14-day comparisons! Photo 1 (2/27-3/12) is the baseline. #2 (3/13-3/26) is full #closedloop with Humalog and #DynamicISF. #3 (4/20-5/03) is full closed loop w/a 50/50 Humalog/Lyumjev mix. Best results with faster insulin. TIR gains are minimal (starting from a high baseline). 1/? 
What jumps out at me are the reductions in variability. Also, the Lyumjev/Humalog results probably would have been even higher TIR and less variable had the restaurant not messed up last night, delivering basically a whole chicken rather than the 2 chicken burritos I ordered. 2/?
Common conditions in full #closedloop- basically no manual boluses. Even if a meal sent me close to 300 (16.7). Manual boluses were only applied when the loop was not running (comms failures, #CGM warmup, etc) maybe 2/week. No carb counting at all. Ate when hungry. 3/?
The variation of the model w/r/t the insulin in use is actually important- it explains why this effect was only seen minimally in clamp studies- the pharmacokinetics matter. The earlier the peak, the higher it is. Relative concentrations at peaks of insulin and glucose seem 4/?
to be a driving factor. This is also likely contributing at least some of the effect: reduced INSR expression in response to insulin exposure. (faseb.onlinelibrary.wiley.com/doi/full/10.10…) In any event, with a mechanism in place to scale ISF/CorrF relative to insulin exposure and glucose level 5/?
...appears to enable better dosing, reduce glycemic variability, and improve TIR- at least in this N=1. Improvements in sensor accuracy will likely help in the future, as will improvements in insulin PK. Obviously, there are opportunities to build better algorithms around 6/?
#DynamicISF. In fact, that's already started. This is just basically dropping the core dynamic function into an existing algorithm (oref1 - SMB/UAM). Better ML/AI wrapped around it will likely enable better adjustments for circadian patterns in insulin sensitivity and longer- 7/?
duration cycles (e.g. monthly ones 😉). Blanc, et al (including folks from @Diabeloop noted this effect in their DTM Poster "INSULIN SENSITIVITY, A FUNCTION OF BLOOD GLUCOSE LEVEL: PROOF OF CONCEPT ON SP7 MULTI-CENTRIC CLINICAL STUDY." from 2018. 8/?
With that, it's time to plan for some wider data collection. #WeAreNotWaiting 9/9
• • •
Missing some Tweet in this thread? You can try to
force a refresh
