FDA Device Regulation

In the interests of length I had to sacrifice a few points in my WSJ review of Innovation Breakdown by Joseph Gulfo (excerpted on MR yesterday). In the review, I argued that the FDA could speed the approval of medical devices and reduce uncertainty by not reviewing directly but becoming a certifier of certifiers as is done in Europe.

In fact, a US model is already in place. OSHA, the Occupational Safety and Health, requires that a range of electrical products and materials meet certain safety standards but it outsources certification to Underwriters Laboratories and other Nationally Recognized Testing Laboratories. We could and should do the same for medical devices and for drugs. Indeed, if a device or drug is permitted in a developed, advanced economy such as in Europe, Australia and Japan then I see no reason why it ought not to be provisionally approved in the United States (and vice-versa).

My paper with DiMasi and Milne showed that some FDA drug divisions appear to be much more productive than other divisions suggesting possibilities for substantial improvements if best practices were uniformly adopted. There also appear to be substantial differences between the regulation of drugs and devices especially in recent years. Ian Hathaway and Robert Litan have a new paper on Entrepreneurship and Job Creation in the U.S. Life Sciences Sector that shows that new firm creation in the medical device sector has fallen drastically since 1990 and far more than in the drug sector. Although there are likely many causes, the drop in the number of new firms is consistent with Gulfo’s experience of regulatory uncertainty and may suggest increases in regulatory cost for devices relative to drugs. Here is Hathaway and Litan:

The medical devices and equipment sector, on the other hand, saw new firm formations decline steadily and persistently between 1990 and 2011—falling by 695 firms or 53 percent during that period. Its share of new life sciences firms fell to 31 percent in 2011 from 50 percent in 1990. Unlike its life sciences sector counterparts, the decline in new firm formations in this segment appears to stretch beyond the cyclical effects of the Great Recession.

new device firms


Yet the EU has recognized the major deficiencies of their system and are reforming their regulations now.

Would readers be surprised to know that the majority of devices are approved without any clinical data?

And that is in the US.

"the majority of devices are approved without any clinical data".

More information is needed. What kind of devices are we talking about here? Do they all need clinical data?

I agree you don't need clinical data to prove a new kind of tongue depressor isn't going to kill someone, but this issue actually extends to many high risk devices, such as cardiovascular products, like pacemakers.


Thanks, that is helpful. However, what should the correct approach be? Requiring a new clinical trial for every small change to an electronic device like a pacemaker would be prohibitively expensive, and would probably slow down the improvement of pacemakers.

Part of the problem in this space is that innovation that does not happen also harms people's health, but unlike failures of approved devices, this harm is not visible (and is hard to measure). The regulatory system tends to be biased towards preventing the visible harm, rather than striking a balance between allowing innovation and preventing failures. The policy this article refers to seems like a reasonable attempt to strike such a balance.

Yes, it is a tough issue to develop a policy on that strikes a fair safety/innovation balance. I think one of the main issues is "device creep", whereby there can be literally dozens of serial updates to an original device, each approved by FDA without clinical data, because each change on its own is seemingly superficial. Eventually, through all the additive changes, you end up with a product that can be very different from the original and has never been tested in people. I think there should be some limit to the number of changes without re-testing.

Another issue is the weak post-market monitoring of products. If our surveillance system was better, FDA could more comfortably approve changes with the confidence safety problems could be detected soon.

Third, there are two whole classes of devices (Class I and Class II), which make up of the vast majority of approvals, that are deemed to be low and medium risk and generally do not require clinical data, even when they are _first_ approved. I think devices need to be periodically re-assessed to determine whether they are in the correct category. For example, all-metal artificial hips on the market blew up in recent years due to a safety issue, causing tons of revision surgeries and affecting tens of thousands of patients. These types of products were cleared through FDA's less stringent review pathway because they had been grandfathered in to the low-risk category, meaning clinical data was never mandatory upon first approval and only required on an as-needed basis by FDA. (Cardiac devices are already in the high-risk category, so this wouldn't apply to them, but one would think hip implants should have been considered high-risk products).

All good points, I agree. Especially about the post-approval monitoring.

I would like to see a system that is less about a specific "approval" line and more about continuous monitoring of all devices. There would be devices with various levels of evidence available and less of a clear distinction between trial and use of the product. Instead, patients would be offered an informed choice based on the level of available evidence for products.

Jan, can you point us to any cases where this "design creep" has caused harm?

With the many device recalls due to design problems over the years, it would be extremely difficult to say that any particular design issue was precisely because of design creep (as opposed to a design issue more generally). I am not aware of anyone having attempted such an analysis. However, it is clear that if companies are permitted to bypass clinical testing based on similarity to older devices, then that decreases the level of information available on these products and ability to spot safety problems before they are marketed.

