A Glimpse of the Future – local Artificial Pancreas study
After hearing about how promising artificial pancreas trials have been, I was very excited to hear a talk here in Seattle by Dr. Mauseth, a Puget Sound pediatric endocrinologist, about results from his trial (with Dr. Carla Greenbaum) using fuzzy logic as the brain behind the dosing decisions. He was turned on to fuzzy logic by a couple of Boeing engineers who instantly saw a parallel between fuzzy logic used to help plans with takeoffs and landings and its potential to help control blood sugar ups and downs. (I’m always amazed at how quickly engineers – whether software or other – grasp the challenges of managing type 1.) Fuzzy logic is promising, because rather than trying to create an absolutely correct, predictable mapping of an individual’s blood sugar variations (which is like trying to predict the weather with complete precision), it treats everything as a matter of degree, as part of a spectrum. Rather than relying on an absolutely precise blood sugar reading from a CGM (nowhere near close yet!), it looks at blood sugar trends and steepness: if blood sugar is falling quickly, fuzzy logic tells the pump to ease up on or suspend insulin, etc. This isn’t too different from what we do now with our CGM – if it’s double arrows straight down, even at 200, we start pulling out the graham crackers.
I’d assumed that these artificial pancreas trials must be using secret Bat Cave prototypes to get results like Dr. Mauseth’s – blood sugar staying in a 100-14o range through the night, spikes in the low 200s after a carb-heavy meal, etc. But no – Dr. Mauseth used an Omnipod and Dexcom 7+ (because they were compatible with the software platform), with the only modification being that the Omnipod was set to give a basal bolus every 5 minutes (like a real pancreas), rather than every half hour like current pumps.
I was floored that they could use the current generation of Dex, which can often be off 100 points off at times. I could understand fuzzy logic compensating for some of that, but surely they must’ve been calibrating the Dex every half hour or so…?
No, was the answer – “We calibrate every 6 hours or if sensor is more than 20% variance from a venous blood sample, which is done every 15 minutes. We have only had to calibrate out of cycle once, and that is when our original calibration was done when the blood sugar was changing.” They did have some glitches with the Dex dropping out of range when a patient went to the bathroom and found it less accurate when patients were sleeping rolled over on the sensor site. All lines up with our Dex experiences as well.
And when could we get that lovely bit of interpretive software, and would it take up an entire harddrive?
It would fit on an iPod, Dr. Mauseth said.
He has only done 3 of 10 (proof of concept) trials – if Luke were over 18, I’d sign him up. I’m sure there’s no shortage of volunteers, but FYI, here’s the site: http://www.benaroyaresearch.org/diabetes-research/find-study/fuzzy-logic
There are other intriguing trials going on with JDRF funding, including one using both glucagon and insulin. It’s not a cure, but after 6 months of essentially napping through the night due to ever-shifting toddler blood sugars, it would be a very welcome leap forward!