4 ms·

For some background - I'm a T1D working on a search engine and conversational interface for integrating a bunch of new data sources and models into metabolic de

For some background - I'm a T1D working on a search engine and conversational interface for integrating a bunch of new data sources and models into metabolic decision making: https://replica.health https://replica.health. I've also been a user of and worked on various open source artificial pancreas systems through the years, and am currently on Loop. >I would have thought by now they'd have hacked 2 patch-pump AIDs to work simultaneously. As you pointed out, the problem is not really hardware. It could technically be done in a straightforward way using two independently controlled insulin pumps, but the complexity and risk of the whole operation goes way up if you are taking way more insulin. Taking a bunch of insulin and glucagon at the same time is not necessarily a great idea either - they don't just annihilate each other without consequence and you could end up with secondary effects like gaining a bunch of weight. >Because this feels like a holy grail / functional cure Unfortunately it is not; even dual hormone systems have problems keeping up w/ the kinetics of glucose absorption and to address this there is also research into tri-hormonal systems, w/ amylin as the third hormone. In any case you will still need some a-priori info about meals and planned activities, though less so than with a single hormone system. Integration of exogenous data sources to provide this info to the APS is what we are working on at Replica. Also, hate to be the bearer of bad news but beta bionics has shelved their dual hormone ambitions for now; their prototype device soon to be released is insulin-only. On the bright side there is a small Dutch company whose tech predates beta-bionics. They sell a dual hormone device and will give it to you for a ton of $ (and probably have you sign a bunch of waivers): https://www.inredadiabetic.nl/en/discover-the-ap/ https://www.inredadiabetic.nl/en/discover-the-ap/

