Insulin Bolus Administration in Insulin Pump Therapy: Effect of Bolus Delivery Speed on Insulin Absorption from Subcutaneous Tissue

Abstract

Background:

This study assessed subcutaneous absorption kinetics of rapid-acting insulin administered as a bolus using bolus delivery speeds commonly employed in commercially available insulin pumps (i.e., 2 and 40 s for delivering 1 insulin unit).

Materials and Methods:

Twenty C-peptide-negative type 1 diabetic subjects were studied on two occasions, separated by at least 7 days, using the euglycemic clamp procedure. After an overnight fast, subjects were given, in random order, a subcutaneous insulin bolus (15 U of insulin lispro, Eli Lilly) either for 30 s using an Animas IR2020 pump (fast bolus delivery) or for 10 min using a Medtronic Minimed Paradigm 512 pump (slow bolus delivery).

Results:

Fast bolus delivery resulted in an earlier onset of insulin action as compared with slow bolus delivery (21.0 ± 2.5 vs. 34.3 ± 2.7 min; P < 0.002). Furthermore, time to reach maximum insulin effect was found to be 27 min earlier with fast bolus delivery as compared with slow bolus delivery (98 ± 11 vs. 125 ± 16 min; P < 0.005). In addition, the area under the plasma insulin curve from 0 to 60 min for fast bolus delivery was greater than the one for slow bolus delivery (10,307 ± 1291 vs. 8192 ± 865 min·pmol/L; P = 0.027).

Conclusions:

Results suggest that insulin bolus delivery with fast delivery speed may result in more rapid insulin absorption and, thus, may provide a better control of meal-related glucose excursions than that obtained with bolus delivery using slow delivery speeds. Our findings may have important implications for the future design of the bolus delivery unit of insulin pumps.

Introduction

Continuous subcutaneous insulin infusion (CSII) using an external insulin pump has become a popular treatment option for patients with type 1 diabetes.^1,2 Modern insulin pumps for CSII are small programmable battery-operated devices^1
–3 that allow insulin to be delivered at a high rate for meals (bolus insulin delivery) and at a low rate between meals and during night (basal insulin delivery). The bolus insulin is usually administered for a short period of time. The length of this period (bolus duration) depends on both the bolus size chosen by the patient and the insulin pump model used. For example, pump models manufactured by Animas deliver 1 U of insulin for 2 s.³ Thus, when a typical bolus amount of 10–20 U of insulin is administered through such a pump, the bolus duration will be <1 min. Pump models featuring longer bolus durations (e.g., 1 U of insulin for a period of 40 s) are manufactured by Medtronic Minimed and Insulet.³ Thus, with such pumps, a typical bolus amount of 10–20 U of insulin is administered for a period of 7–14 min. Reasons for delivering boluses for longer periods of time may include power consumption limitations, considerations on flow rate accuracy, and the prevention of burning sensation at the infusion site.

It has been shown previously that the absorption of insulin administered as a subcutaneous infusion for 30 min is much slower than the absorption of the same amount of insulin administered as an infusion for a period of 5 min.⁴ Thus, a potential drawback of a lower bolus delivery speed may be slower insulin absorption from the subcutaneous infusion site. In this previous study, the slower insulin absorption associated with lower bolus delivery speeds was observed in healthy humans for porcine insulin. To the best of our knowledge, evaluations of the absorption kinetics associated with rapid-acting insulin administered with bolus delivery speeds typically employed in commercially available insulin pumps have not been carried out. Therefore, the purpose of this study was to evaluate the pharmacodynamics (PD) and pharmacokinetics (PK) of rapid-acting insulin administered as subcutaneous boluses with bolus delivery speeds commonly employed in current insulin pumps.

Materials and Methods

Study subjects

Twenty subjects were included in the study. They were men and women required to be in the age group of 18–60 years and diagnosed with type 1 diabetes. They had to have a HbA_1C of <10% (<86 mmol/mol), have a body mass index (BMI) between 20 and 30 kg/m², and be treated with CSII or multiple daily injections of insulin (MDI). Subjects were excluded if they had evidence of clinically overt diabetic complications, had local lipodystrophy at insulin injection sites, had C-peptide levels in blood plasma >30 pmol/L, and had used insulin lispro before the study. Each subject signed a written consent form after the purpose, nature, and potential risks of the study had been explained. The study was approved by the Ethics Committee of the Medical University of Graz and the Austrian Agency for Health and Food Safety (Clinical Trials registration no. NCT01792323).

