Algorithmen für die automatisierte Insulindosierung (AID): ein Überblick

Abstract

For automated insulin dosing (AID), an insulin pump with appropriate accessories and a continuous glucose monitoring (CGM) system are needed on the hardware side. Both devices must communicate with each other, either directly or via a smartphone with a corresponding app. To implement automatic insulin delivery, an algorithm integrated into the software is also required to ensure appropriate “translation” of the glucose data so that the correct insulin dosage is delivered at a given time to keep the glucose level in the body within the target range. This software can either be integrated into the insulin pump (e.g., Medtronic MiniMed™ 670G/770G/780G or Tandem t:slim X2 Control IQ™) or as an app on a smartphone connected to the insulin pump and CGM system. It may seem easy to calculate the correct insulin dose and control the pump to deliver the currently required insulin dose; however, after administration into the subcutaneous adipose tissue, insulin is absorbed rather slowly and only then becomes subsequently metabolically active. The specifics of this dynamic process must be taken into account by the software. One goal for the AID system is that it should maximize the proportion of time (at least 70%) that glucose levels are in the target range 70–180 mg/dl (3.9–10.0 mmol/l), while simultaneously minimizing the proportion of time (at most 4% <70 mg/dl and 1% <54 mg/dl) in the hypoglycemic range <70 mg/dl (3.9 mmol/l) or <54 mg/dl (3.0 mmol/l). In addition, AID systems should be safe to use, i.e., in particular, no severe hypoglycemia should occur. In this review article, the different types of AID algorithms that have been developed in recent decades are described and classified using a systematic approach. A systematic evaluation of the performance of the different algorithms in the sense described is lacking not only in the clinical development of AID systems, but also in daily practice. It is likely that other factors that are important in daily practice will decide which AID algorithms will be used in the respective AID systems and which will succeed on the market.