fyhn 3y ago
This is very interesting, and I (T1D) would like to learn more. However, the e-mail signup doesn't work for me. ("Oops! Something went wrong while submitting the form.")
haldujai 3y ago
> In any case you will still need some a-priori info about meals and planned activities Not necessarily, at least not via patient input. In the albeit small Inreda studies manual announcement of exercise and meals wasn't required (or an option). Medtronic also has a meal prediction algorithm on their newest offering that's a step towards a fully automated process and currently more or less obviates carb counting but isn't at the point where you don't have to announce a meal (yet). Rather than integrating external data sources the algorithms are predicting based on historical glucose levels and/or insulin administration and it seems to be working. https://jamanetwork.com/journals/jamasurgery/article-abstract/2795853 https://jamanetwork.com/journals/jamasurgery/article-abstrac...
- jablongo 3y ago
  Agree that patient input shouldn't be necessary, but to replace it we will need to include other inputs besides CGM in a systematic way to get the optimal results. My company is working on how to use contextual info automatically collected by your devices to help (detected activity, measured calorie burn, geofencing, data from meal-ordering apis, etc.). This is especially true given that the CGM data themselves are lagged due to averaging and/or kalman filtering going on under the hood. This is a fundamental problem; Inreda uses two identical CGMs for noise reduction purposes just so they can get clean data with less of a lag. None of the systems claiming you don't have to do anything in terms of meal announcement are _working_ in the sense of achieving euglycemic parity, which should be the goal. I can say with certainty that the cgm logs from people who don't announce meals on the Inreda device do not look like they are from non-diabetics: there are still often large post-prandial spikes. Inreda likely does better than any single hormone system, but the problem is not solved in any sense.
  - haldujai 3y ago
    > Agree that patient input shouldn't be necessary, but to replace it we will need to include other inputs besides CGM in a systematic way to get the optimal results. I'm not going as far as to claim Medtronic's approach (I believe the only one commercially available with so-called meal prediction based on historical CGM and offers full correction boluses) is the optimal one, just that it is an approach that is at least very good (~80% time in target) and while it still requires meal announcements it's just the first step of what they're trying to do. Clearly we can expect further iterations of these algorithms as the technology matures. > Inreda uses two identical CGMs for noise reduction purposes just so they can get clean data with less of a lag. Just giving an example that this is possible without external input or data, your statement was that you will need a-priori information which is not necessarily the case. Whether such a system is optimal is a different question. I haven't seen the raw data and highly doubt enough of it even exists for anyone to make a claim whether or not such a system can be optimized to the point necessary. > None of the systems claiming you don't have to do anything in terms of meal announcement are _working_ in the sense of achieving euglycemic parity, which should be the goal. For clarity to any less knowledgeable readers while time spent in euglycemia is a very important outcome measure it cannot come at the expense of severe hypoglycemia or severe hyperglycemia/diabetic ketoacidosis (i.e. an algorithm that improves euglycemia to 95% but has a 2% severe hypo time is less acceptable than 80% euglycemia and 0.5% severe hypo.) To my knowledge no system on the market/generally available right now is claiming to be completely input free. The closest to my knowledge is again the MiniMed 780G discussed in my first point which will assuredly be iterated on. Also to be clear I'm not being dismissive of what your company is working on, it's a very interesting and novel approach. It may even be necessary to achieve the optimal product. I look forward to reading about your results when you publish them. I'm just presenting alternatives and a brief overview of what other approaches are for HN readers who are likely unfamiliar with the topic being discussed.
Communitivity 3y ago
So I do not have a medical background. I have however worked a number of Industrial Control Systems (ICS) projects, and what you describe kind of sounds like a PID control loop, which also is not a simple push up/down approach. From Wikipedia:"A proportional–integral–derivative controller (PID controller or three-term controller) is a control loop mechanism employing feedback that is widely used in industrial control systems and a variety of other applications requiring continuously modulated control. A PID controller continuously calculates an error value e(t) as the difference between a desired setpoint (SP) and a measured process variable (PV) and applies a correction based on proportional, integral, and derivative terms (denoted P, I, and D respectively), hence the name." Getting PID control loops to work has a lot of research behind it, but it's still hard to get right with new hardware. I would imagine a PID control loop involving organics (wetware) would be much order, and harder still a PID control loop in organics with life-threatening failures possible.
- jablongo 3y ago
  At this point all the major solutions are using an algorithm that would probably fall under the umbrella of "Model Predictive Control" rather than a vanilla PID controller. An absurd spate of patent trolling occurred back in the early 2000s related to Controller definitions like these though. Researchers patented the use of a "PID controller in artificial pancreas systems" [1][2] which slowed down the development of APSs by many years. The way the Supreme Court recently changed patent law [3] for software has definitely had a positive effect for APS development. [1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2769814/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2769814/ [2] https://patents.google.com/patent/US20150306314 https://patents.google.com/patent/US20150306314 [3] https://en.wikipedia.org/wiki/Alice_Corp._v._CLS_Bank_International https://en.wikipedia.org/wiki/Alice_Corp._v._CLS_Bank_Intern...
  - tremon 3y ago
    An absurd spate of patent trolling occurred back in the early 2000s related to Controller definitions like these That's good news, right? That means that everything that people now come up with has well-documented prior art?
nahsra 3y ago
> https://replica.health https://replica.health Oh my gosh, this is the startup I considered starting last year when this issue became personal to me and I wrapped my brain around the complexity of insulin dosing. There are a lot of challenges here but we absolutely need an external "brain" to correlate many data points, some trends, and reasonably estimate current insulin sensitivity.
- jablongo 3y ago
  I'd love to chat - send me an email at sam@replica.health . An accurate time-varying insulin sensitivity model is one of our big projects.
  - selimthegrim 3y ago
    I formerly used to work in the diabetes management space, and I too would be interested in chatting seeing as I have acquired some modeling expertise in the meantime.
  - andrewchilds 3y ago
    Good luck with Replica! I built something sort of similar for my family (my 8 yo has T1D and a lot of life-threatening food allergies): https://saltie.app https://saltie.app. Happy to chat as well, I can create an account for anyone interested, and I can share what I’ve learned in making this.
gitfan86 3y ago
This is awesome. The obvious next step here would be integration with the pump to send the correct dose after the user confirms the app is correct in the food they are eating
agawish 3y ago
Sorry I know this is off topic, but I'm a recently diagnosed T1D and I would like to get more information about Loop and other loop-like open-source systems.
- wombatpm 3y ago
  My son is T1D since age 7. He is now on the OmniPod pump and the Dexcom G6 sensor. It’s been a very effective combination that’s covered by insurance. The challenge with the open source projects is limited equipment that can be hacked.
  - jablongo 3y ago
    Most people on loop have omnipod and dexcom g6 also so I don’t think it’s that limited. It’s true that it takes a few months for new devices to get integrated but open source APS are the only systems lets you choose what devices you want to use with them.
- Jarmsy 3y ago
  One thing I wish I'd been told more about when first diagnosed was the possibility of prolonging the honeymoon period. It was decades ago, and back then the advice was to eat a fairly normal diet and just dose as much insulin as needed. If I was back in that position I'd go low carb high exercise from the start, and do everything I could to reduce my insulin requirements as much as possible. There are cases of new T1s coming of insulin completely, or just a low basal dose for multiple years.
thecosas 3y ago
Bummer to hear about beta bionics! My understanding was that the hardware they are releasing does have a place for glucagon, just not the software and additional reservoir required. I'm considering a switch to their pump since it seems like they've made it as automated as possible with insulin-only. Would love to see/hear more about what you're working on!
Loic 3y ago
My father was an MD (psychiatrist) and had this saying: - 1 drug, it works ok, still very patient specific. - 2 drugs at the same time, only if you validated correctly at least one alone before. - 3 at the same time: clusterfuck/total mess. Base rule: do everything possible to use only one active ingredient/drug at a time.
- robobro 3y ago
  Yes, much easier to control a single variable. I'm a type 1 diabetic on multiple daily injections -- a pump just introduces a lot of inconvenience for me, price aside. You have to track batteries and change tubing and make sure the tubing doesn't snag on things, and unplug the machine before taking a shower or bath, and ... blah blah blah... All I need for MDI is to carry my syringes and vial around in a cold pack (I live in the tropics, lol) and I don't need to worry about hardware beyond that.
  - Jarmsy 3y ago
    I'm in the same boat re. MDI over pumping. I can still keep very tight control without a bunch of cumbersome attachments (I do wear a flash sensor though). Why vials and syringe rather than a pen though? I usually just carry the pen in my pocket and phone for monitoring. Even in hot weather I've not had problems with insulin losing potency noticeably before finishing a cartridge.

