New technologies for MDI. Have Roche and Medtronic missed a trick?

At ATTD, we saw the Medtronic Experience (which sounds like a bad rock tribute band) and Roche’s SmartGuide describing themselves as the next great thing in the MDI world.

The way Medtronic want you to think about InPen is as the greatest option in the world of MDI.

Roche, on the other hand, want you to think about SmartGuide as reaching the parts that other CGMs fail to reach with their longer term predictive alerts.

There’s no doubt that these both make sense, as these types of technology go, but the implementation feels like both have missed the target. Let’s go into why.

The InPen

The InPen was bought by Medtronic and integrated into their platform. It’s sold with the idea that it alerts you when you’ve missed dosing for meals by using the glucose data from Simplera (or Guardian G4 I presume) to try and maintain better time in range.

The idea makes a lot of sense, and provides actionable suggestions for users (perhaps explaining why the Simplera app seems less capable of this). But yet again, either there’s an opportunity missed, or a deliberate decision been taken to shorten the time to market by missing features that would be true game changes to MDI users.

Internet Requirement

We’ll come back to this with Roche, but in a world where you can run a machine learning based Automated Insulin Delivery system on your phone, why does an app to give CGM based dosing guidance require an internet connection?

It feels like a complete misunderstanding of users and when some of these features might be most useful. On a long distance flight? Then factor in paying for internet as part of your ability to do normal diabetes care? Going camping up country? Better check you have that internet connectivity if you want any guidance…

This is a missed opportunity.

Settings

Given that Medtronic have been selling the 780G now for a number of years, it seems astonishing that the therapy settings menu is manual, and according to the app, should be checked with a healthcare professional before making changes. Especially when you consider the number of times per year you see a healthcare professional and the frequency with which you need to make changes (at a minimum monthly if you’re a woman).

There are no options to create profiles that might be easily switchable, and given what the 780G is capable of on pump, it seems like a hugely missed opportunity where machine learning could have been used to identify when various settings might need updating, taking a huge amount of burden off users.

Oh well. At least you can choose an active insulin display…

I’d like to think that there are some major upgrades in the pipeline that address some of these concerns, but of course they probably require clinical trials and cost, and who knows whether that might take place.

Roche SmartGuide

The big industry symposium of the event was Roche’s entry into the CGM market, with SmartGuide (much of which was predicted here).

As their slides showed, the key feature is long term predictions to reduce worry about hypos longer term (compared with 30 minutes predictions from most systems).

While this seems like an applaudable solution, and makes sense for a group of users on MDI or non-automated pump solutions (which to be fair, is the majority of people using insulin), it comes with a couple of caveats.

Internet Connection

Once again, you need an internet connection for the longer term predictions to function. On that flight, or camping in the middle of nowhere when they might be most useful? Sorry, no predictions. The models run on the cloud and while we understand that not all phones can handle some of these types of heavy lifting, this functionality only working when online misses the target.

Calibration at startup

Surprisingly, this CGM needs a fingerprick at start up, which puts it about 6 years behind Dexcom and 10 behind Abbott.

Requiring an initial calibration suggests that there’s variability across the sensor manufacturing process that Roche haven’t yet got on top of, which is hardly a surprise, given this is their first commercial offering in this market.

But unless the price is incredibly low, or you have very specific needs, this feels like a missed opportunity.

Some thoughts on these technologies

While these offerings from Roche and Medtronic are both good starts, they could have been so much more.

The inability to offer on-phone capabilities into the systems is a massive hole in both offerings, especially as the key selling points of their products rely on a network connection. In the case of the InPen it’s particularly crazy as the CGM data is already on the phone.

Given that both of these solutions use machine learning to do what they do, it seems particularly odd that the writers of the code behind them haven’t optimised it to run locally on the phone, when commercial machine learning models for AID already do this. And given this opportunity for the InPen to provide so much more insight, especially if the ML has been taken off the phone and put in the cloud, was quite a surprise.

As the title suggests, while these may well be steps forward for MDI users, they could have been so much more.