To give you an example how weak the links between an original device and its progeny can be, here is an excerpt from a paper.

The Pathwork Tissue of Origin Test, cleared in 2008, is a microarray kit that compares the RNA expression pattern from a tumor with an unknown primary to the expression patterns of 15 common tumors [43]. This device's predicate device was the BioPlex 2200 Medical Decision Support Software, a software algorithm cleared in 2005 that assists in diagnosing autoimmune disorders by matching enzyme-linked immunoassay results to a database of sera from patients with autoimmune disorders [44]. This device, in turn, had been declared substantially equivalent to the Remedi HS Drug Profiling System, an algorithm-based diagnostic kit cleared in 1995 that tests for illicit drugs. Thus, a screening test for illicit drugs ultimately allowed for the clearance of a malignancy diagnostic test, simply because both use computer programs to compare samples to an existing database.


Recognizing the difficulty of attributing clinical failure specifically to design creep may be difficult, your example mostly tells me the FDA is simply incompetent. Those do not sound like incremental design refinements. The fact that they were considered as such seems more an issue with FDA mismanagement than an issue with auto-approval of incremental design changes.

It's a tough balance whenever the regulatory framework allows a lot of reviewer judgment. The agency undoubtedly makes ill-advised decisions from time to time, but if they don't allow certain products on the market or require testing and data that aren't explicitly required by law or regulations the companies will certainly avail themselves of the opportunity to complain about regulatory burden and arbitrary demands.

Jan, you are a clown. That is all.

Or metal on metal artificial hips:


If someone wants to write a paper on how the FDA has killed innovation look at the hearing aid business in the US versus Europe.

Audio device for the hard of hearing? — A medical device, must regulate, delay, raise costs.
Audio device for other human needs? —A consumer product, let people choose.
Wheeled vehicle for fun? — Great! Just satisfy some safety regulations.
Wheeled vehicle called a “wheel chair”? — It’s a “medical device”, so we must delay for approval and make it needlessly expensive.

Does anyone know/recall why Tyler rejected this kind of thing as a possible cause of "The Great Stagnation"?

I read the book but the only solution to TGS I recall being offered was that we should dramatically increase the status and income of academics.

Are new firm formations really the best way to measure device innovation? When I consider drugs, I note that it often may make sense for a small company to form around a single potential drug, betting the farm on whether or not that drug will end up working. So you may see an increase in drug firms.

Yet it doesn't follow to me that devices would follow this same structure. For example, if many device innovations are modifying or working off previous devices, a market with a lot of innovation may end up being dominated by fewer firms piling up more and more patents.

Wouldn't a more accurate measure of innovation then be something along the lines of the # of new device applications per year over time?

" Indeed, if a device or drug is permitted in a developed, advanced economy such as in Europe, Australia and Japan then I see no reason why it ought not to be provisionally approved in the United States (and vice-versa)."

Doesn't this encourage forum shopping? Apply to whichever regulatory agency you think is most likely to approve the drug, then present your approval to the stricter agencies and say, "Look, I passed!"

Yes, and this has been a major criticism of the EU device regulatory system. http://www.ft.com/cms/s/0/71730f00-7fd2-11e0-b018-00144feabdc0.html#axzz3AIyRAxrW

There is a lack of transparency of review processes and the bases for approval in the EU, so it is hard to even know if these third party reviewing companies (who are paid by the regulated companies) are effectively evaluating the products : http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.1001276

http://pharmamkting.blogspot.com/2006_06_01_archive.html has a nice chart for medical device patents filed by year. Unfortunately it only goes through 2003. However the chart clearly shows medical device patents leaped from 4500 in 1990 to 9091 in 2003. During that period the graph clearly shows new medical device company formation dropping a lot.

This hints that we haven't seen a dry up of medical device innovation due to a dysfunctional FDA. I'd love to see the data from 2003 to 2014 but just looking at the 90's where new device companies started a 'persistent' decline we nonetheless saw plenty of new devices.

This is only a good measure of innovation if the use of patents stayed the same over time. Because this time frame corresponds with the proliferation of software patents, there is good reason to suspect it did not.

You are saying some patents classified 'medical device' are actually software patents instead? Presumaby they are medically related software patents to get lumped in as 'medical devices'. And that would still count as innovation unless you are saying the same number of software based innovations were happening in the early 90's but they just didn't bother to patent them.

Regardless, I'm not seeing why new company formation is any better as a measure.

Good day! Do you use Twitter? I'd like to follow you if that
would be ok. I'm undoubtedly enjoying your blog and look forward
to new updates.

Comments for this post are closed