FDA’s surgical-robotics draft and the Galleri advisory review: what the evidence supports
FDA’s draft guidance on robotically-assisted surgical devices proposes an evidence framework for locally teleoperated robots, and an advisory panel gave GRAIL’s Galleri test favorable but qualified votes. Neither is an implementation deadline or an approval.
Principal Consultant
In brief
The robotics draft ties system architecture, latency, lifetime reliability, training and procedure-specific clinical evidence together; comments are requested by November 24, 2026. The Galleri panel voted 10–0, 6–4 and 7–2 with one abstention, and the candidate’s sensitivity figures need their version and endpoint context.
Status: November 12, 2026
Two FDA developments from late September 2026 look similar on the surface but differ in legal effect. A draft guidance proposes an evidence framework for locally teleoperated surgical robots. An advisory panel gave a favorable but qualified recommendation on GRAIL’s Galleri multi-cancer detection test. The robotics document is a draft, expressly not for implementation. The Galleri votes concern one premarket approval application (PMA) reviewed by the panel on September 23; they neither authorize the product nor set a generally binding evidence standard for multi-cancer tests. [1] [9] [10]
For development planning, the substance still matters. The robotics draft connects system architecture, performance, lifetime reliability, training and procedure-specific clinical evidence. The Galleri review shows why high specificity cannot be read apart from detection performance, the proposed claim, version changes and downstream consequences. In both cases the evidence has to be built around the specific claim, and a technically complete submission is not the same as an established clinical benefit.
Dates to watch
- November 24, 2026: comments on the robotically-assisted surgical devices draft (docket FDA-2026-N-9505) [2] [3]
- December 2–3, 2026: FDA public workshop on benefit–risk evaluation of robotic medical devices with autonomous or remote teleoperation capabilities; in-person registration closes November 23, or earlier once maximum capacity is reached [8]
Key takeaways
- Robotics: a proposal, not a mandate. The draft is nonbinding, marked “not for implementation,” and does not suspend existing statutory duties such as section 524B for cyber devices. [1] [5]
- “Local” teleoperation is defined by the room. The operator console and the patient-side subassembly must be in the same operating room. Remote teleoperation and robots that perform significant aspects of a procedure independently of the practitioner are outside the draft; some automated functions remain within scope. [1]
- Evidence leverage is limited. The proposed umbrella and covered-procedure approach can reduce procedure-specific in vivo data, but FDA states that there is no fixed list of suitable procedures. [1] [7]
- Galleri: favorable votes are not approval. The panel voted 10–0 on safety, 6–4 on effectiveness and 7–2 with one abstention on benefit–risk. The final decision rests with FDA. [9]
- Numbers need their context. FDA’s candidate analysis reports 12-month episode sensitivity of 35.0% and 31.6% with specificity above 99.7%. Version, endpoint and weighting belong next to each figure. [10]
- A randomized trial missed its primary endpoint. NHS-Galleri did not demonstrate a statistically significant reduction in stage III/IV cancer incidence across 12 prespecified cancer types after three screening rounds. This is a different question from the performance estimates in the PMA. [11] [13]
Surgical robotics: a proposed framework, not an implementation deadline
CDRH issued Robotically-Assisted Surgical Devices: Premarket Submissions on September 25, 2026. The Federal Register notice (FR Doc. 2026-19704) was filed for public inspection on September 24 and published on September 25. [1] [2]
Comments are requested within 60 days of publication, which yields November 24, 2026, under docket FDA-2026-N-9505. That date serves consideration before FDA begins work on the final guidance. It is a consultation milestone, not a compliance deadline, and comments on guidance can be submitted at any time. [1] [3] [4]
The draft creates no enforceable duties of its own, and its recommendations must be kept apart from the requirements it cites. If a surgical robot meets the definition of a cyber device under section 524B(c) of the Federal Food, Drug, and Cosmetic Act, its premarket submission must meet the cybersecurity requirements of section 524B(b). Clinical investigations remain subject to the applicable requirements of 21 CFR Parts 812, 56 and 50, taking account of their scope and any applicable exemptions. “Draft: not for implementation” does not suspend those requirements. [1] [5] [14]
The same operating room is the decisive boundary