Study design

Eligible subjects were studied on two occasions, separated by 5–21 days, using the euglycemic clamp procedure.⁵ On each occasion, subjects were admitted to the clinical research center in the morning after an overnight fast. Subjects treated with MDI administered their last dose of rapid- or short-acting insulin at least 5 h before the study visit. Subjects using long-acting insulin analogs were transferred to Neutral Protamine Hagedorn (NPH) insulin 2 days before the study visits. The last dose of NPH insulin was administered at least 20 h before the study visit. Subjects with CSII treatment stopped their insulin delivery at least 5 h before the study visit. At ∼7:00, an arm vein was cannulated with an 18-gauge catheter to be used for insulin and glucose infusion. A second catheter was inserted into a vein in the opposite forearm for blood sampling. The forearm with the sampling catheter was then placed under a heating blanket (55°C) to ensure arterialization of the venous blood. After catheter insertion, an intravenous infusion of regular human insulin (Actrapid; Novo Nordisk, Bagsvaerd, Denmark) was started using a syringe pump (Perfusor fm; B. Braun, Melsungen, Germany). During a subsequent baseline period lasting 4–6 h, the insulin infusion rate was adjusted on the basis of frequent plasma glucose measurements to slowly achieve and maintain target glucose levels between 4.4 and 6.7 mmol/L (80–120 mg/dL).

Once the plasma glucose concentration had been maintained in the target range for at least 40 min, a soft 9-mm cannula of an infusion set (Quick-set; Medtronic Minimed, Northridge, CA) was placed in the abdominal subcutaneous adipose tissue. Twenty minutes later, the intravenous insulin infusion was decreased at 5-min intervals to 75% and 50% of the basal infusion rate that maintained target glucose levels before the subcutaneous cannula insertion. Thirty minutes after subcutaneous cannula insertion, the intravenous insulin infusion was discontinued and a subcutaneous bolus of 15 U of insulin lispro (Eli Lilly Nederland BV, RA Houten, Netherlands) was administered, in random order, either for 30 s (fast bolus) using an Animas IR2020 pump (Animas Corp., West Chester, PA) or for 10 min (slow bolus) using a Medtronic Minimed Paradigm 512 pump (Medtronic Minimed). When the plasma glucose concentration decreased by at least 0.28 mmol/L (5 mg/dL) from the baseline value (defined as the plasma glucose value determined 10 min before the insulin bolus administration), an intravenous infusion of a 10% glucose solution (Glucosteril 10%; Fresenius Kabi, Bad Homburg, Germany) was started using a peristaltic pump (Infusomat fmS; B. Braun). The intravenous glucose infusion was then periodically adjusted to maintain the plasma glucose concentration at the baseline level for the next 8 h.

Blood samples for the measurement of the plasma glucose concentration were drawn every 5–15 min throughout the experiment. Additional blood samples for the insulin determination were collected at 15 and 1 min before the insulin bolus administration, and at 5, 10, 15, 20, 25, 30, 40, 50, 60, 70, 80, 90, 105, 120, 150, 180, 240, 300, 360, 420, and 480 min afterward. Before the placement of the subcutaneous infusion cannula, the reservoir of the insulin pump (either an Animas 2.0 mL Cartridge or a Medtronic Paradigm Reservoir 1.8 mL) was filled with about 0.75 mL ( = 75 U) of insulin lispro and attached to an infusion set (600-mm tube length, Quick-set; Medtronic Minimed). To facilitate the attachment of the infusion set to both reservoir types, infusion sets with Luer Lock connectors and connectors specific to the Medtronic pump were used. After inserting the reservoir into the insulin pump, the tube and cannula of the infusion set were filled with insulin. To minimize differences in the rate of insulin absorption from different areas within the abdominal wall (e.g., absorption may be faster from the upper than the lower abdomen⁶), the cannula placement site selected on the second study day was in proximity to the placement site chosen on the first study day (25–50 mm apart). After the administration of the insulin bolus, the infusion set tube and insulin pump were disconnected from the cannula. The cannula was left in place throughout the experiment.

