Blockchain: behind the buzzwords14 December 2018
As the hype around blockchain has exploded, its actual applications have become harder to discern. However, specific aspects of the technology have the potential to address the ongoing problem of reproducibility in clinical trials. Dr Steve Arlington and Dr Richard Shute of the Pistoia Alliance, and epidemiologist and blockchain specialist Dr Mehdi Benchoufi tell Tim Gunn what lies beyond the buzzwords.
It sounds almost dystopian, but more people trust the organisations that build their phones than those that take care of their bodies. Across the 28 countries surveyed for communications marketing company Edelman’s 2018 Trust Barometer, only 55% of consumers had faith in the pharmaceutical industry to “do what is right”. In the US, the world’s biggest pharmaceutical market, the rate was 38%, and in Germany, the fourth-largest, a paltry three out of ten people were confident that the sector has patients’ best interests at heart.
By contrast, despite the ongoing backlash against election meddling and the way data is used online, three quarters of respondents continued to believe in the moral scruples of technology companies.
Edelman’s barometer is a blunt measure, but it reflects real problems. A 2016 study into outcome switching found that just 13% of the clinical trials published in the top five medical journals between October 2015 and January 2016 reported their results correctly. The same year, a survey carried out by Nature indicated that more than 70% of researchers have tried and failed to reproduce another scientist’s experiments, and more than half have been unable to replicate their own findings.
For Dr Mehdi Benchoufi, a researcher looking for ways to tackle the reproducibility crisis at Hôpital Hôtel-Dieu in Paris, “it’s something like a curse”.
“Poor-quality studies compromise the whole system,” he continues. “As a medical professional, you don’t necessarily have the time to assess a publication, so you rely on the opinion of your peers. It is very strongly bound to trust.”
A question of trust
Appropriately, the most talked-about target for technological disruption is now trust itself.
Blockchain technology, of which the cryptocurrency bitcoin is the most famous example, is “trustless” in that it distributes trust through a whole network rather than relying on a particular guarantor. Of course, trustless isn’t quite a synonym for trustworthy, but blockchain is designed to make the distinction moot by automating how contracts are carried out and ensuring that no party can alter information without alerting all the others.
“It means you don’t have to rely on a single entity to do the right thing,” summarises Benchoufi.
That simple advantage is a powerful draw for the pharmaceutical industry. According to the Pistoia Alliance, a not-for-profit dedicated to supporting innovation in life sciences, 60% of pharma professionals are currently using or experimenting with blockchains, up from only 22% last year. What is more, in 2016 IBM found that clinical trial management was the most popular medical use case for blockchain among healthcare executives.
The statistics reflect something of Pistoia president Dr Steve Arlington’s gruffly qualified optimism. As he puts it, “Blockchain gives us an opportunity to build trust levels to a point whereby the first thing a regulator says isn’t, ‘So, how did they fiddle with this?’”
Arlington stresses the need to be “realistic” about what this technology can do for medicine. Blockchain is not so much a panacea as a type of shared database, and a slow one at that. The key selling point is that it can be directly and safely shared by entities that do not trust each other without relying on a central administrator, but this is paid for by the fact that blockchains are expensive to maintain and hard to upgrade.
Equally, the technology wasn’t built with medicine in mind. It was first developed to serve as the public transaction ledger for bitcoin, a digital currency cleared and controlled by a distributed network of computers rather than a central institution. That said, ‘trustlessness’ might be more relevant to clinical trials than currency. Centralised institutions and middlemen are far better at managing transactions than regulators and publications are at ensuring clinical trials follow the right protocols.
Timestamping using blockchain
In fact, Dr Greg Irving and Dr John Holden of the University of Cambridge have demonstrated that bitcoin itself can be used to unalterably timestamp original clinical trial documentation and automate the process of spotting changes and edits.
It’s a far more precise version of tracking criminals with marked money. Like other blockchains, bitcoin uses ‘hashes’, unique codes of a determined length (in this case 256 characters) that can be created from any form of digital media, to record transactions and attach them to a timestamp. As each block also contains the hash and address of its predecessor, the network continually fingerprints its entire history. This makes it easy to find and check individual transactions, and nearly impossible to falsify them.
