search

ConsenSysMesh / web3studio-projects

Discussion board for web3studio projects
4 stars 1 forks source link

TrialBlind #48

Open humbitious opened 5 years ago

humbitious commented 5 years ago

Pharmaceuticals conducting clinical trials have a problem. They need to know that nobody in their trial is in more than one trial, because it would be dangerous for the patient to be taking more than one untested drug (unknown interactions), and a person "double-spending" themselves into multiple trials will invalidate the trials. The trick here is for a system to be able to give each party (CRO, Pharma, etc) a signal, "Person is or is not in another trial," without giving any party any other information about any other. For example, Pfizer should not be able to learn through use of the system anything meaningful about Merck's clinical trials, including metadata like the number of trials or the volume of patients in them. This would have to contemplate gaming of the system, like making many patient/participant queries and using them to map the landscape of participants and then use some kind of external information to train an AI to create a high-confidence mapping of people to other parties' trials.

This is a trivial problem if you can trust a third party to manage the system and keep different parties' data separate while running a deduplicated list of participants.

brian-lc commented 5 years ago

I'm not sure this is a trivial problem even if you can trust a third party. We dealt with this exact issue at at a previous place I worked and it was with a trusted internal system. The core of the reason that this is actually incredibly challenging is that it involves identifying people. If you go off info a person provides; they can use different names, data can be incorrect or incomplete, data changes over time (like names, addresses, phone numbers, etc...). Maybe if you used some bio-metric system you might side-step some of those problems. But then you're back to getting all the clinical trial operators to have/use the same system.

humbitious commented 5 years ago

Yea

I'm not sure this is a trivial problem even if you can trust a third party. We dealt with this exact issue at at a previous place I worked and it was with a trusted internal system. The core of the reason that this is actually incredibly challenging is that it involves identifying people. If you go off info a person provides; they can use different names, data can be incorrect or incomplete, data changes over time (like names, addresses, phone numbers, etc...). Maybe if you used some bio-metric system you might side-step some of those problems. But then you're back to getting all the clinical trial operators to have/use the same system.

Yeah, this is the key issue, because a single company can set up a protocol for ensuring (to some hopefully-acceptable degree) that nobody in one of their trials is in another trial, but this is irrelevant to the FDA and other authorities. They care if the patient is in ANY other trial, regardless of which CRO or pharma is running it. There is in the industry a centralized company (may be others, but folks I know only know of one) that handles this for many CROs and pharmas. They do provide the protocols for identity verification and maintain that black box data, so that participating organizations can get that "true/false" signal on the candidate without knowing anything else about the other providers' trials. The question is, a) would doing this with web3 technology remove some of the liability and cost from such an org and help ensure that such entities have no way to jigger with the data in the event something goes wrong that would give them a strong incentive to change history (of course, if this were a major issue, then their clients could just insist on them integrating with Chainpoint and hashing state to prove that no tampering occurred after the fact); b) would advanced-to-come in digital identity verification (self-sov, biometric, etc etc) make a web3 solution 10x more certain to reduce patients trying to get into multiple trials?

brian-lc commented 5 years ago

I would like to know more about "patients trying to get into multiple trials." If the people trying to get onto multiple trials are sufficiently motivated to do so I wonder what those reasons are and if there are other non-technical ways to discourage them. If the solution must be technical then some biometric ID mapped to a distributed trial ledger should be enough? The only information in the ledger would be the hashed bio-id, and an expiration date, presence of a non-expired id should be sufficient to produce a true/false and not give any other information, right?

On Sat, Apr 13, 2019 at 10:01 AM John Wolpert notifications@github.com wrote:

Yea

I'm not sure this is a trivial problem even if you can trust a third party. We dealt with this exact issue at at a previous place I worked and it was with a trusted internal system. The core of the reason that this is actually incredibly challenging is that it involves identifying people. If you go off info a person provides; they can use different names, data can be incorrect or incomplete, data changes over time (like names, addresses, phone numbers, etc...). Maybe if you used some bio-metric system you might side-step some of those problems. But then you're back to getting all the clinical trial operators to have/use the same system.

Yeah, this is the key issue, because a single company can set up a protocol for ensuring (to some hopefully-acceptable degree) that nobody in one of their trials is in another trial, but this is irrelevant to the FDA and other authorities. They care if the patient is in ANY other trial, regardless of which CRO or pharma is running it. There is in the industry a centralized company (may be others, but folks I know only know of one) that handles this for many CROs and pharmas. They do provide the protocols for identity verification and maintain that black box data, so that participating organizations can get that "true/false" signal on the candidate without knowing anything else about the other providers' trials. The question is, a) would doing this with web3 technology remove some of the liability and cost from such an org and help ensure that such entities have no way to jigger with the data in the event something goes wrong that would give them a strong incentive to change history (of course, if this were a major issue, then their clients could just insist on them integrating with Chainpoint and hashing state to prove that no tampering occurred after the fact); b) would advanced-to-come in digital identity verification (self-sov, biometric, etc etc) make a web3 solution 10x more certain to reduce patients trying to get into multiple trials?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ConsenSys/web3studio-projects/issues/48#issuecomment-482811815, or mute the thread https://github.com/notifications/unsubscribe-auth/AAAGWFNAXdsU-o7IqCsYfigLW-jJBLptks5vgeM7gaJpZM4cXtim .

humbitious commented 5 years ago

There's no question that the industry works hard on discouraging people from entering multiple trials. And every manner of identification (including biometric) protocol is being tried by many CROs. The problem, confirmed by Heather, is that whatever one company's or one group's protocol may be, they have no basis upon which to coordinate. One CRO might have excellent ways of discouraging or identifying people trying to enter multiple trials, but they simply don't know if they are being effective, because they can't confirm their patients aren't in trials outside their purview.

(The motives for attempting to get into multiple-trials are typically monetary (you're a student and want the money and aren't thinking about what it could do to your body), placebo gaming (you really want to get an actual therapeutic, and you know there is a good chance you are getting the placebo, so you sign up for different studies), and desperation (you are very sick and you conspire with your doctor to get into many trials in a last-ditch effort...not ethical, but not all doctors are ethical, and not all terminal patient studies are as lenient about entrance -- because while the patient is going to die anyway, the mixing of trials can cause results that will confuse and warp the data -- as the patient would want).