The draft defines a robotically-assisted surgical device (RASD) as a teleoperated, software-controlled system that assists a qualified practitioner in positioning and controlling surgical instruments for open, minimally invasive or endoluminal procedures. It uses a leader/follower architecture under the practitioner’s direct control or supervision, and it may contain some automated functions that enhance user capabilities. RASDs with different physical architectures are within scope if they meet the definition. [1]
Local teleoperation means that the operator console is co-located with the bedside subassembly in the same operating room, connected by wired or wireless signal. A console that is not in that room makes the system remotely teleoperated, whatever the distance. [1]
The draft does not address remotely teleoperated RASDs or autonomous robots that perform significant aspects of a procedure independently of the practitioner. Pre-operative planning and stereotaxic systems are not RASDs, and systems assisting the placement or use of interventional devices such as interventional catheters are also outside the scope. “Endoluminal” use does not extend to positioning and controlling multiple instruments or interventional devices through blood vessels. Exclusion from this draft is not an exemption from regulation, and some recommendations may still be relevant to excluded products. For RASDs outside the scope, FDA strongly recommends Q-Submission engagement; developers who are unsure whether their device is a RASD are told to contact FDA through the Q-Submission Program. The final Q-Submission guidance of May 29, 2025, provides that voluntary route; it is neither a marketing authorization nor a pre-review of the full application. [1] [6]
Testing the robot as an integrated system
The proposed device description starts with the whole system and then covers subassemblies and components. FDA asks for kinematic and connection diagrams, including connections to external services and servers, and lists the sterile–nonsterile interface, safety features, emergency patient access and instruments. Features that enhance human capability, such as tremor reduction or motion scaling, should be described in their own sections. [1]
For a 510(k), FDA recommends a side-by-side comparison with the predicate and a rationale for each technological difference. The two example tables in the draft are organizational aids rather than mandatory templates; their practical value is traceability from a changed feature to the hazard, the testing and the place in the submission. [1]
Prior standalone authorization of an accessory or third-party device does not establish its compatibility with the robot. FDA recommends listing each accessory in the labeling and demonstrating compatibility and interoperability through performance testing of the whole RASD, including any adapters and drapes. The draft refers to FDA’s interoperability guidance for further considerations. [1]
Latency, kinematics and reliability need justified acceptance criteria
Non-clinical testing is recommended at subassembly and system level before in vivo evaluation. Test reports should state the objective, methods, sample size, predefined pass/fail criteria with clinical justification and the rationale for the instruments tested; the draft builds on FDA’s general guidance on bench performance testing reports. Test conditions should simulate worst-case conditions of clinical use. [1]
Total system latency covers the delay from the surgeon’s hand input to instrument motion and from instrument motion to the displayed visual feedback. Testing should show that it stays within predefined, clinically justified limits under representative and worst-case conditions. The draft sets no universal millisecond ceiling. It recommends testing a minimum of three RASDs, or a rationale for fewer, for latency, for accuracy, repeatability and resolution, and for workspace verification. It does not make this sample recommendation for every test. [1]
Further testing addresses motion scaling, tremor filtering, stability under disturbances, fault detection and recovery, emergency removal, safety stops, velocity limits, position tracking, collision detection and singularity avoidance. Reachability should be validated across the intended patient population, including clinically relevant extremes of body mass index and anatomy. Acceptance criteria therefore have to be tied to the intended tasks and hazards. [1]
Reliability is considered over two periods. Service life is the period specified by the manufacturer during which the RASD is expected to perform as intended. Use life is the predefined duration of active use or number of repeated uses of instruments, accessories and compatible devices. The draft calls for worst-case loading and actuation, retained control of the instruments, maintained integrity of essential surgical information and, for reusable instruments, reprocessing between uses over the labeled use life. For grip and cutting reliability it recommends a sample size sufficient to demonstrate 95% reliability with 95% confidence and allows a different level with a rationale. That recommendation should not be generalized to every reliability test. [1]
