Day-to-day variability of insulin requirements in the inpatient setting: observations during fully closed-loop insulin delivery

Objective To characterise variability of exogenous insulin requirements during fully closed-loop insulin delivery in hospitalised patients with type 2 diabetes (T2D) or new-onset hyperglycaemia, and to determine patient-related characteristics associated with higher variability of insulin requirements. We retrospectively analysed data from two fully closed-loop inpatient studies involving adults with T2D or new-onset hyperglycaemia requiring insulin therapy. The coefficient of variation quantified day-to-day variability of exogenous insulin requirements during up to 15 days using fully automated closed-loop insulin delivery. Data from 535 in participants were analysed. The coefficient of variation of day-to-day exogenous insulin requirements was 30±16% and was higher between nights than between any daytime (56±29% overnight compared with 41±21% in the morning 39±15% in the afternoon and 45±19% during the evening all p<0.01).


Introduction
Safe and effective management of diabetes and stress-related hyperglycaemia in hospitalised patients can be challenging due to the impact of metabolic responses to acute illness, inconsistent oral intake and use of nutritional support, scheduled or unscheduled fasting periods, and medications affecting insulin sensitivity, for example corticosteroids (1,2). Exogenous insulin requirements may vary considerably from day-to-day as a result of these factors. To date, characterisation of the variability of day-to-day insulin requirements in the inpatient setting has not been reported.
Automated closed-loop insulin delivery systems can be used as a tool to provide an estimate of exogenous insulin requirements. Closed-loop systems incorporate an algorithm to modulate insulin delivery in response to real-time sensor glucose levels, reflecting the amount of insulin required to achieve in-hospital treatment targets.
Fully closed-loop insulin delivery has been evaluated in inpatients with type 2 diabetes or new-onset hyperglycaemia in the non-critical care setting (3)(4)(5)(6).
Randomised controlled trials comparing closed-loop insulin delivery with usual care on the general wards have demonstrated superior glycaemic control without increasing the risk of hypoglycaemia, even in patients requiring enteral/parenteral nutrition and haemodialysis (3)(4)(5)(6).
In this retrospective analysis, we quantify the day-to-day variability of exogenous insulin delivery in adult inpatients with type 2 diabetes or new-onset hyperglycaemia during a period of up to 15 days of fully closed-loop insulin delivery (4,5). We compare patient-related characteristics between those with higher and lower variability of insulin requirements and relationship to glycaemic endpoints. This article is protected by copyright. All rights reserved.

Research Design and Methods
This retrospective post-hoc analysis evaluated closed-loop directed insulin delivery, as a marker of exogenous insulin requirements, from two multi-national randomised controlled trials (4,5).
Approvals were received from independent research ethics committees and national regulatory authorities in the UK and Switzerland prior to study start. All participants provided written informed consent. Inclusion criteria were adult inpatients on noncritical care wards (medical or surgical) with type 2 diabetes or new-onset hyperglycaemia requiring subcutaneous insulin therapy and for one study, an additional requirement for nutrition support (enteral/parenteral nutrition). Inpatients with type 1 diabetes were excluded. Only data from participants assigned to receive fully closed-loop insulin delivery were analysed in the present study. continuously for up to 15 days without any meal announcements or prandial insulin boluses. Participant's usual insulin therapy and/or sulphonylurea medication, if prescribed, was discontinued on the day of closed-loop initialisation. All other medications were continued. Standard insulin aspart (Novorapid, Novo Nordisk, Denmark) was used in one study (4), and fast-acting insulin aspart (Fiasp, Novo Nordisk, Denmark) was used in the other (5).

Data analysis and statistical methods
This article is protected by copyright. All rights reserved. This article is protected by copyright. All rights reserved.
Of those participants included, 31.3% received enteral/parenteral nutrition, 13.4% haemodialysis, and 13.4% corticosteroid therapy during the study period.  Post-hoc test comparing high and low CV groups demonstrated that inpatients in the highest tertile of insulin variability were younger than those in the lowest tertile This article is protected by copyright. All rights reserved.
(65±10 vs 71±11year; p=0.035). Body mass index (BMI), gender, HbA1c, diabetes and insulin duration, use of steroids, requirement for dialysis or nutrition support were comparable between the high and low insulin variability groups (Table 1).
Participants with high variability of day-to-day insulin requirements had comparable This article is protected by copyright. All rights reserved.

Conclusions
This analysis reports considerable variability of day-to-day exogenous insulin requirements during use of a fully automated closed-loop insulin delivery system in inpatients with type 2 diabetes or new-onset hyperglycaemia.
We observed higher variability of insulin requirements between night-time periods (CV of 56%) compared to between daytime periods (CV of 39-45%), in the context of lower insulin requirements overnight. Identifying higher risk periods, where increased attention needs to be given to glucose management, is important to prevent adverse glycaemic events in inpatients. The variability of overnight exogenous insulin requirements in people with type 2 diabetes or new-onset hyperglycaemia in the inpatient setting in our study is even greater than the variability of overnight insulin requirements reported in adults with type 1 diabetes, 56% compared with 31% and 36%, (7,8).
These results enhance our understanding of why attainment of recommended glucose targets during the hospital admission is challenging. The workload associated with regularly adjusting insulin doses to meet treatment goals is a significant burden in the inpatient setting. Inpatient dysglycaemia is a poor prognostic marker, associated with increased morbidity and mortality, length of stay, and healthcare costs (2). Our observations may help to further understand why, despite frequent capillary blood glucose monitoring and regular insulin dose adjustments, dysglycaemia is common in people with type 2 diabetes and new-onset hyperglycaemia during the hospital admission (1).
High variability of insulin requirements was associated with lower participant age in our analysis. We hypothesise that this may reflect greater caloric intake in younger This article is protected by copyright. All rights reserved.
inpatients (9) although other reasons may apply. No other demographic factors significantly influenced variability of insulin requirements in our analysis.
The high day-to-day variability of insulin requirements is difficult to overcome with conventional therapeutic tools, multiple daily injections and insulin pumps. Therefore, our results emphasise the importance of advanced technologies such as closed-loop systems to safely and effectively manage inpatient diabetes. The advantage of automated, algorithm-directed insulin delivery systems is the frequent modulation of insulin delivery according to real-time sensor glucose concentrations, thereby accommodating variability of insulin delivery to achieve glycaemic consistency. We have shown in this analysis that fully closed-loop insulin delivery systems can accommodate highly variable day-to-day insulin requirements without compromising glucose control or increasing the risk of hypoglycaemia.
The strengths of our investigations include the heterogeneity of participants included and the multinational study design, which supports generalisability of our findings.
Limitations include minor differences in study design that were not controlled for, and a relatively short follow-up period. We did not evaluate the impact of individual noninsulin glucose-lowering therapies. The study was not powered to assess the impact of individual factors (dialysis, nutrition support, steroid therapy) on variability of insulin requirements.
In summary, there is high day-to-day variability of exogenous insulin requirements in the inpatient population, particularly overnight. Diabetes management approaches should account for this variability, and consider adoption of closed-loop systems in the inpatient setting. This article is protected by copyright. All rights reserved. Table 1. Baseline demographics and glycaemic outcomes between different tertiles of variability of exogenous insulin requirements during fully automated closed-loop.