Welcome to the ATTD 2023 Interactive Program
Closing the loop – From cradle to mature age (ID 220)
IS026 - Closed-loop with adjunct therapies (ID 221)
Abstract
Abstract Body
Automated insulin delivery systems improve glycemia in type 1 diabetes but daytime control remains suboptimal and carbohydrate counting is still needed. Glucose control could be improved and carbohydrate counting burden could be reduced with the addition of adjunct therapies such as pramlintide, SGLT2i, and GLP-1. The amylin analogue pramlintide delays gastric emptying, suppresses nutrient-stimulated glucagon secretion, and increases satiety in people with type 1 diabetes. Adjunct use of pramlintide with closed-loop therapy improves glucose control during the day and has the potential to alleviate carbohydrate counting. SGLT2i inhibits glucose reabsorption in the kidney, which allows more glucose to be excreted in the urine and thus lowers blood glucose levels in an insulin-independent manner. Adjunct use of SGLT2i with closed-loop therapy improves glucose control during the day and night but increases ketone concentration and ketosis compared to placebo. Data on the adjunct use of GLP-1 with closed-loop therapy is lacking.Postprandial glucose control in advanced hybrid closed-loop systems (ID 222)
IS027 - Leveraging behavioral and physiologic patterns collected from wearable sensors and a smart-home to augment next-generation closed-loop algorithms (ID 223)
Abstract
Abstract Body
Wearable sensors and smart-home based sensors are becoming ubiquitous but they have not yet been integrated into automated insulin delivery (AID) or decision support systems (DSSs). We present a new algorithm called BlockRQA that is used to identify patterns from multi-variate, multi-modal data collected from both wearable sensors and smart-home sensors to identify behavioral patterns that can lead to negative health outcomes or other events important for glucose management. A total of 30 people with type 1 diabetes were recruited to be monitored for 4 weeks while wearing a CGM, an insulin pump, an Apple Smart Watch, and using a custom food and exercise tracking app while having their movement tracked with a beacon-based smart home monitoring system called MotioWear (MotioSens, Portland OR). Twenty-four participants had data that were usable for analysis of patterns. We found that BlockRQA was able to identify patterns that led to 61% of low glucose (<70 mg/dL) events on average. We found that meals could be anticipated 30-60 minutes in the future when utilizing movement data from the smart home along with other physiologic data within the BlockRQA algorithm whereby an average of 46.2% of meals could be anticipated. We derived a hypoglycemia risk score that is defined as a prior-conditioned ratio of likelihood of a pattern leading to low glucose (<70 mg/dL) relative to likelihood of a pattern leading to high glucose (>180 mg/dL). We also derived a meal anticipation score that is defined as a prior-conditioned ratio of likelihood of a pattern leading to a meal relative to likelihood of the pattern leading to low glucose. The hypoglycemia risk score and the meal anticipation score may ultimately be used to increase the aggressiveness of insulin delivery for an AID algorithm in anticipation of a meal, or may be used to decrease insulin aggressiveness in anticipation of a behavioral pattern that has led to a problem event like hypoglycemia.
Q&A (ID 755)