issue device-specific master-protocol guidance, it would need to address at minimum:
- IDE structure for multi-device master protocols
- Significant-risk determination for individual device arms within a master protocol
- Statistical considerations for shared control arms with heterogeneous device interventions
- Adaptive rules for adding and dropping device arms
- Device accountability across multiple arms
- Adverse-event attribution in shared-control designs
- Notified-body and IRB interaction models for platform trials with evolving arms
Frequently Asked Questions
Is There an FDA Master-Protocol Guidance for Medical Devices?
No. FDA's only master-protocol guidance (March 2022, Doc 2022-04398) is explicitly titled "Master Protocols: Efficient Clinical Trial Design Strategies to Expedite Development of Oncology Drugs and Biologics." Even the broader revised draft ("Master Protocols for Drug and Biological Product Development") is limited to drugs and biologics. CDRH has issued no device-specific master-protocol guidance.
Can Multiple Investigational Devices Be Run Under One IDE in a Master Protocol?
There is no clear regulatory guidance on this. The IDE regulations (21 CFR Part 812) were designed for single-device investigations. A sponsor proposing a multi-device master protocol should request a Pre-Sub meeting with CDRH to negotiate the IDE structure. Options include a single IDE with device-arm-specific supplements, or multiple coordinated IDEs sharing a common protocol and statistical analysis plan.
Do Device Registries Like the TVT Registry Count as Platform Trials?
No. Registries are observational data sources that capture real-world device outcomes — they are not prospective experimental platforms with randomization and shared control arms. However, registries can serve as the infrastructure for registry-embedded prospective trials, which could incorporate platform-trial elements. The distinction is between a registry (observational, no randomization) and a registry-embedded trial (prospective, randomized, using registry infrastructure for enrollment, randomization, and outcome capture).
What Is the Difference Between a Platform Trial and an Adaptive Trial?
An adaptive trial modifies parameters (sample size, randomization ratio, dose level) within a single experimental arm based on interim data. A platform trial manages multiple experimental arms under a single overarching protocol, with the ability to add new arms and drop existing arms over time. A platform trial is inherently adaptive (it adapts by adding/dropping arms), but an adaptive trial is not necessarily a platform trial (it may have only one experimental arm).
Are There Any Device-Specific Master Protocol Examples?
Very few, and most are in radiation oncology — the device niche closest to platform-trial adoption. The SMART, CONFIRM, and JUMP trials evaluate MR-guided adaptive radiotherapy; Gilbert et al. (JNCI 2024) identified 12 radiotherapy master protocols. Outside oncology, the modified-platform diabetic-foot-ulcer trial (NCT07086443) tests multiple wound-care products with a shared control arm. These remain rare exceptions, not an established practice.
Related Guides
For deeper context on device clinical evidence and trial regulation:
- Medical device clinical trials on ClinicalTrials.gov — Quantifies the device-trial landscape; the present article adds trial-design adoption analysis
- Real-world evidence for medical devices — Covers retrospective real-world data sources; distinct from the prospective platform designs discussed here
- FDA early feasibility studies for novel devices — A different device-trial efficiency program; complementary to master protocols
- Investigational device rules compared — IDE regulations that any device master protocol must navigate
- GCP and ISO 14155 for device trials — Master protocols run under GCP and ISO 14155
- Sample size calculation for device investigations — Shared control arms change the sample-size math; links to methods
- Choosing a medical device CRO — Platform trials need specialized CRO and biostatistics partners