JDBT Journal

Introduction

The importance of achieving adequate glycemic to prevent longterm complications has been clearly established in people with T1D [1]. However, achieving these recommended glycemic targets remains challenging. This is highlighted in a recently published data from the T1D Exchange consortium demonstrating that only 21% of adults achieve the desired HbA1c < 7%, without significant hypoglycemia [2], despite a widespread use of continuous glucose monitors. The inability to achieve optimal glycemic in diabetes is multifaceted, with challenging post meal glucose management, being a key contributing factor [3]. The complexity of meal management is related to the complexity of nutritional constituents’ absorption, the interactive effect of meal components on the overall glycemic effect, poor reproducibility of meal absorption all leads to unpredictability of post meal glucose excursion.  Physiological, 90% of dietary carbohydrates are absorbed within 1–2 hours after eating [4], the attempt to mimic this physiological response to meals with subcutaneous insulin injections using currently available insulins and delivery systems is sub-optimal causing disturbing and unpredictable post meal excursions and hypoglycemia. 

Dietary fat and protein effect postprandial glycemia in patients with T1DM [5,6] by the attenuation of blood glucose elevation in the early postprandial phase (2-3 hours post-meal) with delayed peak and prolonged glycemic excursion later in the postprandial phase (>3 hours post- meal) [7]. Possible explanation for effects includes:

1. Transient insulin resistance attributed to dietary fat and Free Fatty Acids (FFAs) and altered response of other hormones involved in glucose regulation including glucagon, Glucagon Like Peptide 1(GLP1), ghrelin and Glucose-Dependent Insulinotropic Peptide (GIP) [8].
2. High dietary fat-mediated delayed gastric emptying leading to late post-prandial glycemic response [6].
3. Directly conversion of amino acids to glucose

(gluconeogenesis) increases blood glucose levels [8].

CHO counting with early, pre-meal insulin injections is one of the recommendations to mitigate some of these variabilities [9]. Indeed, the Diabetes Control and Complications Trial (DCCT) found that CHO counting is effective in meeting outcome goals and allows flexibility in food choices [1]. In addition, different conversion algorithms for fat and protein were also described.  For example, the Pankowska Equation and Food Insulin index has demonstrated a reduction in postprandial glucose [10]. However, these measures of CHO counting and conversions, and meal bolus timing emerged as a major source of burden to PWD.   Patients related factors such as fear of hypoglycemia also challenges meal glycemia [11]. Injecting before a meal, when the glucose levels might be normal, frighten some patients that are not certain if and what meal they will eventually consume, leading to   late bolusing in relation to food consumption, and/or intentional underestimation of CHO intake, skipping meal bolus, leading to a mismatch between meal absorption and exogenous insulin action [12].

Beyond addressing the nutritional content of meals, different strategies of meal blousing were devised to address the delayed gastric emptying, such as splitting the meal bolus, in various ways, [13] adding to the complexity meal handling.

 AID systems, such as the MM780G system, combine a closedloop algorithm controller with continuous subcutaneous insulin infusion and continuous glucose monitor (CGM) to provide automated basal insulin delivery and an automated correction every five minutes. Meal   boluses follow CHO announcement that effectively targets postprandial hyperglycemia by mitigating different factors that might affect the post meal glucose glycemia [14]. The concepts governing the development of the MM780G algorithm were reducing patient burden while   significantly increasing the efficacy of the therapy, this included announcing only carb content of meals, with a single premeal bolus while the algorithm-driven controller autonomously manage all other parameters associated with meal variability.  (Grossman B. et al manuscript accepted for publication DT&T). 

This study was aimed at providing proof of the above algorithm design concepts by comparing the management of HPHF meals by the most elaborate non algorithmic driven therapy versus the use     of MM780G system    either in its intended use mode with accurate CHO counting or with a modified and simplified approach that precludes CHO counting. Identifying simplified approaches for complex meal management will aid in reducing the burden associated with meal management, while maintaining effectiveness and safety of insulin delivery [14].

Research Design and Methods

Methods: This was an open label, single-center, repeatedmeasures prospective meal trial.