Software, networks and the operating-room team are part of performance
FDA generally considers RASD software to warrant the Enhanced documentation level under its premarket software guidance; a Basic level would need a scientific rationale based on risk. The draft emphasizes unit, integration and system-level testing, justification of regression versus full testing for changes, traceability of unresolved anomalies, testing with the final commercial software version, and documented version control and update deployment. [1]
The cybersecurity assessment, including penetration testing, should consider all relevant network elements, accessories and compatible devices that connect to or interact with the RASD during pre-deployment, normal operation and maintenance, including back-end infrastructure such as servers, databases and cloud services and the public and private networks used for communication, remote access, teleoperation and updates. The draft gives simulation of communication with a compromised hospital network as an example. These recommendations come on top of the statutory section 524B requirements for devices that meet the cyber-device definition. [1]
FDA proposes a validated training program that helps users achieve device competency before clinical use. It should reach beyond surgeons to bedside assistants and scrub nurses, and training needs should be re-evaluated when interfaces, instruments or user groups change. [1]
Human-factors work should start early, run iteratively and end in validation testing of the final interface with representative users in a realistic environment, which may include teams working together as in a normal operating room. Where several user groups work with the system at the same time, the draft asks that this be considered when selecting validation participants. Named considerations include negative knowledge transfer from other RASDs, instrument exchange and port switching, testing of all human-capability-enhancing or augmenting features even when they are optional, and emergency procedures such as conversion to open or laparoscopic surgery, emergency withdrawal, power loss and emergency stops. The draft also covers electrical safety and electromagnetic compatibility, optical radiation, sterility, reprocessing, shelf life and packaging, pyrogenicity and biocompatibility according to the product’s characteristics. [1]
Clinical evidence and the limits of umbrella procedures
FDA states that clinical data are often needed for new RASDs, meaning systems that have not previously received marketing authorization in the United States. For modifications to an authorized system, clinical data may be requested in situations such as a new intended use, special populations, different technology, unresolved non-clinical concerns or specific performance claims such as superiority or treatment of a disease or condition. Clinical data can come from prospective clinical studies, real-world evidence or clinical literature. [1]
FDA generally considers the investigational RASDs addressed by the draft to be significant-risk devices that are not eligible for the abbreviated requirements of 21 CFR 812.2(b). This is FDA’s position applied under existing law, not a new classification. The draft generally recommends evaluating a common set of surgical endpoints: length of stay, intraoperative adverse events, transfusion rate (with estimated blood loss also identified), conversion rate, postoperative adverse events, readmission and reoperation (each through 30 days), mortality and operative time. The clinical section also notes that intravascular procedures may warrant different or additional endpoints; this reference does not bring those applications within the draft’s scope, which excludes the specified interventional-device and intravascular uses. Endpoints should be analyzed alongside comparator information where available, and robotic comparator data are preferred over laparoscopic or open data. For clinical data from outside the United States, FDA stresses that patients and users should be representative of the intended U.S. population and practice, and notes that the requirements of 21 CFR 812.28 may apply. [1] [14]
The umbrella and covered procedure paradigm is a proposed way to leverage in vivo data. A covered procedure is one of lesser complexity and risk in the same surgical specialty as the umbrella procedure, and additional in vivo performance data for it may not be needed. According to the draft, umbrella and covered procedures generally contain identical or very similar procedural steps. For developers seeking authorization for multiple procedure types, FDA recommends considering this approach and, for a 510(k), recommends listing all procedures in the indications for use or in referenced labeling. The benefit–risk profile of the RASD for the umbrella procedure should be similar to or greater than for the covered procedures. FDA cautions that there is no set list of suitable umbrella and covered procedures and that relying on the examples in Appendix A will not always be appropriate; early discussion with FDA is advisable. [1] [7]
