A Quick Look at a Critical FDA Discussion Paper

It seems a good time to update everyone on the FDA discussion paper for autonomous and remote teleoperation surgical systems. There’s a lot in this paper that I won’t go into here, but here are a few matters worth mentioning.

First, the FDA’s Center for Devices and Radiological Health released this paper in advance of a Dec. 2-3 meeting on these themes. Both days are scheduled to run from 10 a.m. to 4 p.m. Eastern, so there will be plenty of time for stakeholders to offer their perspectives.

Second, CDRH released a draft guidance for premarket submissions for robotically assisted surgical devices (RASDs), which is exclusive of autonomous surgical devices and remotely operated telesurgical systems. However, the agency stated that “certain recommendations” found in the RASD draft may apply to these out-of-scope systems. With that out of the way, let’s dive into what the Nov. 6 discussion paper says.

First, there are a few definitions of interest. One is that remote teleoperation can mean that the patient and the operator of the console can be in different rooms in the same clinical site or in entirely different nations. As a practical matter, there’s a huge difference between an operator located 2,500 feet from the surgical suite and one that’s 2,500 miles away.

Obviously, the reliability and the robustness of the data connection is a consideration, but it’s not clear whether those considerations will force a do-over of this definition. We might also ask how U.S. state boards of medical licensure will react to interstate telesurgery, but that’s for another day.

CDHR brings up the question of levels of autonomy, offering several examples without staking out a position. A recent paper on the subject in Science Robotics offers what might seem a plausible approach, starting with mere robot assistance at L-1. This consists of low-level mechanical operations as seen in existing FDA-cleared systems. Following that are levels 2 through 5, described as task autonomy, supervised autonomy, high-level autonomy and full autonomy, respectively.

Whether stakeholders will express a preference for a scalar approach as in the Science Robotics article as opposed to the approach taken in IEC/TR 60601-4-1 is tough to predict, but we’ll know soon enough. There are other options, but both industry and FDA might conclude that simplicity is a virtue one ignores at one’s own peril.

Risk as the pivotal question regarding FDA oversight

The question of risk pops up on several occasions in the discussion paper. For instance, page 3 states that uncertainty plays a role in the agency’s understanding of the risks associated with a device type. Page 9 highlights some specific considerations of risk for autonomous systems, such as reduced user vigilance and limited adaptability to intra- and inter-procedural variability.

At the risk of stating the obvious (ahem), risk may be the central and defining question regarding autonomous and remote telerobotic surgery. The key question there might be whether autonomy comes across to CDRH as elevating risk high enough to reflexively apply a class III, high-risk designation to autonomous surgical robots, at least in the early going. We remind ourselves that CDRH routinely (if sometimes belatedly) down-classifies device types for risk when circumstances warrant, so a class III designation isn’t necessarily interminably sticky.

Before you make the point that FDA regulates non-autonomous robotic surgical systems as medium-risk, class II devices, allow me to point out that there are surgical meshes that are designated class III for some purposes and class II for others. It’s not as if there’s no precedent and it wouldn’t be at all difficult for a regulatory agency to see a significant difference in risk, particularly when full autonomy is part of the picture.

One imagines that special controls might suffice to deal with the risks of remote telesurgery, but this scenario brings the provider of the data connection into the picture. We’ll see how much interest bandwidth providers have in being dragged into the FDA regulatory regime without a big boost in fees, particularly given the fact that they’d likely have to upgrade their liability insurance policies.

For a moment, let’s consider the question of adoption of autonomous surgical systems. One thing that bears mentioning – and CDRH doesn’t bring it up this in the paper – is that autonomous systems can’t replace surgeon intuition. I know how squishy that sounds, but it is difficult to ignore the fact that neural networks still aren’t even close to replicating the integration of learnings that we take for granted in our daily lives, both at work and away from work. Surveys of outcomes, such as this one, demonstrate that experience is critical in outcomes for many procedures, a fact of surgical life with which CDHR staff are surely familiar.

Lingering questions about the next gatekeeper

We might also ask about payer coverage as coverage policies might add drag to this picture. There might not be a Medicare benefit category problem, but it’s not clear whether the software needed for either autonomous or remote teleoperation falls into an existing category for software as a service (which CMS will rephrase as software as a medical service, or SaMS, as of Jan. 1, 2027).

Bear in mind that even remote teleoperation might necessarily incorporate AI, but whether payers will cover the virtual/augmented reality software and equipment for remote telesurgery is up for grabs. The solution might be in the practice expense part of the Medicare physician fee schedule, but we’ll have to wait and see how that washes out.

So that’s a quick look at the FDA discussion paper and some of the implications. If you think I’ve kicked wide on any of this, please feel free to ping me at LinkedIn. I’d love to hear from you.