Participants were included if they were 20 to 70 years of age, had body weight of 45-120 kg, had a clinical diagnosis of T1D for a minimum of 36 months prior to enrollment, used the MiniMed™ 780G insulin pump with real time continuous glucose monitoring (Medtronic GS3) for at least 6 months with the use of a rapidacting insulin analogue, a minimum daily insulin requirement (total daily dose; TDD) equals to or greater than 8 units and had an HbA1c ≤ 10.0% measured at screening visit. Participants were excluded if they had unresolved adverse skin condition, a history of hypoglycemic seizure or hypoglycemic coma within the past 12 months, a history of seizure disorder unrelated to diabetes within the past 12 months, had any condition, including screening lab values that in the opinion of the investigator may preclude them from participating in the study and completing study related procedures, and pregnant or lactating women. All participants provided written informed consent and were willing to comply with all study procedures and were proficient in the English language. Trial was registered at clinicaltrials.gov (ClinicalTrials. gov Identifier: NCT04901143).

Study protocol

Run-in period: Subjects using AID, who met inclusion and exclusion criteria, were enrolled after providing a written informed consent. Demographics, anthropometric measures, medical history, and blood samples for HbA1c were collected. Subjects were trained on meal challenges procedures, and optimal pump settings were verified (glucose target=100 mg/dL and active insulin time=2 hours).  During this period overnight OL settings were optimized by requiring the subjects to spend one night on OL after consuming a low fat, low protein, and low fiber meal with a regular bolus and setting the OL setting, to meet optimal glucose targets of 70-180 mg/mL.

Study period: Identical  test meals  were consumed during three treatment phases; Open loop (OL), Closed Loop (CL) and simplified CL with a predefined bolus (CLP), as following: (1) consume the meal test on the same day  at supper; (2) no food consumption within 4 hours prior to, and during at least 6 h after the test meal (overnight); (4) consume the test meal 10 min after delivery of the insulin bolus; (5) avoid exercise for 24 h before the test meal and for 12 h following the meal. Participants were contacted daily by a member of the research team to ensure the protocol was being followed correctly. In addition, daily study diary was filled. Each meal test was repeated 3 times under each treatment phase for a total of 9 meals. Figure 1 depicts the study design.

Figure 1: Study overview.

Test Meals:

A standardized HPHF test meal of pre-made frozen pizza was provided to subjects by the study team at the beginning of the study. The meal contained 41 g fat, 38 g protein and 60 g carbohydrates, with a total energy content of 761 kCal. The macronutrient content of the meal tests was analyzed by a certified chemical laboratory (Bactochem Ltd. Israel)

 In the OL phase, insulin delivery was used with the PLGM feature switched on, meal bolus dose   was based on common practice and recommendations  (10), i.e. Participants were instructed to enter the accurate amount of CHO (60g) through  “bolus wizard” feature, to increase the recommended amount of insulin units by 30% to account for the protein content of the meal and to use a combination bolus (dual wave bolus) with a 50/50% split over 2.5 h. Insulin was delivered 10 minutes prior to consuming the meal. The participants were required not to give a correction bolus within 3 hours post meal unless glucose level meet the pre-defined safety criteria as described below. Following 3 hours of meal initiation and throughout the night, participants could correct high glucose level   by using the “Bolus Wizard” feature.

In the CL phase, bolus was given with an accurate CHO content of the meal (60 g) entered to the “Bolus Wizard” calculator. Mealtime insulin was delivered 10 minutes prior to consuming the meal. Participants were instructed not to interact with the insulin delivery during the hours post meal and throughout the night.

In the CLP phase, a predefined carbohydrate meal estimation was delivered 10 minutes prior to each meal. This estimation was individually calculated for each participant by first determining their fixed insulin dose, which was derived by dividing their TDD by 6 (representing 50% of TDD allocated to bolus insulin, further divided by 3 main meals). This insulin dose was then converted to a carbohydrate value by multiplying by an insulin-to-carb ratio of 7. For safety purposes, the calculated carbohydrate value was capped at 60g to prevent potential excessive insulin delivery and minimize the risk of hypoglycemia.

 The universal meal dose per participant is presented in Table 1). As in the CL phase, participants were instructed not to interact with the insulin delivery during the hours post meal and throughout the night.

Table 1: Optimal priming dose.

Safety measures:

Hypoglycemia- In the event of grade 2 hypoglycemia, defined as blood glucose level < 54 mg/mL confirmed by a glucose meter, subjects were instructed  to consume approximately 15 grams of carbohydrates and if self-monitoring of blood glucose shows continued hypoglycemia fifteen minutes following treatment, CHO consumption should be repeated.

Hyperglycemia- In the event of hyperglycemia, defined as glucose values that remain above 250 mg/dL (13.9 mmol/L) for more than 3 hours, a correction bolus was given using the “Bolus Wizard” feature recommendation. Blood beta hydroxybutyrate concentration was measured using a handheld meter and if blood ketones > 3.8mmole/L the subjects were to be treated as if there is a concern of diabetic keto-acidosis per investigator discretion.

