There’s another reason why Pfizer actually did the right thing with Enbrel in Alzheimer’s

The Washington Post recently ran a story about a potential use of Enbrel, an anti-inflammatory drug, to reduce the risk of Alzheimer’s disease. The article reports that a non-randomized, non-interventional, retrospective, non-published, non-peer-reviewed, internal review of insurance claims was correlated with a reduced risk for Alzheimer’s disease. This hypothesis was dismissed after internal consideration, on the basis of scientific (and probably business) considerations.

You can probably guess my take on it, given the way I describe it, but for people who are unfamiliar with the hierarchy of medical evidence, this kind of data mining represents pretty much the lowest, most unreliable form of medical evidence there is. If you were, for some reason, looking for even lower-level evidence than this, you would need to go into the case study literature, but even that would at least be peer-reviewed and published for public scrutiny. If you need further convincing, Derek Lowe has already published a fairly substantial debunking of why this was not a missed Alzheimer’s opportunity on his blog.

Many people, even after being convinced that Pfizer probably made the right call in not pursuing a clinical trial of Enbrel in Alzheimer’s, still think that the evidence should have been made public.

There’s been lots of opinions thrown around on the subject, but there’s one point that people keep missing, and it relates to a paper I wrote in the British Medical Journal (also the final chapter of my doctoral thesis). Low-level evidence that causes suspicion of activity of a drug, when it is not swiftly followed up with confirmatory testing, can create something that we call “Clinical Agnosticism.”

Not all evidence is sufficient to guide clinical practice. But in the absence of a decisive clinical trial, well-intentioned physicians or patients who have exhausted approved treatment options may turn to off-label prescription of approved drugs in indications that have not received regulatory sanction. This occurs where there is a suggestion of activity from exploratory trials, or in this case, extremely poor quality retrospective correlational data.

Pfizer should not be expected to publish every spurious correlation that can be derived from any data set. In fact, doing so would only create Clinical Agnosticism, and potentially encourage worse care for patients.

On techbros in healthcare and medical research

My thoughts on the following may change over time, but at least for me, I have found it helpful to think about the threats that techbros pose to healthcare and medical research in terms of the following four major categories. These aren’t completely disjoint of course, and even the examples that I give could, in many cases, fit under more than one. I am also not claiming that these categories exhaust the ways that techbros pose a threat to healthcare.

1. Medical-grade technosolutionism, or “medicine plus magic”

When Elizabeth Holmes founded her medical diagnostics company Theranos, she fit exactly into the archetype that we all carry around in our heads for the successful whiz-kid tech-startup genius. She was not just admitted to Stanford University, but she was too smart for it, and dropped out. She wore Steve Jobs-style black turtlenecks. She even founded her company in the state of California—innovation-land.

She raised millions of dollars for her startup, based on the claim that she had come up with a novel method for doing medical diagnostics—dozens of them—from a single drop of blood. Theranos claimed that the research backing up these claims occurred outside the realm of peer-review. This practice was derisively called “stealth research,” and generally criticized because of the threat that this mode of innovation might have posed to the enterprise of medical research as a whole.

It was, of course, too good to be true. The company has now been shut down, and Theranos has been exposed as a complete fraud. This sort of thing happens on a smaller scale on crowdfunding sites with some regularity. (Remember the “Healbe” Indiegogo?)

While I’m not entirely sure what is driving this particular phenomenon, I have a few pet theories. For starters, we all want to believe in the whiz-kid tech-startup genius myth so much that we collectively just let this happen out of sheer misguided hope that the techbros will somehow get it right. And on some level, I understand that impulse. Medical research progress is slow, and it would be wonderful if there actually were a class of smart and talented geniuses out there who could solve it by just applying their Apple Genius powers to the matter. Alas, it is not that easy.

And unfortunately, there is a certain kind of techbro who does think like that: “I’m a computer-genius. Medicine is just a specialized case that I can just figure out if I put my mind to it.” And there’s also a certain kind of medical professional who thinks, “I’m a doctor. I can figure out how to use a computer, thank-you-very-much.” And when those two groups of people intersect, sometimes they don’t call each other out on their lack of specialized knowledge, but rather, they commit synergy.

And worse, all this is happening under an extreme form of capitalism that has poisoned our minds to the extent that the grown-ups—the people who should know better—are turning a blind eye because they can make a quick buck.

Recommended reading: “Stealth Research: Is Biomedical Innovation Happening Outside the Peer-Reviewed Literature?” by John Ioannidis, JAMA. 2015;313(7):663-664.

