The 2020 annual conference of the Petrie-Flom Center for Health Law Policy, Biotechnology, and Bioethics (at Harvard Law School) will focus on the future of medical device regulation. If you’re interested in participating, abstracts are due on October 14, final papers are due March 27, and the conference is May 8. (Here’s a link to the call for papers.) I’m working on an empirical project relating to device premarket approval and patent term restoration, but sadly the dataset won’t be ready in time. Meanwhile, here’s a run-down of some interesting legal scholarship on medical device regulation in the last year and a half.
Unlike the drug provisions of the FDCA, the device provisions include an express preemption provision. Under section 521 of the statute, or 21 U.S.C. § 360k, no state may establish or continue in effect … any requirement (1) which is different from, or in addition to, any requirement applicable under the FDCA to the device, and (2) which relates to the safety or effectiveness of the device or to any other matter included in a requirement applicable to the device under the FDCA. A trilogy of Supreme Court cases grappled with this provision between 1996 and 2001, and you can get the skinny on all the subsequent lower court action through the Drug & Device Law blog.
Several pieces of legal scholarship in the last year have considered the relative roles of tort law and federal regulatory requirements in assuring the safety of medical devices in the market. I describe four below. In brief: Professors Gentry and McMichael show empirically that the Supreme Court’s 2008 decision in Riegel — finding preemption of a state law failure-to-warn claim relating to a Class III device approved by FDA — may have prompted companies to seek (and obtain) approval of even higher risk devices. Preemption is so pervasive, now, that Professor Rabin (Stanford) suggests device safety must be assured through changes in the FDA regulatory paradigm, rather than through application of tort law. Professor Tschider agrees tort liability is nearly completely foreclosed but she argues tort law has an important complementary role to play. So she lays out an argument that patients injured by AI-enabled medical devices might use in court to avoid the preemption rulings. Finally, Professor Uzdavines has more of a burn-it-down-and-start-over approach; she would rewrite federal and state law so that tort law packs more of a punch, and then use preemption to entice more companies to submit to (more aggressive) regulatory oversight. The details below.
Professor Marilyn Uzdavines, at Nova Southeastern University, argues that the current medical device regulatory framework is clumsy in design — that safe devices are over-regulated while unsafe devices are under-regulated. To give a concrete example, she would have FDA focus on cryotherapy chambers rather than exam gloves. The title of her piece (Dying for a Solution: The Regulation of Medical Devices Falls Short in the 21st Century Cures Act, posted September 2018) is a bit misleading; the article is not really about the 21st Century Cures Act. The first 20 pages tell the history of FDA’s medical device authority, including the initial 1938 enactment, the Dalkon Shield controversy and enactment of the 1976 Medical Device Amendments, various later device safety controversies prompting enactment of the Safe Medical Devices Act of 1990, and the call for faster access that prompted changes in the Food and Drug Administration Modernization Act of 1997. She adds a few pages describing three provisions of the 21st Century Cures Act relating to medical devices, concluding that the statute was “one sided” — ensuring faster access to medical devices while reducing regulatory oversight.
But the heart of the article is Section III. And I found myself wishing she had skipped or shortened sections I and II — which are clear and interesting but have been done before — and blown up Section III into a full article. Here she argues that Class I and II devices should be exempt from premarket notification, and that manufacturers of Class II devices that voluntarily seek out premarket approval should benefit from express preemption of state law tort claims. Preemption could serve as a carrot, she argues, encouraging companies to submit to greater regulatory scrutiny. She also favors a damages multiplier and punitive damages for unapproved medical devices that cause injury, thus using tort law as a “stick” to prompt companies to produce safer products. These proposals — which would require changes in both federal and state law — merit much more than the six pages she gave them, so let’s hope she has more work planned.
In Reassessing the Regulation of High-Risk Medical Device Cases (posted May 2019), Professor Robert Rabin of Stanford Law School, with Alyssa Picard (expecting her JD from Stanford next May) consider the express preemption provision in 21 U.S.C. § 360k, as interpreted in Medtronic v. Lohr and Riegel v. Medtronic. In the first case, the Supreme Court declined to find preemption of a state tort claim when FDA had cleared a device for market pursuant to a premarket notification and a finding of “substantial equivalence” to another marketed device. In the second, it found preemption of a failure-to-warn claim relating to a device for which FDA had approved a premarket approval application (PMA). The article is brief, but the authors question the assumption that devices subject to premarket approval have been the subject of rigorous scrutiny for safety and effectiveness. New evidence may arise after regulatory approval, they point out, as Justice Ginsburg did in the Riegel dissent. Post-market clinical studies are inconsistently completed, they add, and the adverse event reporting system for medical devices may be under-used. The authors wrap up with the third major device preemption case from the Supreme Court: Buckman v. Plaintiffs’ Legal Committee, in which the Court found implied preemption of a state law claim grounded exclusively on violation of FDA requirements. What’s left? They briefly survey the landscape in the lower courts, comment on the “strength of the preemption bar,” and conclude that tort law can play only a “limited, complementary role in improving the safety of medical devices on the market.” This leads to their take-away, that reform at the agency level is the best way to improve device safety.