fyhn 3y ago

This is very interesting, and I (T1D) would like to learn more. However, the e-mail signup doesn't work for me. ("Oops! Something went wrong while submitting the form.")

haldujai 3y ago

> In any case you will still need some a-priori info about meals and planned activities Not necessarily, at least not via patient input. In the albeit small Inreda studies manual announcement of exercise and meals wasn't required (or an option). Medtronic also has a meal prediction algorithm on their newest offering that's a step towards a fully automated process and currently more or less obviates carb counting but isn't at the point where you don't have to announce a meal (yet). Rather than integrating external data sources the algorithms are predicting based on historical glucose levels and/or insulin administration and it seems to be working. https://jamanetwork.com/journals/jamasurgery/article-abstract/2795853 https://jamanetwork.com/journals/jamasurgery/article-abstrac...

jablongo 3y ago
Agree that patient input shouldn't be necessary, but to replace it we will need to include other inputs besides CGM in a systematic way to get the optimal results. My company is working on how to use contextual info automatically collected by your devices to help (detected activity, measured calorie burn, geofencing, data from meal-ordering apis, etc.). This is especially true given that the CGM data themselves are lagged due to averaging and/or kalman filtering going on under the hood. This is a fundamental problem; Inreda uses two identical CGMs for noise reduction purposes just so they can get clean data with less of a lag. None of the systems claiming you don't have to do anything in terms of meal announcement are _working_ in the sense of achieving euglycemic parity, which should be the goal. I can say with certainty that the cgm logs from people who don't announce meals on the Inreda device do not look like they are from non-diabetics: there are still often large post-prandial spikes. Inreda likely does better than any single hormone system, but the problem is not solved in any sense.
- haldujai 3y ago
  > Agree that patient input shouldn't be necessary, but to replace it we will need to include other inputs besides CGM in a systematic way to get the optimal results. I'm not going as far as to claim Medtronic's approach (I believe the only one commercially available with so-called meal prediction based on historical CGM and offers full correction boluses) is the optimal one, just that it is an approach that is at least very good (~80% time in target) and while it still requires meal announcements it's just the first step of what they're trying to do. Clearly we can expect further iterations of these algorithms as the technology matures. > Inreda uses two identical CGMs for noise reduction purposes just so they can get clean data with less of a lag. Just giving an example that this is possible without external input or data, your statement was that you will need a-priori information which is not necessarily the case. Whether such a system is optimal is a different question. I haven't seen the raw data and highly doubt enough of it even exists for anyone to make a claim whether or not such a system can be optimized to the point necessary. > None of the systems claiming you don't have to do anything in terms of meal announcement are _working_ in the sense of achieving euglycemic parity, which should be the goal. For clarity to any less knowledgeable readers while time spent in euglycemia is a very important outcome measure it cannot come at the expense of severe hypoglycemia or severe hyperglycemia/diabetic ketoacidosis (i.e. an algorithm that improves euglycemia to 95% but has a 2% severe hypo time is less acceptable than 80% euglycemia and 0.5% severe hypo.) To my knowledge no system on the market/generally available right now is claiming to be completely input free. The closest to my knowledge is again the MiniMed 780G discussed in my first point which will assuredly be iterated on. Also to be clear I'm not being dismissive of what your company is working on, it's a very interesting and novel approach. It may even be necessary to achieve the optimal product. I look forward to reading about your results when you publish them. I'm just presenting alternatives and a brief overview of what other approaches are for HN readers who are likely unfamiliar with the topic being discussed.