The paradigm permits extrapolation of in vivo performance data to support authorization of specified surgical procedures; it does not confer automatic authorization or establish disease-treatment benefit. It may support the addition of specified surgical procedures, but not other changes to the indications for use, such as claims for the treatment of a specific disease or condition. In our reading, suitability depends on representative procedural steps and the RASD’s benefit–risk profile, not solely on the surgical specialty; the draft itself says there is no set list of suitable procedures. For a specific procedure, FDA recommends evidence of a successfully completed procedure in its entirety; for a specialty-wide indication it recommends covering the full range of extirpative or reconstructive elements in the anatomical location. [1]
For remote and autonomous capabilities outside the draft’s scope, FDA will host a public workshop on December 2–3, 2026, in person or virtually, on evaluating benefit–risk in robotic medical devices with autonomous or remote teleoperation capabilities. On November 6, 2026, FDA published a discussion paper and request for feedback for the workshop, which is not guidance; comments are invited through January 5, 2027 [15]. The workshop page now carries a schedule, in-person registration closes on November 23, 2026, or earlier once maximum capacity is reached, and virtual registration stays open. The workshop is a consultation, not an extension of the surgical draft’s scope. [8]
Galleri: favorable advisory votes, not an FDA approval
On September 23, 2026, FDA’s Molecular and Clinical Genetics Panel discussed the PMA for GRAIL’s Galleri test, submitted on January 29, 2026. The panel’s 24-hour summary describes it as a first-of-a-kind device. [9] The proposed indication is a prescription, qualitative, next-generation-sequencing-based in vitro diagnostic that detects cancer-specific methylation patterns in cell-free DNA, intended for screening for the early detection of multiple types of cancer in adults aged 50 or older and for predicting the cancer signal origin (CSO) in positive cases. A positive result, including the predicted origin, calls for diagnostic work-up; it does not diagnose cancer. A negative result does not rule out cancer. [10]
The panel’s 24-hour summary records three separate votes: 10–0 on reasonable assurance of safety, 6–4 on reasonable assurance of effectiveness and 7–2 with one abstention on whether benefits outweigh risks. These were advisory votes on the application before the panel on September 23, not an agency decision. [9]
The qualifications behind the votes matter. The panel’s unanimous safety assessment was also qualified: it depended on adequate directions for use, labeling, and physician and patient education. On effectiveness, the panel generally expressed doubt that the term “early” in the intended use was supported. Some “yes” votes on effectiveness were conditioned on removing the “early” claim and on additional post-approval studies, and some panelists who voted “no” said they could support “yes” if the indication were modified to remove “early.” The “yes” votes on benefit–risk were conditioned on post-approval studies, appropriate labeling and provider and patient education, while panelists who voted “no” cited uncertainty about whether the cancers detected by Galleri would lead to clinically meaningful improvements in patient outcomes, and the one panelist who abstained indicated uncertainty about the benefit–risk balance in the absence of long-term clinical outcomes data. The panel also judged the clinical significance greatest for cancers without guideline-recommended screening and the value limited for cancers with established screening. These are panelists’ expressed conditions, not FDA approval conditions. [9]
FDA’s executive summary states that no FDA-authorized device is indicated for screening for multiple types of cancer and that Galleri has not previously received marketing authorization in the United States or any foreign country. An earlier version of the test has been available by prescription in the United States since June 2021. That earlier availability must not be confused with approval of the PMA candidate. GRAIL’s third-quarter results release of November 4, 2026, still states that the Galleri test has not been cleared or approved by FDA, and the company anticipates an FDA decision by early Q1 2027; that is the company’s expectation, not an FDA commitment. We found no later report of an approval. [16] [10]
FDA reviewed the sponsor’s analytical studies and had no questions for the panel on analytical performance, which does not announce closure of all review issues. FDA also had no outstanding panel questions on the concordance assessment between the study version and the candidate. The formal discussion questions concerned the clinical significance of performance overall and by cancer type, stage-specific performance and the “early” claim, risk mitigations, communication and post-approval studies. [10]
