Abstract
P
The authors quantitatively compared the adaptive AP approach (R2R-AP) against nonadaptive AP (NA-AP), in 18 individuals with type 1 diabetes already using the latter system under free-living conditions. 11 The primary endpoint of this 4 weeks extended phase study was the proportion of time when sensor glucose was in the range 3.9–10.0 mmol/L for 24 h, during the last week of each intervention. The Diabetes Assistant (DiAs) AP system 12 was initialized using the participants' basal insulin pattern, CIR, and correction factor. Insulin delivery in the comparator arm was directed by a nonadaptive Modular MPC algorithm, which implements real-time corrections to the preset basal and meal bolus insulin delivery based on past sensor glucose data and projected glycemic level. 13 During the R2R-AP period, the nocturnal basal insulin and diurnal CIR were automatically adapted and updated by the controller. 14 The primary rule for adaptation, reduction of time spent hypoglycemic (<3.9 mmol/L), was put in place for safety. Once this was achieved, the secondary updating rule was then implemented to reduce time spent hyperglycemic (>10.0 mmol/L). Several adaptation functionalities and rules were imposed during the study. For example, the basal profile was only updated during the overnight period. However, this would not be implemented if meals were announced to the controller during this period or temporary basal insulin modifications occurred overnight. The CIR adaptation would be aborted if manual insulin delivery was instigated during the adaptation period, or if the R2R controller preset algorithm rules were not met (e.g., having multiple meals in the same CIR interval).
The primary endpoint was not significantly different between the R2R-AP and NA-AP approaches. The R2R-AP, however, had a significant impact during the overnight period. Both time in target and above target were improved by use of the adaptive controller system by 9.7 and 10.3 percentage points, respectively, compared with the nonadaptive controller. This perhaps was unsurprising, as others have similarly shown that the benefit of day-and-night closed loop use was more apparent during the overnight period. 5,15,16 Despite adaptive controllers, the barriers to daytime closed-loop performance are the relative delay of subcutaneous insulin action profile 17 and sensor glucose lag compounded by rapid glucose perturbations induced by daytime activities. 18 This may be mitigated by faster insulin 19 and better sensor performance in the future. The R2R-AP performance may also have been limited by the safety restraints imposed (time spent hypoglycemic was low and similar between the two treatments). It is hoped that further development of the R2R-AP approach, which is still in its early stages, can be further optimized while preserving safety aspects.
The technical performance of the R2R-AP controller also showed better outcomes overnight than the daytime period. Overnight basal rate adaptations were successfully completed more than 60% of the time during the whole study period. In comparison, daytime adaptation of the R2R-AP was lower, with only 23% of updates successfully completed. This was attributed to participants' eating and bolusing habits, as in more than a third of the cases additional meals or snacks were consumed within a 3-h period from the last meal. Interestingly, from a user-behavior perspective, the suggested meal bolus was overruled by 26% of participants. As user interaction can hamper controller adaptation during daytime, one could argue that minimal user input, for example, using a simplified bolus strategy, would help facilitate controller adaptation. This strategy has demonstrated satisfactory postprandial glycemic control in recent dual-hormone AP clinical studies. 20,21
There were several limitations that the authors rightfully acknowledged. Cautious interpretation of the study results is needed due to the small sample size and nonrandomized study design. Adults were exclusively studied, although the pediatric population may also benefit from adaptive controllers due to the challenging transition phase from childhood to adolescence. The adopted CIR adaptation assumes that daily carbohydrate intakes are generally consistent. Other factors known to attribute to glycemic variability, such as exercise and macronutrient meal composition, were not accounted for. The results also highlight the limitation of in silico experiments. There is no doubt that considerable advances have been achieved, thanks to data and results derived from in silico studies. 22,23 However, real-world studies during free-living conditions are still needed and irreplaceable especially for regulatory approval, due to the unpredictability and variability of human behavior and practice.
The human factor is now recognized as an important aspect when designing and developing AP systems, especially as hybrid AP systems are still dependent on user interaction. 24,25 Thus at present, there will remain an ongoing need for education and training of this novel technology, adopting individualized strategies and approaches when using adaptive AP systems during meals and exercise, for example. Adaptive fully closed-loop systems that are able to cope with meal-related excursion with minimal or no user input 20,26 may help reduce the burden of postprandial glucose management 27 and benefit those frequently missing meal boluses. 28 It is hoped that progress and advancements in AP controllers and insulin kinetics may lead to further reduction in diabetes burden and address the different needs of people with type 1 diabetes. 29
Footnotes
Author Disclosure Statement
No competing financial interests exist.