Clinical Trial on a Public Blockchain | Hacker News Reader

Clinical Trial on a Public Blockchain

medium.com

26 pointsjedixit18 comments

The authors of this article don't seem to understand that "Blockchain technology" don't prevent people from simply entering fake data. Many of the things discussed in this article, for example informed consent forms, can be easily bypassed by simply either fabricating consent forms or entering data with fake dates.

At the end of the day, the hard part with clinical trial security isn't securing the data/documents (the FDA as a centralized trusted authority does quite a good job); the hard part is ensuring that the data entered actually matches reality. Blockchain technology doesn't help at all with that problem. Blockchain technology is only really useful when you don't have a centralized trusted authority like the FDA. However, in most legal scenarios (like clinical trials), a trusted centralized authority is readily and cheaply available and thus blockchain technology is worse than useless: it only adds additional cost and complexity with minimal benefit.

PS: I highly recommend people check out https://clinicaltrials.gov/ It's the US's public repository of clinical trials that contains details about many clinical trials and results. That one website probably contains a good fraction of our total human knowledge of what drugs/procedures do.

Also a visualized summary of all the clinical trials conducted by 10 of the largest pharma companies in a digestable format:

https://www.aerodatalab.org/birds-eye-view-of-research-lands... https://www.statnews.com/2019/07/18/clinical-trials-birds-ey...

A centralized, trusted authority does a good job until it doesn't.

Unlike what else? I don't think you are saying anything.

The main property of a blockchain is that the data is copied over and over to lots of places owned by many different entities. However, the "magic" already fades when it comes to the important part of the technology, deciding what version is authoritative made even harder by lots of otherwise unconnected people having to agree. The algorithms used for this indeed are "magical" and not easy to understand in real life (incl. various attack scenarios; even if such an algorithm itself appears to be simple on its own the complexity arising of running it in the distributed system is not), not to mention the resource use (of both storing the same data many times over and of consensus algorithms, even the ones without "proof of work" requirement).

Of course, this second part is necessary because of the first part, because you now have lots and lots of copies you now have the problem of having to make the decision. That adds a lot of unforeseeable consequences because making that decision - and in breadth, in a distributed system - is a hard problem. You trade one issue that you have with a single place with a lot of fuzzy and hard to foresee issues in a complex system. It seems to me to many people "blockchain" simply is attractive because it is harder to understand and has "magical" properties, just like writing notes on paper automatically appears to be "so last millennium" compared to storing digital notes even if for a given situation it would do the job just as well (and require only low-tech hardware and no electrical power).

Also reminds me of "Why did you climb that mountain?" - "Because it was there" to some degree. I mean, it's not bad that there seems to be an ingrained attitude to favor new shiny things in parts of humanity, that's probably good as a progress-driving factor for all things where outcomes cannot be predicted, so that they are attempted to be used anyway (similar to some "errors" that seems like waste of resources in some biological systems where inefficient pathways are nevertheless left open a bit, inefficient at the moment but enabling better reactions to changes). So I don't mind all the blockchain (and other) hype(s) too much, but let's keep it in check a bit, let there be experiments but let's not go overboard.

Published hashes can prevent people from entering fake data at a later time though, which is quite interesting. (It has nothing to do with blockchain though.)

That problem is already solved though by the FDA (and clinicaltrials.gov). Once you submit your documents to the trusted authority you can't modify them without it being super obvious.

This is only done after all data has been collected an analysed (from multiple clinical trials!) and when a company is looking for market authorization. So clearly if anyone wants to alter their data, they would do this prior to submitting their study dossier to the FDA. This has happened time and time again....

It is therefore crucial to make sure data cannot be altered AS SOON AS IT IS COLLECTED. At this point, one would not yet know how to alter the data and if that would steer results in the 'right' direction (researchers are blinded for instance).

So I definitely see value in creating a cryptographic hash of a data point, and immediately registering it in some immutable database (which could be blockchain, or a different solution I suppose)

Data might pass through many different hands between collection and submission the the FDA. I can totally imagine a clinical trial where one lab tech enters some results into an Excel spreadsheet, and that sheet gets emailed to another person, and that sheet gets modified in some way, on and on. There aren’t a ton of guarantees you can make about how accurate the data is once it makes it to submission, so I could see some sort of hashing/blockchain tech being mandated within an organization to certify results.