Analytical procedures

Plasma glucose concentrations were measured at the bedside using a glucose oxidase method (Super GL 2; Dr. Müller Gerätebau GmbH, Freital, Germany) with a coefficient of variation (CV) of <2%. Plasma concentrations of insulin lispro were determined as described previously,⁷ using two two-site enzyme immunoassays (Iso Insulin ELISA, Insulin ELISA; Mercodia AB, Uppsala, Sweden) with within-assay CVs of <6%. The plasma C-peptide concentrations were determined by a two-site sandwich chemiluminescent immunoassay using an ADVIA Centaur platform (Siemens AG, Erlangen, Germany) with a lower limit of quantification of 20 pmol/L. HbA_1C was measured by high-performance liquid chromatography (Menarini HA-8160; Menarini Diagnostics, Florence, Italy).

The two pump models used in the study delivered either a bolus of 1 U every 2 s (Animas pump) or a microbolus of 0.05 U every 2 s (Medtronic pump). Thus, the two pumps did not differ in the frequency with which they dispensed the boli/microboli, but they did significantly differ in the size of the boli/microboli they delivered. To determine how accurately and reproducibly the insulin bolus doses were delivered by the two insulin pumps, 15-U boluses of insulin lispro ( = 150 μL) were dispensed from the pumps and deposited into plastic vials (PCR softtube 0.2 mL; Biozyme Diagnostik, Oldendorf, Germany). The vials were weighed before and after bolus delivery using an analytical balance (CP225D; Sartorius, Göttingen, Germany). The observed differences in weight were then divided by the specific density of insulin lispro (1.004 mg/μL and 10.04 mg/U) to obtain volumes and unit values. Measurements were made after the clamp experiments (n = 40) using the insulin-filled reservoirs and infusion sets employed in the patients. There were no differences in the insulin dose delivered by the two pumps (14.90 ± 0.05 U for Medtronic pump vs. 14.84 ± 0.04 U for Animas pump; mean ± standard error; P = 0.39) and both pumps exhibited a relatively low CV value for the delivery of 15-U boluses (Medtronic pump: 1.6% and Animas pump: 1.4%).

Statistical methods

From the obtained time courses of the intravenous glucose infusion rate (GIR) and plasma insulin concentration, the PD and PK study endpoints were derived. The primary endpoint in this study was the time to maximum GIR (T_GIRmax). Secondary PD endpoints included the area under the curve of the GIR from time zero to time t (AUC_{GIR 0-t}, with t = 60, 120, 180, 240, and 480 min), maximum GIR (GIR_max), time to 10% of maximum GIR (T_{GIRmax 10%}), time to 50% of maximum GIR (T_{GIRmax 50%}), and the onset of insulin action, calculated as the time from the start of the bolus administration until the start of the glucose infusion (T_GIRstart). Secondary PK endpoints included the area under the curve of the plasma insulin concentration from time zero to time t (AUC_{INS 0-t}, with t = 60, 120, and 480 min), maximum plasma insulin concentration (C_INSmax), time to C_INSmax (T_CINSmax), time to 10% of C_INSmax (T_{CINSmax 10%}), and time to 50% of C_INSmax (T_{CINSmax 50%}). AUCs were calculated using the trapezoidal rule.

The Kolmogorov–Smirnov test was used to test for normality of data distribution. As all data sets followed a normal distribution, statistical comparisons were performed using the two-tailed paired t-test. Results obtained with this test were consistent with those obtained using the nonparametric Wilcoxon signed-rank test. A P value <0.05 was considered to indicate statistical significance. All data are presented as means ± SEM, unless otherwise indicated. All reported P values come from the two-tailed paired t-test. Data analysis was performed using an ORIGIN software package (Version 8.5; OriginLab Corporation, Northampton, MA). Sample size calculations were based on previous studies on the subcutaneous absorption of insulin lispro, which showed a mean value of 94 min and an intrasubject CV of 35% for T_GIRmax.⁸ To detect a 20% change in T_GIRmax with a power of 80% at a P value of 0.05 using a paired t-test, the sample size was estimated to be 20. Sample size calculations were performed using the freely accessible G*Power 3 software package.⁹