2. The hype of the week (these days, it’s mostly “medicine plus blockchain”)

In 2014 I wrote a post on this very blog that I low-key regret. In it, I suggest that the blockchain could be used to prospectively timestamp research protocols. This was reported on in The Economist in 2016 (they incorrectly credit Irving and Holden; long story). Shortly thereafter, there was a massive uptick of interest in applications of the blockchain to healthcare and medical research. I’m not claiming that I was the first person to think about blockchain in healthcare and research, or that my blog post started the trend, but I am a little embarrassed to say that I was a part of it.

Back in 2014, being intrigued by the novelty of the blockchain was defensible. There’s a little bit of crypto-anarchist in all of us, I think. At the time, people were just starting to think about alternate applications for it, and there was still optimism that the remaining problems with the technology might still be solved. By 2016, blockchain was a bit passé—the nagging questions about its practicality that everyone thought would have been solved by that point just, weren’t. Now that it’s 2019, and the blockchain as a concept has been around for a full ten years, and I think it’s safe to say that those solutions aren’t coming.

There just aren’t any useful applications for the blockchain in medicine or science. The kinds of problems that medicine and science have are not the kinds of problems that a blockchain can solve. Even my own proposed idea from 2014 is better addressed in most cases by using a central registry of protocols.

Unfortunately, there continues to be well-funded research on blockchain applications in healthcare and science. It is a tech solution desperately in search of its problem, and millions of research funding has already been spent toward this end.

This sort of hype cycle doesn’t just apply to “blockchain in science” stuff, although that is probably the easiest one to spot today. Big new shiny things show up in tech periodically, promising to change everything. And with surprising regularity, there is an attempt to shoehorn them into healthcare or medical research.

It wasn’t too long ago that everyone thought that smartphone apps would revolutionize healthcare and research. (They didn’t!)

3. “The algorithm made me do it”

Machine learning and artificial intelligence (ML/AI) techniques have been applied to every area of healthcare and medical research you can imagine. Some of these applications are useful and appropriate. Others are poorly-conceived and potentially harmful. Here I will gesture briefly toward some ways that ML/AI techniques can be applied within medicine or science to abdicate responsibility or bolster claims where the evidence is insufficient to support them.

There’s a lot of problems that could go under this banner, and I’m not going to say that this is even a good general overview of the problems with ML/AI, but many of these major problems stem from the “black box” nature of ML/AI techniques, which is a hard problem to solve, and it’s almost a constitutive part of what a lot of ML/AI techniques are.

The big idea behind machine learning is that the algorithm “teaches itself” in some sense how to interpret the data and make inferences. And often, this means that many ML/AI techniques don’t easily allow for the person using them to audit the way that inputs into the system are turned into outputs. There is work going on in this area, but ML/AI often doesn’t lend itself well to explaining itself.

There is an episode of Star Trek, called “The Ultimate Computer,” in which Kirk’s command responsibilities are in danger of being given over to a computer called the “M-5.” As a test of the computer, Kirk is asked who he would assign to a particular task, and his answer differs slightly from the one given by the M-5. For me, my ability to suspend disbelief while watching it was most thoroughly tested when the M-5 is asked to justify why it made the decision it did, and it was able to do so.

I’ve been to the tutorials offered at a couple different institutions where they teach computer science students (or tech enthusiasts) to use the Python library for machine learning or other similar software packages. Getting an answer to “Why did the machine learning programme give me this particular answer?” is really, really hard.

Which means that potential misuses or misinterpretations are difficult to address. Once you get past a very small number of inputs, there’s rarely any thought given to trying to figure out why the software gave you the answer it did, and in some cases it becomes practically impossible to do so, even if you wanted to.

And with the advent of “Big Data,” there is often an unspoken assumption is that if you just get enough bad data points, machine learning or artificial intelligence will magically transmute them into good data points. Unfortunately, that’s not how it works.

This is dangerous because the opaque nature of ML/AI may hide invalid scientific inferences based on analyses of low-quality data, causing well-meaning researchers and clinicians who rely on robust medical evidence to provide poorer care. Decision-making algorithms may also mask the unconscious biases built into them, giving them the air of detached impartiality, but still having all the human biases of their programmers.

And there are many problems with human biases being amplified while at the same time being presented as impartial through the use of ML/AI, but it’s worth mentioning that these problems will of course harm the vulnerable, the poor, and the marginalized the most. Or, put simply: the algorithm is racist.

Techbros like Bill Gates and Elon Musk are deathly afraid of artificial intelligence because they imagine a superintelligent AI that will someday, somehow take over the world or something. (I will forego an analysis of the extreme hubris of the kind of person who needs to imagine a superhuman foe for themselves.) A bigger danger, and one that is already ongoing, is the noise and false signals that will be inserted into the medical literature, and the obscuring of the biases of the powerful that artificial intelligence represents.