The data in a clinical trial are collected at research sites with many different investigators, spreading many different countries. These investigators are generally not the people who 'alter the data', it are the CROs or Pharmaceutical companies that alter the data set AFTER all data has been collected and analysed. So if the collected data is registered in some immutable manner at the research sites, it becomes (near) impossible for the CRO to alter that data after analysis. This is where blockchain adds value.

The clinicaltrials.gov repository only records high level findings of a clinical trial, again AFTER data analysis. This in no way prevents researchers from entering fake data, if the results are not as they want (registering it at the moment it is collected, when researchers don't know what the results would be yet, would make this more difficult).

The comparison with the FDA (that supposedly makes it impossible to alter data after submission) does not make a lot of sense. This is only done prior to market application after years and years of performing clinical trials (where in the mean time one could alter the data in anyway they want). Moreover, the FDA actually stipulates that all data and documents need to be recorded in a time-stamped and immutable manner, so I'm certain they would welcome the use of blockchain technology here (which they do, have a look at their website).

I completely agree, however, that making sure the entered data matches the actual data remains a persistent challenge (not because of fraudulent behaviour but because of human errors)

If you read their academic paper linked at the bottom, they actually undermine their entire system in a laughably silly way:

> For each user, a pair of private-public keys were provided (https://msdn.microsoft.com/fr-fr/library/windows/desktop/aa3...). These are asymmetric cryptographic data that enable authentication on Blockchain. These were randomly attached in one-to-one correspondence to the user’s emails. We focused on Blockchain’s usage in the time-stamping and archiving logic. We did not let users create or use their own Blockchain authentication setup (i.e., if a user owned a Bitcoin account, the key and Bitcoin address were not allowed to be used). This restriction was related not to the Blockchain complexity but rather to maintain a simple and common email-focused authentication process. Other ways for authentication include the physical devices USB keys or cell phone fingerprints, but this would have been outside the focus of our protocol-related problematics.

So basically, we didn't want users to have to bother with this whole private key thing, so we'll hold on to those private keys for them. In other words, they have full control of whats written on the blockchain and users just trust them - which is the exact thing blockchains are trying to get rid of.

https://f1000research.com/articles/6-66/v5?source=post_page-...

Except very few want to manage their keys and are more than happy to have a custodian.

Blockchain doesn't solve human behavior, or much of anything IRL.

>Blockchain doesn't solve human behavior, or much of anything IRL.

Blockchain solves a very select few problems, which are important, but not every problem.

Imagine if I told you I could turn on your coffee machine with a SQL database. Why the hell would you care?

When you think blockchain, think database. When you think database, think Excel. Can your problem be solved with Excel, and your stakeholders made happy with it? If the answer is yes, YOU DON'T NEED BLOCKCHAIN.

You are comparing two different authors here....

One is a company using blockchain technology in clinical trials, the other is a researcher postulating a situation for using blockchain in clinical trials.

Don't compare apples with oranges

This isn't their whitepaper at all. The whitepaper you mention is from a separate researcher. Has nothing to do with Triall and or this thread at all.

The Triall whitepaper can be found at https://www.triall.io/#whitepaper

Why use that tech - "But, but, but, but it's blockchain!" (I don't have a smiley for beating head against wall.) Sounds like almost all blockchain projects - doesn't solve any of their problems ... but it's blockchain ... fashionable amongst the ignorant. The only practical purpose for blockchain is to gain project funding.

How about https://blog.smartdec.net/you-do-not-need-blockchain-eight-p...

That fact that you can prove to all stakeholders that the data about a triall is complete, hasn't been altered is very powerful when you have many parties and many documents, where you still need to provide proofs in the future.

This solution allows you to prove it on an individual document level, all the way up till dossier or clinical trial level. Meaning you can prove that not a single document has been added, edited or deleted in ten years time for instance.

Clinical trials are all about collecting trustworthy data that demonstrate safety and efficacy. Large trials can collected over a million data points.

I personally see a lot of value in creating cryptographic hashes of these data points and documents, and registering them in an immutable database. This would prove that the data/document is not altered at a later time, and increases its trustworthiness.

However, I see quite a few commenters here who do not see that value. For you I have the following question:

"How would you prove that a data point or document has not been altered"?

(No, this cannot be done by the FDA or Clinicaltrials.gov)

Please bear in mind that researchers could indeed use fake data in these cryptographic hashes. But at the moment it is collected, they would have no idea what they should alter because they are double-blinded. The results and implications only become clear when analysing the data after de-blinding. So proving the validity prior to analysis is where the value lays.

Reply on news.ycombinator.com