Results

Subject characteristics

Twenty‐one type 1 diabetic patients were enrolled in the study. One patient was discontinued from the study due to an intravenous infusion site reaction (infusion site infiltration and mild phlebitis). The 20 patients completing the study (5 women and 15 men) had an average age of 41 ± 9 years (mean ± standard deviation [SD]; range 22–53 years) and an average BMI of 24.9 ± 2.6 kg/m² (range 20.1–29.3 kg/m²). Their mean duration of diabetes was 20.7 ± 12.3 years (mean ± SD; range 2–39 years), and their HbA_1c averaged 7.5% ± 0.9% (58 ± 10 mmol/mol) [mean ± SD; range 5.4%–8.7% (36–72 mmol/mol), normal range 4.3%–5.9% (23–41 mmol/mol)]. Sixteen patients were treated with MDI and four with CSII.

Glucose and regular human insulin concentrations at baseline

Baseline plasma glucose concentrations were comparable on the study days with fast and slow bolus delivery (5.58 ± 0.09 vs. 5.45 ± 0.06 mmol/L; P = 0.18). Similarly, plasma concentrations of regular human insulin at baseline did not differ on the two study days (57 ± 8 pmol/L for fast bolus delivery vs. 51 ± 7 pmol/L for slow bolus delivery; P = 0.41).

Pharmacodynamics

After bolus delivery of insulin lispro, plasma glucose concentrations were maintained at baseline levels in all clamp experiments. Thus, average plasma glucose concentrations measured during the course of the clamp procedure were comparable on the study days with fast and slow bolus delivery (5.56 ± 0.07 vs. 5.49 ± 0.05 mmol/L; P = 0.22). Furthermore, the variability in the plasma glucose concentration during the clamp procedure, as measured by the CV, did not differ on the study days with fast and slow bolus delivery (8.9% ± 0.5% vs. 8.8% ± 0.5%; P = 0.21). In addition, the mean differences between the glucose level observed at baseline and the glucose concentrations observed during the clamp period were similar on the two study days (0.007 ± 0.006 vs. −0.001 ± 0.004 mmol/L; P = 0.20).

As can be seen in Figure 1A, speed of bolus delivery had a substantial effect on the PD profile of insulin lispro. The T_GIRmax, T_GIRstart, T_GIRmax10%, and T_GIRmax50% were all significantly longer with slow bolus delivery as compared with fast bolus delivery (Table 1). Furthermore, the AUC_{GIR 0–60min} and AUC_{GIR 0–120min} for slow bolus delivery were significantly smaller than those for fast bolus delivery (Table 1). There were no statistical differences in GIR_max or AUC_GIRtotal between the two bolus delivery speeds (Table 1).

FIG. 1.

(A) Average time course (n = 20, means ± SE) of GIR after s.c. administration of 15 U of insulin lispro with short (30 s; solid line) and long (10 min; dashed line) bolus durations. (B) Average time course (n = 20, means ± SE) of the insulin lispro concentration in plasma after s.c. administration of 15 U of insulin lispro using short (black circles) and long (open circles) bolus durations. GIR, glucose infusion rate; SE, standard error.

Table 1.