Communitivity 3y ago

So I do not have a medical background. I have however worked a number of Industrial Control Systems (ICS) projects, and what you describe kind of sounds like a PID control loop, which also is not a simple push up/down approach. From Wikipedia:"A proportional–integral–derivative controller (PID controller or three-term controller) is a control loop mechanism employing feedback that is widely used in industrial control systems and a variety of other applications requiring continuously modulated control. A PID controller continuously calculates an error value e(t) as the difference between a desired setpoint (SP) and a measured process variable (PV) and applies a correction based on proportional, integral, and derivative terms (denoted P, I, and D respectively), hence the name." Getting PID control loops to work has a lot of research behind it, but it's still hard to get right with new hardware. I would imagine a PID control loop involving organics (wetware) would be much order, and harder still a PID control loop in organics with life-threatening failures possible.

jablongo 3y ago
At this point all the major solutions are using an algorithm that would probably fall under the umbrella of "Model Predictive Control" rather than a vanilla PID controller. An absurd spate of patent trolling occurred back in the early 2000s related to Controller definitions like these though. Researchers patented the use of a "PID controller in artificial pancreas systems" [1][2] which slowed down the development of APSs by many years. The way the Supreme Court recently changed patent law [3] for software has definitely had a positive effect for APS development. [1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2769814/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2769814/ [2] https://patents.google.com/patent/US20150306314 https://patents.google.com/patent/US20150306314 [3] https://en.wikipedia.org/wiki/Alice_Corp._v._CLS_Bank_International https://en.wikipedia.org/wiki/Alice_Corp._v._CLS_Bank_Intern...
- tremon 3y ago
  An absurd spate of patent trolling occurred back in the early 2000s related to Controller definitions like these That's good news, right? That means that everything that people now come up with has well-documented prior art?

nahsra 3y ago

> https://replica.health https://replica.health Oh my gosh, this is the startup I considered starting last year when this issue became personal to me and I wrapped my brain around the complexity of insulin dosing. There are a lot of challenges here but we absolutely need an external "brain" to correlate many data points, some trends, and reasonably estimate current insulin sensitivity.

jablongo 3y ago
I'd love to chat - send me an email at sam@replica.health . An accurate time-varying insulin sensitivity model is one of our big projects.
- selimthegrim 3y ago
  I formerly used to work in the diabetes management space, and I too would be interested in chatting seeing as I have acquired some modeling expertise in the meantime.
- andrewchilds 3y ago
  Good luck with Replica! I built something sort of similar for my family (my 8 yo has T1D and a lot of life-threatening food allergies): https://saltie.app https://saltie.app. Happy to chat as well, I can create an account for anyone interested, and I can share what I’ve learned in making this.

gitfan86 3y ago

This is awesome. The obvious next step here would be integration with the pump to send the correct dose after the user confirms the app is correct in the food they are eating

agawish 3y ago

Sorry I know this is off topic, but I'm a recently diagnosed T1D and I would like to get more information about Loop and other loop-like open-source systems.

wombatpm 3y ago
My son is T1D since age 7. He is now on the OmniPod pump and the Dexcom G6 sensor. It’s been a very effective combination that’s covered by insurance. The challenge with the open source projects is limited equipment that can be hacked.
- jablongo 3y ago
  Most people on loop have omnipod and dexcom g6 also so I don’t think it’s that limited. It’s true that it takes a few months for new devices to get integrated but open source APS are the only systems lets you choose what devices you want to use with them.
Jarmsy 3y ago
One thing I wish I'd been told more about when first diagnosed was the possibility of prolonging the honeymoon period. It was decades ago, and back then the advice was to eat a fairly normal diet and just dose as much insulin as needed. If I was back in that position I'd go low carb high exercise from the start, and do everything I could to reduce my insulin requirements as much as possible. There are cases of new T1s coming of insulin completely, or just a low basal dose for multiple years.

thecosas 3y ago

Bummer to hear about beta bionics! My understanding was that the hardware they are releasing does have a place for glucagon, just not the software and additional reservoir required. I'm considering a switch to their pump since it seems like they've made it as automated as possible with insulin-only. Would love to see/hear more about what you're working on!

Loic 3y ago

My father was an MD (psychiatrist) and had this saying: - 1 drug, it works ok, still very patient specific. - 2 drugs at the same time, only if you validated correctly at least one alone before. - 3 at the same time: clusterfuck/total mess. Base rule: do everything possible to use only one active ingredient/drug at a time.

robobro 3y ago
Yes, much easier to control a single variable. I'm a type 1 diabetic on multiple daily injections -- a pump just introduces a lot of inconvenience for me, price aside. You have to track batteries and change tubing and make sure the tubing doesn't snag on things, and unplug the machine before taking a shower or bath, and ... blah blah blah... All I need for MDI is to carry my syringes and vial around in a cold pack (I live in the tropics, lol) and I don't need to worry about hardware beyond that.
- Jarmsy 3y ago
  I'm in the same boat re. MDI over pumping. I can still keep very tight control without a bunch of cumbersome attachments (I do wear a flash sensor though). Why vials and syringe rather than a pen though? I usually just carry the pen in my pocket and phone for monitoring. Even in hot weather I've not had problems with insulin losing potency noticeably before finishing a cartridge.