Irving and Holden generated a hash from an original clinical protocol, converted it into a bitcoin key, and used that key to transfer bitcoin. This action wrote the key into a block on the chronological chain of hashes and recorded it across the network with a precise timestamp. From here, Irving and Holden showed that any change to the protocol document would be reflected in the fact that any future hash it might generate would differ from the timestamped one on the blockchain.
As Dr Richard Shute, a Pistoia Alliance consultant specialising in blockchain, explains, “You can prove categorically that a specific trial document generated a specific hash at a specific time, and you can be sure that if the data is ever tampered with in the future, the hashing around it will change.”
There are a number of applications designed to make bitcoin timestamping easy to do at scale, but using the world’s largest digital currency to manage trial data is a bit like holding athletics competitions in the Large Hadron Collider because it’s a circuit with good clocks. Benchoufi and Shute both recommend more flexible and fine-tuned blockchain platforms.
By way of an example, the Ethereum blockchain makes it possible to create non-transactional smart contracts: pieces of code that automate agreements where trust could traditionally be manipulated, executing as soon as pre-agreed conditions are met.
A method article by Timothy Nugent has shown that a private blockchain consisting of regulators, pharma companies and research associations could implement smart contracts to submit and cement trial protocols, agree and issue trial contracts, manage patient recruitment, read all of the trial data and provide a summary that timestamps every element.
As such, each step (or block) of a clinical trial can be chained together to verify that the designated methodology has been followed, keeping the trial as transparent as possible and preventing any reconstruction or beautification of data.
Shute emphasises the interest the US Food and Drug Administration is showing in leveraging these capabilities.
“If regulators were able to see an entire clinical package with all of its documents traced back through the blockchain,” he says, “they could track the trial’s history with confidence.” The result could be a significant decrease in the time and money necessary to develop a drug.
Nugent’s Ethereum method relies on data either being held on-chain or on a shared storage network. By contrast, Irving and Holden only recorded proof of their trial protocol on the blockchain, not the protocol itself.
“That’s a critical difference,” stresses Shute, “because, given privacy concerns and legal requirements, actually storing personal data on a blockchain, where you cannot remove it, is a big risk. You could just end up with a public database.”
As secure as it might be, data stored on blockchains and the storage networks linked to them cannot be changed, which means it is likely to fall foul of the EU’s General Data Protection Regulation, which enshrines individuals’ right to be forgotten.
Nevertheless, at a recent “blockchain bootcamp” hosted by the Pistoia Alliance, members discussed using the permissioned Hyperledger platform to facilitate clinical data sharing with searchable metadata and a blockchain-managed requesting system.
It’s an approach already being implemented by healthcare start-ups Timicoin and Medicalchain, which enable patients to finely control and even revoke access to their healthcare data, using Hyperledger to ensure each interaction with it is auditable, transparent and secure. For those organising clinical trials, these initiatives show promise for streamlining recruitment, which currently takes up 30–35% of their budgets, a considerable expense to make sure 20% of trials meet their enrolment deadlines. Shute calls it a “blockchain-enabled dating agency”.
It’s an interesting comparison. Dating apps store and sell a huge amount of embarrassingly personal information. And if the well-trusted technology industry has taught consumers anything, it’s that data, medical or otherwise, is a valuable asset. Blockchain allows patients to take control of it, putting the onus on those running trials to communicate with them effectively.
“Again, it builds trust,” says Shute. “There is a risk on individual organisations if they do not adopt blockchain technology in appropriate places. It shows they are not big bad ogres and wolves, but that they’re really trying to do the right thing by patients.”
That’s not to say it’s only patients that will benefit. One of the most intriguing applications for blockchain is ensuring the integrity of data recorded by devices like Apple watches and Fitbits through a process similar to timestamping. Shute notes that pharmaceutical companies are still being a little “coy” about how they are implementing the technology, but they certainly have the incentive to do it right.
“I don’t know the answers,” admits Arlington, “but I do know everything gets much more exciting when you can start using population-sized data.”