Summary of Pharmacokinetic and Pharmacodynamic Parameters

Parameter ^a	Fast bolus mean (SE)	Slow bolus mean (SE)	P^b
Pharmacodynamics
T_GIRmax (min)	98 (11)	125 (16)	0.005
T_GIRstart (min)	21.0 (2.5)	34.3 (2.7)	0.002
T_{GIRmax 10%} (min)	22.7 (3.0)	35.9 (2.7)	0.005
T_{GIRmax 50%} (min)	49.9 (3.5)	67.4 (8.4)	0.012
GIR_max [mg/(kg·min)]	7.49 (1.07)	7.43 (1.01)	0.923
AUC_{GIR 0–60min} (mg/kg)	123 (28)	74 (25)	0.018
AUC_{GIR 0–120min} (mg/kg)	474 (78)	353 (47)	0.019
AUC_{GIR 0–180min} (mg/kg)	807 (119)	674 (77)	0.087
AUC_{GIR 0–240min} (mg/kg)	1082 (139)	950 (90)	0.152
AUC_{GIR total} (mg/kg)	1356 (147)	1271 (98)	0.388
Pharmacokinetics
T_{CINSmax 10%} (min)	13.3 (1.6)	18.1 (1.5)	0.003
T_{CINSmax 50%} (min)	30.3 (2.1)	34.8 (2.2)	0.046
T_CINSmax (min)	75.0 (5.2)	86.0 (6.8)	0.099
C_INSmax (pmol/L)	345 (33)	333 (23)	0.518
AUC_{INS 0–60min} (min·pmol/L)	10,307 (1291)	8192 (865)	0.027
AUC_{INS 0–120min} (min·pmol/L)	27,456 (2614)	25,119 (1766)	0.154
AUC_{INS total} (min·pmol/L)	55,072 (3200)	54,186 (2476)	0.604

GIR_max, maximum glucose infusion rate; T_GIRmax, time to reach GIR_max; T_{GIRmax 10%}, time to reach 10% of GIR_max; T_{GIRmax 50%}, time to reach 50% of GIR_max; AUC_{GIR total}, area under the GIR curve from 0 min until clamp end; AUC_{GIR 0-tmin}, area under the GIR curve from 0 until a specified time; GIR_start, onset of insulin action, defined as the time from bolus administration until the plasma glucose concentration had decreased at least by 5 mg/dL from baseline. C_INSmax, maximum insulin concentration; T_CINSmax, time to reach C_INSmax; T_{CINSmax 10%}, time to reach 10% of C_INSmax; T_{CINSmax 50%}, time to reach 50% of C_INSmax; AUC_{INS total}, area under the insulin concentration curve from 0 min until clamp end; AUC_{INS 0-tmin}, area under the insulin concentration curve from 0 until a specified time.

With paired t-test; similar values with Wilcoxon signed-rank test.

SE, standard error.

Pharmacokinetics

As shown in Figure 1B, insulin bolus delivery with low delivery speed resulted in a slower increase in the plasma insulin concentration as compared with that with fast delivery speed. Both T_{CINSmax 10%} and T_{CINSmax 50%} were significantly longer with slow bolus delivery than with fast bolus delivery (Table 1). Furthermore, the AUC_{INS 0–60min} was significantly lower for the slow bolus delivery than for the fast bolus delivery (Table 1). The T_CINSmax for slow bolus delivery was also increased compared with that for fast bolus delivery, but this difference did not reach significance (P = 0.099; Table 1). There were no statistical differences in C_INSmax or AUC_INStotal between the slow and fast bolus delivery (Table 1). In no one of the subjects did any insulin leakage occur from the infusion site after bolus administration.

Discussion

The magnitude of the speed at which a subcutaneous insulin bolus is administered varies considerably in commercial insulin pumps,³ with a 20-fold difference between the insulin pump model employing the lowest (40 s for 1 insulin unit) and highest speed (2 s for 1 insulin unit). In this study we examined in type 1 diabetic subjects whether such a difference in the bolus delivery speed would affect insulin absorption from the subcutaneous tissue. To this end, a subcutaneous bolus of 15 U of rapid-acting insulin was administered twice to each subject, once using the pump model with the lowest delivery speed (Medtronic Minimed Paradigm 512 pump; delivers a bolus of 0.05 U every 2 s) and another time using the pump model with the highest delivery speed (Animas IR2020 pump; delivers a bolus of 1 U every 2 s). After the bolus administrations, the subcutaneous insulin absorption was assessed by measuring changes in the plasma insulin concentration and determining the time action profile of insulin using the glucose clamp technique. We found that the bolus delivery with low delivery speed (bolus duration of 10 min) substantially reduced the rate of subcutaneous insulin absorption, resulting in a much later onset of insulin action and a markedly increased time to peak insulin action in comparison with the delivery of the same bolus amount with fast delivery speed (bolus duration of 30 s).

