Final stats (N=1) of two weeks in (almost) full closed-loop on #AAPS #DynamicISF with Humalog. Two manual boluses were administered during the 14 days- one during a CGM warmup, and another when the loop disconnected due to a dying pump battery. 
Further, one (mild, exercise/alcohol-induced) low was treated with a pack of Swedish Fish. Comparison from #Dexcom #Clarity shows no significant change from baseline (prior 14 days) other than a reduction in SD of 19%. (Baseline on left, Full closed-loop on the right)
The 7% highs were almost completely due to high-calorie, high-carb, complex meals. (Bacon King and fries (chips) from Burger King, Pizza, Chicken McNuggets and fries from McDonalds). Notably, none of those resulted in a "rebound low".
Baseline was a commercially-available HCL system with lots of manual intervention, so this is a significant reduction in burden, with no significant change (and actually a slight improvement by some measures) in outcome.
For further comparison, the #nightscout data from the baseline period shows that even though the total number of above-range and below-range readings were similar in the two periods, the highs were not as high, and the lows not as low.
So, in summary, dynamically adjusting ISF appears to enable an automated system to perform at least as well as a human at managing #T1D. The next 2-wk period will examine a proposed refinement to the ISF = f(TDD, BG) function based on observations from other real-world testers.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