Recommended reading: Weapons of Math Destruction by Cathy O’Neil.

4. Hijacking medical research to enable the whims of the wealthy “… as a service”

I was once at a dinner party with a techbro who has absolutely no education at all in medicine or cancer biology. I told him that I was doing my PhD on cancer drug development ethics. He told me with a straight face that he knew what “the problem with breast cancer drug development” is, and could enlighten me. I took another glass of wine as he explained to me that the real problem is that “there aren’t enough disruptors in the innovation space.”

I can’t imagine being brazen enough to tell someone who’s doing their PhD on something that I know better than them about it, but that’s techbros for you.

And beyond the obnoxiousness of this anecdote, this is an idea that is common among techbros—that medicine is being held back by “red tape” or “ethical constraints” or “vested interests” or something, and that all it would take is someone who could “disrupt the industry” to bring about true innovation and change. They seriously believe that if they were just given the reins, they could fix any problem, even ones they are entirely unqualified to address.

For future reference, whenever a techbro talks about “disrupting an industry,” they mean: “replicating an already existing industry, but subsidizing it heavily with venture capital, and externalizing its costs at the expense of the public or potential workers by circumventing consumer-, worker- or public-protection laws in order to hopefully undercut the competition long enough to bring about regulatory capture.”

Take, for example, Peter Thiel. (Ugh, Peter Thiel.)

He famously funded offshore herpes vaccine tests in order to evade US safety regulations. He is also extremely interested in life extension research, including transfusions from healthy young blood donors. He was willing to literally suck the blood from young people in the hopes of extending his own life. And these treatments were gaining popularity, at least until the FDA made a statement warning that they were dangerous and ineffective. He also created a fellowship to enable students to drop out of college to pursue other things such as scientific research outside of the academic context. (No academic institution, no institutional review board, I suppose.)

And this is the work of just one severely misguided techbro who is able to make all kinds of shady research happen because of the level of wealth that he has been allowed to accumulate. Other techbros are leaving their mark on healthcare and research in other ways. The Gates Foundation for example, is strongly “pro-life,” which is one of the strongest arguments I can think of, for why philanthropists should instead be taxed and the funds they would have spent on their conception of the public good dispersed through democratic means, rather than allowing the personal opinions of an individual become de facto healthcare policy.

The moral compass behind techbro incursions into medical research is calibrated to a different North than the one most of us recognize. Maybe one could come up with a way to justify any one of these projects morally. But you can see that the underlying philosophy (“we can do anything if you’d just get your pesky ‘ethics’ out of the way”) and priorities (e.g. slightly longer life for the wealthy at the expense of the poor) are different from what we might want to be guiding medical research.

Why is this happening and what can be done to stop it?

Through a profound and repeated set of regulatory failures, and a sort of half-resigned public acceptance that techbros “deserve” on some level to have levels of wealth that are comparable with nation states, we have put us all in the position where a single techbro can pervert the course of entire human research programmes. Because of the massive power that they hold over industry, government and nearly every part of our lives, we have come to uncritically idolize techbros, and this has leaked into the way we think about applications of their technology in medicine and science. This was all, of course, a terrible mistake.

The small-picture solution is to do all the things we should be doing anyway: ethical review of all human research; peer-review and publication of research (even research done with private funds); demanding high levels of transparency for applications of new technology applied to healthcare and research; etc. A high proportion of the damage they so eagerly want to cause can probably be avoided if all our institutions are always working at peak performance and nothing ever slips through the cracks.

The bigger-picture solution is that we need to fix the massive regulatory problems in the tech industry that allowed techbros to become wealthy and powerful in the first place. Certainly, a successful innovation in computer technology should be rewarded. But that reward should not include the political power to direct the course of medicine and science for their own narrow ends.

Proof of prespecified endpoints in medical research with the bitcoin blockchain

Introduction

The gerrymandering of endpoints or analytic strategies in medical research is a serious ethical issue. “Fishing expeditions” for statistically significant relationships among trial data or meta-analytic samples can confound proper inference by statistical multiplicity. This may undermine the validity of research findings, and even threaten a favourable balance of patient risk and benefit in certain clinical trials. “Changing the goalposts” for a clinical trial or a meta-analysis when a desired endpoint is not reached is another troubling example of a potential scientific fraud that is possible when endpoints are not specified in advance.