The present finding showing that the speed of bolus delivery influences subcutaneous absorption of rapid-acting insulin is consistent with two previous studies examining the effect of bolus delivery speed on the subcutaneous absorption of porcine insulin in healthy humans.^4,10 In one of the two studies,⁴ a subcutaneous insulin bolus of 10 U was delivered with bolus durations of 5 and 30 min using an infusion pump. In the other study, ∼8 U of porcine insulin was given once as a rapid bolus injection using an insulin syringe and another time as a bolus delivered for 17 min using an infusion pump.¹⁰ In both studies, subcutaneous absorption of porcine insulin was found to be significantly slowed down with long bolus durations (i.e., low delivery speeds) as compared with short bolus durations (i.e., fast delivery speeds). However, the results of the present and previous studies seemingly contrast with findings by Hildebrandt et al.,¹¹ who studied the effect of the injection speed on the subcutaneous absorption of porcine insulin in type 1 diabetic subjects. In their study, 10 U of porcine insulin was injected subcutaneously for 30, 3, and <1 s using an insulin syringe. They reported no influence of the injection speed on the subcutaneous absorption of porcine insulin.¹¹ Thus, taken together, the present and previous findings suggest that the subcutaneous insulin absorption is affected at bolus durations that vary from 30 s to 30 min, but variation in bolus duration between 0 and 30 s does not affect subcutaneous insulin absorption.

There are several possible explanations for the observed relationship between the length of the bolus duration and the magnitude of its slowing effect on the subcutaneous insulin absorption. One explanation may be that the tissue pressure generated during bolus delivery with fast delivery speed is much higher than that generated during bolus delivery with slow delivery speed. Since high tissue pressures may significantly decrease the tissue resistance to convective fluid flow,¹² larger insulin distribution volumes may be obtained with fast delivery speeds compared with that obtained with slow delivery speeds. Thus, the larger distribution volume possibly achieved during the bolus delivery with fast delivery speed may expose a higher number of capillaries to insulin, thereby increasing the absorption of insulin. Another possible explanation for our observation may be inferred from the subcutaneous absorption properties of water, the solvent used in insulin formulations. It has been shown previously^13,14 that the absorption of subcutaneously injected water (molecular weight: 18) or sodium (molecular weight: 23) is very fast in comparison with molecules of larger size (e.g., insulin; molecular weight: ∼5800). Within 5 min after injection, about 70% of the injected water was absorbed into the blood stream.¹³ Similarly, within 10 min after subcutaneous injection, about 50% of the injected sodium entered the vascular system.¹⁴ Thus, during bolus administration of insulin with long bolus duration (e.g., 10 min), absorption of fluid (i.e., water) may be very large, resulting in diminished convective fluid flow in the surrounding of the infusion site. Therefore, as the capillary walls in subcutaneous tissue may act as a low permeable barrier to insulin (i.e., the reflection coefficient of the subcutaneous vasculature may be very high for insulin¹⁵), insulin molecules may accumulate in the vicinity of the infusion site (Fig. 8 in Regittnig and Jungklaus¹⁶). Conversely, during bolus administration of insulin with short bolus duration (e.g., <1 min), fluid absorption may be insignificant, which in turn will increase interstitial convective flow in the surrounding of the infusion site, thereby causing the insulin molecules to distribute in a larger volume in the tissue surrounding the infusion site (Fig. 8 in Regittnig and Jungklaus¹⁶). A larger distribution volume may expose a higher number of capillaries to insulin, thereby increasing the absorption of insulin.

It has been shown that, when injected subcutaneously by an insulin syringe or insulin pen, commercially available rapid-acting insulin analogs (lispro, aspart, and glulisine) exhibit much faster PK and PD than regular human insulin.^8,17 For instance, in a study⁸ comparing the absorption of subcutaneously injected insulin lispro, insulin glulisine, and regular human insulin (0.2 U/kg), the average T_GIRmax value obtained for regular human insulin was ∼67% higher than that determined for insulin lispro and insulin glulisine (161 vs. 94 vs. 98 min, respectively). Similarly, in a study examining the subcutaneous absorption of a 0.2 U/kg dose of insulin aspart and regular human insulin,¹⁷ the average T_GIRmax value determined for regular human insulin was about 50% higher than that obtained for insulin aspart (156 vs. 104 min, respectively). Thus, because of their fast subcutaneous absorption properties, the three commercially available insulin analogs can be administered just before a meal, whereas regular human insulin needs to be injected some 30 min or more before a meal for a comparable postprandial glucose control.¹⁸