Statistical Analysis

The mean ± standard deviation is reported for continuous variables. Outcomes were calculated with IBM SPSS Statistics, using the paired-sample t-test. Statistical significance was defined as p <0.05.  Endpoints were the average of the 3 repeated meals per subject and compared between OL and CL, OL and CLP and CL and CLP. The primary outcome measure was mean postprandial glucose excursion from baseline to 5.0 h post-meal using the 5.0 h area under the glucose curve (AU glucose)  Other  endpoints included percent of SG values in various glucose ranges (i.e., 180 mg/dL, and >250 mg/dL), insulin delivered (i.e., total meal dose, and precent of. auto correction and auto basal of total meal dose).

Sample size: Sample size was determined using WINPEPI sample size calculator program based on the following assumptions: 1) Primary outcome: AUC glucose 300 minutes post meal mean difference on open loop therapy using standard of care bolus vs. AID system. 2) Comparison will be conducted using the paired-T statistic. 3) In a previous study of 19 individuals with type 1 diabetes an AUC glucose 300 minutes of 51.2 (mmol/l/min) with a SE of 5.3(mmol/l.) was found [15].  SD =SE (5.3) * √n(√19)=23.1022. Therefore, a sample size of 12 is estimated to provide 80% power at 5% significant level, to detect a potential mean difference in glucose AUC of at least 20 (mmol/l/min).

Results

Participant characteristics

A total of 15 participants were recruited, of whom 13 (11 men) completed the protocol. Two participants withdrew consent and data was excluded due to failure to complete the protocol. Incomplete data due to lost sensor signal or other technical issues were excluded. Baseline characteristics are presented in Table 2. Participants had a mean age of 46.8 ±12.8 years, a diabetes duration of 20.1± 12.3 years, HbA1c of 6.7±1.2% and a BMI of

25.5 ±3.6 Kg/m². The mean duration of insulin pump therapy was 12.8 ±7.0 years. Figures 2&2b demonstrate the mean postprandial glucose for 5.0 h and for the overnight period after the meal tests (0 to 10 hours postprandial).

Table 2: Summary of clinical participant characteristics.

Figure 2: Postprandial glucose profiles (mg/dL) from 0-300 min for 3 different meal handling approaches for a HFHP meals in adults with T1DM.

Figure 2b:  Overnight postprandial glucose profiles (mg/dL) for 3 different meal handling approaches for a HFHP meals in adults with T1DM.

OL Versus CL

No significant differences in the 5.0 h AUC glucose were observed between the groups (6801.1.3±7750.6 mg/dL/min, and 7890.9.6±7210.8 mg/dL/min for OL and CL, respectively; p=0.435). When using OL, there was a tendency for increased time in hypoglycemia (defined as SG <70 mg/dL, % time) at 5.0 h (10±15.3 % vs.4.7±6.2 %; for OL vs, CL, respectively, p=0.098). A similar post-prandial insulin delivery at 2.0 h post-meal was observed between groups (8.1±4.3 units vs.7.4±2.5 units, P=0.15), while significantly higher amount of insulin was delivered using CL between 2.0 to 5.0 h post-meal (2.5±1.5 units vs. 4.8±4.3 units; p=0.015). Time in target range (70-180 mg/dL, %) was comparable for OL and CL (76.2±22.7 % vs.83.6±20 %, p=0.26). (Tables 3)

When using CL, participants spent more time in the desired glucose range for the entire 10 hours post prandial, as reflected by a higher TIR (70-180 mg/dL, %) compared with OL (85.5±16.3% vs. 64.6±23.5 %; p=0.004), and with better mean SG (128.5±20.9 vs. 143.7±35.2 mg/dL; p=0.05). Delivering insulin using OL increased time in hyperglycemia and in extreme hyperglycemia (defined as SG>180 mg/dL and SG>250 mg/dL, respectively). In addition, incidence of hypoglycemia during 10 hours postprandial was increased using OL compared to CL ((6.9±7.5 % vs.3.3±4 %;p=0.029) while insulin delivery was similar (14±6.7 units vs.16.2±9.6 units for OL vs. CL, respectively; p=0.161). Moreover, the standard deviation of glucose and the coefficient of variation were significantly lower with CL as compared with OL (p=0.0095 and p=0.0235; respectively). (Table 4, Figure 3)

Table 3:  5.0-hr post prandial Glycemic outcomes following a high protein and a high fat meal.

Table 4- 10.0-hr post prandial Glycemic outcomes following a high protein and a high fat meal.