The candidate’s percentages need version and endpoint context
The studies used an earlier version, MCED-V2, to test participants prospectively. The Galleri PMA candidate is a different device: differences include the assay workflow, the classifier and the bioinformatics pipeline. Effectiveness estimates come from retrospective testing of frozen plasma samples from PATHFINDER 2 and NHS-Galleri with the candidate. Samples included all eligible participants with a positive MCED-V2 result, all eligible participants diagnosed with cancer and a stratified random subset of participants with a negative result and no cancer diagnosis, with analysis weights for the sampling. Safety data come from MCED-V2 and are treated by FDA as an approximate estimate for the candidate. [10]
In PATHFINDER 2, the candidate’s 12-month episode sensitivity was 35.0% (106 of 303 participants with a qualifying cancer diagnosis; 95% confidence interval 29.8–40.5%), specificity 99.85% (99.75–99.91%) and positive predictive value 77.0%. The analysis comprised 7,766 valid candidate tests representing a weighted total of 21,418. [10]
In the first screening round of NHS-Galleri, sensitivity was 31.6% (259 of 820 participants with a qualifying cancer diagnosis; 28.5–34.8%), specificity 99.74% (99.61–99.82%) and positive predictive value 66.2%, based on 7,926 valid tests representing a weighted total of 50,739. FDA notes that weighted counts are rounded. [10]
Twelve-month episode sensitivity is the proportion of participants with a qualifying cancer diagnosis within 12 months after the blood draw who had a positive candidate-test result. FDA points out that the true cancer status, including stage, at the time of the blood draw is unknown for cancers the test did not detect, so it is uncertain how well the figure represents true sensitivity. Stage-specific estimates can be biased because a missed cancer is staged when it is later diagnosed. FDA asked the panel to consider these uncertainties. [10]
Performance was clearly lower in early stages. For newly diagnosed primary cancers, stage I sensitivity was 21.4% (27 of 126) in PATHFINDER 2 and 13.6% (43 of 317) in NHS-Galleri; stage I–II sensitivity was 24.9% and 20.5%, and stage IV sensitivity 63.2% and 60.0%, respectively. The stage analyses cover newly diagnosed primary cancers, whereas the PATHFINDER 2 overall analysis also includes recurrent cancers. The stage limitation applies to all of these figures, but they help explain the panel’s doubts about the breadth of an early-detection claim. [10] [9]
Performance also varied by cancer type. Breast cancer sensitivity was 26.4% (14 of 53) in PATHFINDER 2 and 19.1% (21 of 110) in NHS-Galleri, and prostate cancer sensitivity was 10.7% in both (8 of 75 and 23 of 215). Combined sensitivity for the cancer types with guideline-recommended screening was 22.9% and 27.5%. Liver and intrahepatic bile-duct cancers had high estimates (81.2% and 80.0%) but rested on only 16 and 15 cases, and in the cancer-type summary tables FDA omits percentage point estimates and confidence intervals for groups with fewer than five cases, although case counts are shown; the stage-by-cancer appendix also reports descriptive percentages for very small groups. Neither high aggregate specificity nor a favorable estimate in a small subgroup establishes broad, precise performance across cancer types. [10]
Two different denominators should not be confused. Among participants with newly diagnosed primary cancers detected by the candidate, stages I–II accounted for 48.9% (44 of 90) in PATHFINDER 2 and 37.4% (97 of 259) in NHS-Galleri, and stages I–III for about 69% in both. That is the stage distribution among detected cancers, not the sensitivity among all cancers diagnosed at a given stage, and neither measure proves a survival benefit. [10]
The version bridge should stay visible. Among participants with a cancer diagnosis, positive percent agreement with MCED-V2 as the reference was 86.4% in PATHFINDER 2 and 80.7% in NHS-Galleri. In the pooled concordance analysis, the candidate was positive in 63.6% and 63.4% of MCED-V2-positive participants, respectively. FDA had no outstanding panel questions on the concordance, but the figures show that earlier-version results are not simply interchangeable with the candidate’s weighted estimates. [10]
Diagnostic consequences and randomized clinical benefit
A low false-positive rate does not remove the burden of investigating a positive result. In FDA’s PATHFINDER 2 safety analysis of MCED-V2, 159 of 218 participants who underwent diagnostic evaluation after a positive result (72.9%) had at least one invasive procedure. Among the 90 participants without a cancer diagnosis within 12 months, 43 had an invasive procedure and four had a surgical procedure. No study-related adverse event in this analysis was serious. [10]