From a comparison of the PK and PD results of this study with those of the previous insulin analog studies, it can be inferred that subcutaneous absorption of rapid-acting insulin after bolus delivery with short bolus duration is similar to that seen after administration by conventional pen or syringe injection. For instance, the mean T_GIRmax value obtained in this study for bolus delivery of insulin lispro with short bolus duration (Table 1; 98 min) compares well with those previously reported for the insulin analogs administered at a similar dose (0.2 U/kg) by insulin pen or insulin syringe injection (lispro: 94 min; glulisine: 98 min; aspart: 104 min^8,17). However, the mean T_GIRmax value obtained for delivery of insulin lispro with long bolus duration (Table 1; 125 min) is about 30% higher than those previously determined for the three insulin analogs. Thus, bolus administration of rapid-acting insulin with long bolus duration slows down subcutaneous absorption so that the resulting PK and PD profile approaches that seen after subcutaneous injection of regular human insulin. Therefore, to preserve the favorable absorption properties of rapid-acting insulins when administered subcutaneously by insulin pumps, we recommend to carry out insulin bolus delivery with bolus durations of 1 min or less. Bolus delivery with such short bolus durations may allow the administered insulin to reach the same distribution volume and absorption rates as with conventional rapid syringe injections.

Bolus delivery with bolus durations of 1 min or less may be incorporated into insulin pumps in two principal ways. One is to fix the bolus duration to a value that is equal to or lower than the proposed upper limit threshold of 1 min and adjust the delivery speed to the bolus size chosen by the user (bolus delivery speed equals bolus size divided by bolus length). Thus, when a larger bolus dose is delivered at fixed bolus duration, a faster delivery speed is needed. A potential disadvantage of using this fixed bolus duration approach may be that, when very large bolus doses are administered, the resulting high delivery speeds could cause leakage of insulin back to the skin surface and induce burning sensation at the cannula insertion site. The other approach to bolus delivery with bolus duration of 1 min or less is to fix the bolus delivery speed to a relatively high value (e.g., 1 s for 1 U) and adjust the bolus duration for the chosen bolus size. A potential drawback of using this fixed delivery speed approach may be that, when very large bolus doses are delivered, the resulting bolus durations exceed the proposed threshold duration, which in turn may cause delayed subcutaneous insulin absorption. Further studies are required to determine which of the two approaches is more appropriate for incorporating the proposed mode of bolus delivery into insulin pumps.

Conclusions

Our data show that subcutaneous absorption of rapid-acting insulin after bolus delivery with long bolus duration (10 min) is substantially slower than the subcutaneous absorption of the same amount of insulin after bolus delivery with short bolus duration (30 s). Thus, to avoid delayed subcutaneous absorption of rapid-acting insulin after bolus administration, we recommend the performance of insulin bolus delivery with bolus durations of 1 min or less. If incorporated in future insulin pumps, this mode of bolus delivery may allow the administered rapid-acting insulin to reach the same absorption rates as found with conventional insulin syringe or insulin pen injection. Further studies are needed to test whether bolus delivery speed also affects the subcutaneous absorption of novel ultrafast-acting insulins.¹⁹

Footnotes

Acknowledgments

This work was supported by funding from the European Commission Framework Program 7 grants: FP7-ICT-2009-4, no. 247138 (www.apathome.eu) and FP7-HEALTH-2012.2.4.3-1, no. 305654 (www.pcdiab.eu).

We are grateful to A. Berghofer, M. Brunner, L. Stach, C. Missbrenner, S. Friedrich, M. Jungklaus, M. Ehrmann, and S. Korsatko, all of the Department of Internal Medicine, Medical University of Graz, for their expert assistance in conducting the studies, and all volunteers for participating in the study.