In Responses to Liability Immunization: Evidence from Medical Devices (posted August 2019), Elissa Philip Gentry (Florida State) and Benjamin McMichael (University of Alabama) consider how manufacturers responded to the Riegel decision in 2008, which they characterize as unexpected. (They base this on the fact that several justices in the earlier Lohr decision had suggested the express preemption provision might preempt only state statutes.) The authors reason that Riegel eliminated some expected costs for devices subject to the PMA requirement, which should have lowered the anticipated profits needed to justify market entry and thus should have increased the number of filings. They hypothesize that firms were more likely to submit applications for high risk devices after Riegel. (They identify “high risk” devices through recall data and adverse event reporting data.) Working with devices approved from 1997 to 2015, they find that — consistent with their hypothesis — the change in the number of approvals for high risk medical device product categories is “significantly positive” relative to the change in approvals for low risk categories. This suggests that device firms reacted to the Riegel decision.
But this is really two studies, in one. The second half of the article considers whether the shifting liability landscape following Riegel — a safe harbor for the manufacturer, presumably causing patients to turn on medical malpractice suits against their doctors — led doctors to use fewer devices that would benefit from preemption. Now they use an all-payer dataset for inpatient care (roughly 1000 hospitals) and focus on treatments selected for heart attack and stroke patients before and after Reigel. The results suggest doctors in areas of the country with high malpractice risk (those with no cap on non-economic damages) chose to use stents less often after the Riegel decision, compared to doctors in low risk areas of the country (those with a cap). The authors conclude that the Riegel decision may ironically undermine the purpose of the Medical Device Amendments: it may lead to approval of riskier devices and at the same time reduce patient access to medical devices.
Finally, in Preempting the Artificially Intelligent Machine (posted September 2019), Charlotte Tschider — a Visiting Assistant Professor at the University of Nebraska — takes us back to more basic questions about the relationship between tort law and FDA oversight. She focuses on FDA regulation of medical devices that use machine learning, and she’s concerned with design-based injuries rather than the “failure to warn” claims at issue in Riegel and Medtronic.
Part I of the article provides a readable overview of medical uses of artificial intelligence, and it concludes by describing FDA’s public statements to date. The focus is FDA’s April 2019 discussion paper, which described a possible regulatory framework for “software as a medical device” that incorporates artificial intelligence (AI) and machine learning technologies. In this paper, FDA proposes a “total product lifecycle” approach to regulation of these software products. For instance, the agency would conduct an initial premarket review to obtain a reasonable assurance of safety and effectiveness, giving the manufacturer an option to submit a plan for modifications, and assessing the company’s ability to manage and control the risks resulting from the modifications. After that, it would view the learning, adaptation, and optimization inherent to these devices as “modifications” to the device. The firm would monitor its device and evaluate the modifications based on the risk to patients, using an approach already described in existing guidance. (Note: this is my description of the discussion paper, not hers!) Part II of her article summarizes the current state of medical device preemption law, her main points being that (1) at this point, tort recovery is nearly completely foreclosed and will probably prevent recovery for AI-enabled medical devices cleared through the premarket approval (PMA) process even if FDA hasn’t reviewed the AI system, but (2) the PMA system isn’t well designed for AI enabled medical devices, especially because these devices will function differently after approval than at the time of approval.
The heart of her article is Part III, which argues that an effective product safety regimen would allow regulatory requirements and tort recovery to complement each other, especially because the tort system provides a mechanism for information dispersion and plaintiff compensation that the regulatory system does not. Ultimately, I think, she is trying to say that preemption isn’t warranted in these cases, because FDA doesn’t completely review the technology (and lacks the expertise to do so) — that this is true as a matter of doctrine, because the agency hasn’t done the full cost-benefit calculation that would merit taking tort law’s complementary role off the table. (This argument also provides the predicate for her solution in Part IV, meaning the roadmap she offers for arguments against preemption that could be made in court, to avoid Riegel and land closer to Lohr with an AI-enabled medical device injury.) Tschider also makes the policy argument: FDA’s planned approach shifts responsibility from the agency to companies, and if this regulatory approach is taken by the courts as equivalent to PMA approval, the companies will enjoy preemption while benefiting from less stringent FDA review. Firms will “benefit twice,” she says, while patients will “lose.”
All four very interesting articles, worth reading together as a set. The Tshider article makes an especially nice contribution by taking the arguments into a new and complicated area of medical device law — and by explaining the technology so clearly.