False reassurance is a different concern. The panel stressed that a positive result does not confirm cancer, a negative result does not rule it out, and guideline-recommended screening must continue. NHS-Galleri did not return negative results to participants, which limits direct assessment of the behavioral effect of a communicated negative result. FDA judged a PATHFINDER 2 questionnaire measure of attitudes toward screening adherence not fit for purpose, because it correlated only weakly with actual screening behavior. [9] [10]
The randomized NHS-Galleri clinical-utility trial, published online on September 22, 2026, randomized 142,250 adults aged 50 to 77, collected blood at up to three annual visits and tested MCED-V2 added to usual care. It did not meet its primary endpoint: it did not demonstrate a statistically significant reduction in stage III or IV cancer incidence across 12 prespecified cancer types (incidence rate ratio 1.03; 95% confidence interval 0.92–1.14; P=0.63). A missed primary endpoint neither proves the absence of an effect nor shows equivalence. The key secondary endpoint, stage IV incidence across the same cancer types, had an incidence rate ratio of 0.86 (0.74–1.00). After the missed primary endpoint, this secondary result should be interpreted descriptively; it does not reverse the primary result or establish a mortality benefit. The authors call for further follow-up. [11] [13]
This randomized outcome is important context, but it is not a contradiction of the PMA performance estimates. FDA based its analysis on first-screening-round performance, stated that the proposed indication includes no claim of stage IV reduction and did not ask the panel to evaluate one. Performance, stage distribution and patient-outcome benefit answer different questions. [10]
The PATHFINDER 2 publication in Nature Medicine, also released on September 22, reports MCED-V2 12-month episode sensitivity of 39.3% (95% confidence interval 34.9–44.0%) for all cancers and specificity of 99.64%, among 32,007 performance-analyzable participants with completed 12-month cancer status assessments. These figures are not corrections of FDA’s candidate values of 35.0% and 99.85%: version, data cut, selection and weighting differ, so no isolated version effect can be derived. Version, cohort, endpoint and weighting must stay attached to every estimate. [12] [10]
What this means for submission planning
For surgical robotics, connect the system configuration to each applicable hazard and verification endpoint. Intended instruments and third-party products, software versions, service and use lives, reprocessing limits and training assumptions should be consistent across the submission. A gap assessment against the draft can identify questions for early FDA feedback; it should not be presented as mandatory implementation of an unfinished guidance. [1]
Clinical strategy should explain why bench or in vivo evidence suffices for each proposed use, or why clinical evidence is needed. Umbrella procedures should be justified procedure by procedure and device by device, not by specialty alone. For remote, autonomous or otherwise uncertain systems, the Q-Submission Program is a more defensible route to feedback than assuming that the local-teleoperation framework applies unchanged. [1] [6]
For multi-cancer and other classifier-based diagnostics, the Galleri review supports keeping the claim, the device version and the clinical evidence inseparable. Bridging should describe what changed and how it affects positive detection, false positives, origin prediction and downstream care. Labeling and evidence plans should address false reassurance and diagnostic follow-up as well as analytical accuracy. These are lessons from one product-specific review, not new requirements for every liquid-biopsy or machine-learning diagnostic. [9] [10]
The panel favored continued data collection on longer-term follow-up, performance across subgroups, real-world performance including the effect of false results, adherence to guideline-recommended screening and patient outcomes. Any approved indication, labeling restrictions or post-approval study obligations would be determined by FDA in its regulatory decision, not by the panel’s recommendations. A favorable advisory vote should be reported as a milestone in the review, not as a market authorization or a general acceptance of an evidence model for multi-cancer tests. [9]
For EU-facing development planning, we regard these U.S. materials as technical context, not as evidence that EU obligations have been met. Any conclusion under the EU Medical Device Regulation or the AI Act requires separate assessment against the applicable EU sources.
Image notice: The illustrative and header images were generated using AI and do not depict real people, places, or events. The portrait photographs show real people.
Relevant for your project?
Similar questions in your current project?
In a first call we clarify what is specifically relevant for your situation, without obligation.