Author Disclosure Statement

T.R.P. is an advisory board member of Novo Nordisk A/S, a consultant for Roche Diabetes Care, Novo Nordisk A/S, Eli Lilly & Co, Infineon, Carnegie Bank, on speaker's bureau of Novo Nordisk A/S and Astra Zeneca and is a cofounder and shareholder of Smart*Med GmbH and decide Clinical Software GmbH. J.K.M. is a member of the advisory board of Boehringer Ingelheim, Becton-Dickinson, Eli Lilly, Medtronic, and Sanofi-Aventis, and has received speaker honoraria from Abbott Diabetes Care, Astra Zeneca, Eli Lilly, Nintamed, Novo Nordisk A/S, Roche Diabetes Care, Sanofi-Aventis, Servier, and Takeda, and is a shareholder of decide Clinical Software GmbH. M.E. is an employee of BBraun AG, and a cofounder and shareholder of Smart*Med GmbH. W.R. has filed patent applications relating to the work described in this article. There are no other potential conflicts of interest relevant to this article.

References

Pickup

: Insulin-pump therapy for type 1 diabetes mellitus. N Engl J Med, 2012; 366:1616–1624.

Cummins

, Royle

, Snaith

, et al.: Clinical effectiveness and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes: systematic review and economic evaluation. Health Technol Assess, 2010; 14:iii-iv, xi-xvi, 1–181.

Diabetes Health–Insulin Pumps Chart. 2017. www.diabeteshealth.com/wp-content/uploads/2017/03/DHInsulinPumpChart20162017-1.pdf (accessed September 24, 2018 )

Chisholm

, Kraegen

, Zelenka

: Programming of insulin delivery with meals during subcutaneous insulin infusion. Diabetes Care, 1981; 4:265–268.

DeFronzo

, Tobin

, Andres

: Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol, 1979; 237:E214–E223.

Frid

, Linde

: Intraregional differences in the absorption of unmodified insulin from the abdominal wall. Diabet Med, 1992; 9:236–239.

Lindström

, Hedman

, Arnqvist

: Use of a novel double-antibody technique to describe the pharmacokinetics of rapid-acting insulin analogs. Diabetes Care, 2002; 25:1049–1054.

Becker

RHA

, Frick

: Clinical pharmacokinetics and pharmacodynamics of insulin glulisine. Clin Pharmacokinet, 2008; 47:7–20.

G*Power 3 software package,. 2014. www.gpower.hhu.de/ (accessed September 24, 2018 )

10.

Home

, Pickup

, Keen

, et al.: Continuous subcutaneous insulin infusion: comparison of plasma insulin profiles after infusion or bolus injection of the mealtime dose. Metabolism, 1981; 30:439–442.

11.

Hildebrandt

, Sestoft

, Nielsen

: The absorption of subcutaneously injected short-acting soluble insulin: influence of injection technique and concentration. Diabetes Care, 1983; 6:459–462.

12.

Levick

: Flow through interstitium and other fibrous matrices. Q J Exp Physiol, 1987; 72:409–437.

13.

Secher-Hansen

: Studies on the subcutaneous absorption in mice. VII. Absorption of 3H2O and 14C-sucrose from non-buffered solutions at different pH values. Acta Pharmacol Toxicol, 1970; 28:97–101.

14.

Forbes

, Deisher

, Perley

, et al.: Effect of hyaluronidase on the subcutaneous absorption of electrolytes in humans. Science, 1950; 111:177–179.

15.

Michel

: Transport of macromolecules through microvascular walls. Cardiovasc Res, 1996; 32:644–653.

16.

Regittnig

, Jungklaus

: Infusion arrangement and method. US patent, 9,320,851, 2016.

17.

Heinemann

, Weyer

, Rauhaus

, et al.: Variability of the metabolic effect of soluble insulin and the rapid-acting insulin analog insulin aspart. Diabetes Care, 1998; 21:1910–1914.

18.

Home

: The pharmacokinetics and pharmacodynamics of rapid-acting insulin analogues and their clinical consequences. Diabetes Obes Metab, 2012; 14:780–788.

19.

Heinemann

, Muchmore

: Ultrafast-acting insulins: state of the art. J Diabetes Sci Technol, 2012; 6:728–742.