Pre-specifying endpoints

Choosing endpoints to be measured and analyses to be performed in advance of conducting a study is a hallmark of good research practice. However, if a protocol is published on an author’s own web site, it is trivial for an author to retroactively alter her own “pre-specified” goals to align with the objectives pursued in the final publication. Even a researcher who is acting in good faith may find it less than compelling to tell her readers that endpoints were pre-specified, with only her word as a guarantee.

Advising a researcher to publish her protocol in an independent venue such as a journal or a clinical trial registry in advance of conducting research does not solve this problem, and even creates some new ones. Publishing a methods paper is a lengthy and costly process with no guarantee of success—it may not be possible to find a journal interested in publishing your protocol.

Pre-specifying endpoints in a clinical trial registry may be feasible for clinical trials, but these registries are not open to meta-analytic projects. Further, clinical trial registry entries may be changed, and it is much more difficult (although still possible) to download previous versions of trial registries than it is to retrieve the current one. For example, there is still no way to automate downloading of XML-formatted historical trial data from www.clinicaltrials.gov in the same way that the current version of trial data can be automatically downloaded and processed. Burying clinical trial data in the “history” of a registry is not a difficult task.

Publishing analyses to be performed prior to executing the research itself potentially sets up a researcher to have her project “scooped” by a faster or better-funded rival research group who finds her question interesting.

Using the bitcoin blockchain to prove a document’s existence at a certain time

Bitcoin uses a distributed, permanent, timestamped, public ledger of all transactions (called a “blockchain”) to establish which addresses have been credited with how many bitcoins. The blockchain indirectly provides a method for establishing the existence of a document at particular time that can be independently verified by any interested party, without relying on a medical researcher’s moral character or the authority (or longevity) of a central registry. Even in the case that the NIH’s servers were destroyed by a natural disaster, if there were any full bitcoin nodes left running in the world, the method described below could be used to confirm that a paper’s analytic method was established at the time the authors claim.

Method

  1. Prepare a document containing the protocol, including explicitly pre-specified endpoints and all prospectively planned analyses. I recommend using a non-proprietary document format (e.g. an unformatted text file or a LaTeX source file).
  2. Calculate the document’s SHA256 digest and convert it to a bitcoin private key.
  3. Import this private key into a bitcoin wallet, and send an arbitrary amount of bitcoin to its corresponding public address. After the transaction is complete, I recommend emptying the bitcoin from that address to another address that only you control, as anyone given the document prepared in (1) will have the ability to generate the private key and spend the funds you just sent to it.

Result

The incorporation into the blockchain of the first transaction using the address generated from the SHA256 digest of the document provides an undeniably timestamped record that the research protocol prepared in (1) is at least as old as the transaction in question. Care must be taken not to accidentally modify the protocol after this point, since only an exact copy of the original protocol will generate an identical SHA256 digest. Even the alteration of a single character will make the document fail an authentication test.

To prove a document’s existence at a certain point in time, a researcher need only provide the document in question. Any computer would be able to calculate its SHA256 digest and convert to a private key with its corresponding public address. Anyone can search for transactions on the blockchain that involve this address, and check the date when the transaction happened, proving that the document must have existed at least as early as that date.

Discussion

This strategy would prevent a researcher from retroactively changing an endpoint or adding / excluding analyses after seeing the results of her study. It is simple, economical, trustless, non-proprietary, independently verifiable, and provides no opportunity for other researchers to steal the methods or goals of a project before its completion.

Unfortunately, this method would not prevent a malicious team of researchers from preparing multiple such documents in advance, in anticipation of a need to defraud the medical research establishment. To be clear, under a system as described above, retroactively changing endpoints would no longer be a question of simply deleting a paragraph in a Word document or in a trial registry. This level of dishonesty would require planning in advance (in some cases months or years), detailed anticipation of multiple contingencies, and in many cases, the cooperation of multiple members of a research team. At that point, it would probably be easier to just fake the numbers than it would be to have a folder full of blockchain-timestamped protocols with different endpoints, ready in case the endpoints need to be changed.

Further, keeping a folder of blockchain-timestamped protocols would be a very risky pursuit—all it would take is a single honest researcher in the lab to find those protocols, and she would have a permanent, undeniable and independently verifiable proof of the scientific fraud.

Conclusion

Fraud in scientific methods erodes confidence in the medical research establishment, which is essential to it performing its function—generating new scientific knowledge, and cases where pre-specified endpoints are retroactively changed casts doubt on the rest of medical research. A method by which anyone can verify the existence of a particular detailed protocol prior to research would lend support to the credibility of medical research, and be one less thing about which researchers have to say, “trust me.”