Request a call →Regulations & standards considered
- FDA, Robotically-Assisted Surgical Devices: Premarket Submissions, draft guidance, September 25, 2026 (docket FDA-2026-N-9505)
- Section 524B of the FD&C Act (21 U.S.C. 360n-2), cybersecurity of cyber devices
- 21 CFR Parts 812, 56 and 50 (clinical investigations), including 21 CFR 812.2(b) and 812.28
- 21 CFR 10.115, good guidance practices
- FDA Molecular and Clinical Genetics Panel of the Medical Devices Advisory Committee, meeting of September 23, 2026 (PMA for the Galleri test)
FAQ
Frequently asked questions
Related expertise
FDA Strategy & Pathways →
Pathway choice between 510(k), De Novo and PMA, the routes a surgical robot or a first-of-a-kind diagnostic can take
Cybersecurity for Medical Devices →
Section 524B requirements for cyber devices, including penetration testing of connected systems
Usability Engineering →
Human-factors validation of the final interface with representative users and teams
Related projects
All case studies →Sources
- [1] FDA, Robotically-Assisted Surgical Devices: Premarket Submissions, draft guidance, document issued September 25, 2026. https://www.fda.gov/media/194987/download
- [2] Federal Register, notice of availability of the robotically-assisted surgical devices draft (FR Doc. 2026-19704, docket FDA-2026-N-9505), published September 25, 2026. https://www.federalregister.gov/d/2026-19704
- [3] FDA, guidance document page for the robotically-assisted surgical devices draft (docket FDA-2026-N-9505). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/robotically-assisted-surgical-devices-premarket-submissions
- [4] 21 CFR 10.115, good guidance practices. https://www.ecfr.gov/current/title-21/section-10.115
- [5] 21 U.S.C. 360n-2 (section 524B of the FD&C Act), ensuring cybersecurity of devices (U.S. Code, 2024 edition). https://www.govinfo.gov/link/uscode/21/360n-2
- [6] FDA, Requests for Feedback and Meetings for Medical Device Submissions: The Q-Submission Program, final guidance (May 29, 2025). https://www.fda.gov/media/114034/download
- [7] Alston & Bird, FDA Proposes Premarket Framework for Robotically Assisted Surgical Devices, advisory of October 13, 2026 (secondary source on the umbrella and covered-procedure paradigm). https://www.alston.com/en/insights/publications/2026/10/framework-robotically-assisted-surgical-devices
- [8] FDA, Public Workshop: Evaluating Benefit-Risk in Robotic Medical Devices with Autonomous or Remote Teleoperation Capabilities, December 2–3, 2026. https://www.fda.gov/news-events/fda-meetings-conferences-and-workshops/public-workshop-evaluating-benefit-risk-robotic-medical-devices-autonomous-or-remote-teleoperation
- [9] FDA, 24 Hour Summary, Molecular and Clinical Genetics Panel of the Medical Devices Advisory Committee, September 23, 2026. https://www.fda.gov/media/195032/download
- [10] FDA Executive Summary for the September 23, 2026, meeting of the Molecular and Clinical Genetics Panel (GRAIL, Galleri). https://www.fda.gov/media/194909/download
- [11] Sasieni P, et al., Effect of screening with multicancer early-detection test on late-stage cancer diagnosis, N Engl J Med, published online September 22, 2026, doi:10.1056/NEJMoa2505723. https://doi.org/10.1056/NEJMoa2505723
- [12] Nabavizadeh N, et al., Performance and safety of a multi-cancer early detection test: the PATHFINDER 2 study, Nat Med, published September 22, 2026, doi:10.1038/s41591-026-04618-w. https://doi.org/10.1038/s41591-026-04618-w
- [13] The ASCO Post, Multicancer Early Detection Screening Does Not Reduce Overall Late-Stage Cancer Diagnosis (secondary source on the NEJM results). https://ascopost.com/news/october-2026/multicancer-early-detection-screening-does-not-reduce-overall-late-stage-cancer-diagnosis/
- [14] 21 CFR Part 812, investigational device exemptions. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-H/part-812
- [15] FDA, Evaluating Benefit-Risk in Robotic Medical Devices with Autonomous or Remote Teleoperation Capabilities, Discussion Paper and Request for Feedback, November 2026 (not guidance), Docket FDA-2026-N-13135. https://www.fda.gov/media/196129/download
- [16] GRAIL, Inc., GRAIL Reports Third Quarter 2026 Financial Results, press release of November 4, 2026. https://www.biospace.com/press-releases/grail-reports-third-quarter-2026-financial-results
Related insights
All insights →Surgical robotics and diagnostics
Where does your evidence plan stand against the robotics draft?
The device, the intended claim and the planned submission are all we need at first.
Your inquiry goes to our team, which brings in the subject-matter experts as needed.
Prefer direct? +49 89 4161170-0
contact@theentourage.de
- Reply usually within one working day
- 3 locations: DE · CH · IT · projects in the US too
- 100% life sciences





