J Artif Organs (2004) 7:91100 DOI 10.

1007/s10047-004-0251-2

The Japanese Society for Articial Organs 2004

ORIGINAL ARTICLE Taiji Sekigami, MD Seiya Shimoda, MD, PhD Kenro Nishida, MD, PhD Yasuto Matsuo, MD Shinji Ichimori, MD Kenshi Ichinose, MT Motoaki Shichiri, MD, PhD Michiharu Sakakida, MD, PhD Eiichi Araki, MD, PhD

Comparison between closed-loop portal and peripheral venous insulin delivery systems for an articial endocrine pancreas

Abstract To establish the ideal insulin delivery route for an articial endocrine pancreas, we examined the effectiveness of closed-loop portal insulin delivery. We investigated the effects of the route of insulin delivery on net hepatic glucose balance (NHGB) in dogs under pancreatic clamp conditions with somatostatin plus basal glucagon and insulin infusions. A constant rate of suprabasal insulin was infused via the portal vein or a peripheral vein, and glucose was infused into the portal vein for 180 min. The mean net hepatic glucose uptake (NHGU) values in the portal insulin infusion group (PI group) were signicantly greater than those in the peripheral venous insulin infusion group (VI group); the changes from the baseline values at 180 min were 3.54 0.66 and 2.45 0.82 mg kg 1 min 1 in the PI and VI groups, respectively, P 0.05. Furthermore, dogs under pancreatic clamp conditions were controlled after a 2-g/kg oral glucose load by applying the closed-loop intraportal (PO) or intravenous (IV) insulin infusion algorithm. There were no signicant differences in glycemic control and insulin requirements between these algorithms. However, the maximum peripheral venous and arterial plasma insulin concentrations with the PO algorithm were signicantly lower than those with the IV algorithm [305.1 68.9 and 468.1 66.9 pmol/l (peripheral vein) and 305.3 62.9 and 469.6 85.1 pmol/l (artery) with the PO and IV algorithms, respectively, P 0.05]. On the other hand, the maximum portal plasma insulin concentration with the PO algorithm was signicantly higher than that with the IV

algorithm (619.9 101.7 and 414.3 79.9 pmol/l with the PO and IV algorithms, respectively, P 0.05). The mean NHGU values with the PO algorithm were signicantly greater than those with the IV algorithm. Our results conrmed that closed-loop portal insulin delivery is feasible with regard to both insulin proles and hepatic glucose handling in vivo, and indicated that the portal vein is the most suitable insulin delivery route for the articial endocrine pancreas. Key words Closed-loop portal insulin delivery Insulin infusion algorithm Articial endocrine pancreas Net hepatic glucose balance

Introduction
Strict glycemic control with intensive insulin therapy can prevent the development of diabetic microvascular complications in both type 1 1and type 2 diabetes.25 However, there are potential adverse effects and inconvenience to patients receiving such intensive insulin treatment, which include episodic hypoglycemia, weight gain, frequent glucose monitoring, and the need for either frequent insulin administration or the use of an external insulin pump. The obvious way to accurately compensate for impaired betacell function in diabetes patients is replacement of islets or pancreas transplantation. However, the lack of availability of these resources is a major problem, and techniques for supplying sufcient functional beta-cells or islets have not yet been perfected. These problems in achieving long-term physiological glycemic control have stimulated us to use a closed-loop control system with an articial endocrine pancreas (AEP).6 Our AEP consists of a needle-type glucose sensor covered with a newly designed highly biocompatible membrane, a microcomputer system, and two precision microroller pumps for insulin and glucagon infusion, and the whole unit is powered by a battery.6,7 Intravenous6,8 or subcutaneous9 infusion of insulin using this system facilitated

Received: February 18, 2004 / Accepted: April 20, 2004 T. Sekigami S. Shimoda (*) K. Nishida Y. Matsuo S. Ichimori K. Ichinose M. Sakakida E. Araki Department of Metabolic Medicine, Graduate School of Medical Sciences, Kumamoto University, 1-1-1 Honjo, Kumamoto 860-8556, Japan Tel. 81-96-373-5169; Fax 81-96-366-8397 e-mail: sshimoda@gpo.kumamoto-u.ac.jp M. Shichiri Research Institute of Life Style-Related Diseases, Osaka, Japan

92

glycemic regulation in ambulatory diabetes patients. However, intravenous or subcutaneous insulin administration, which delivers insulin directly into the systemic circulation, increases glucose utilization mainly in peripheral tissue and has lesser effects on hepatic glucose metabolism compared with portal insulin delivery even when normoglycemia is achieved.10,11 Therefore, with respect to the effect of insulin on the liver and peripheral tissues, the insulin delivery route is a key issue in the development of the AEP to achieve physiological glucose metabolism in vivo. Numerous studies have compared intraportal and peripheral intravenous insulin delivery using open-loop technique or pancreas transplantation;1018 however, the importance of a physiological portal/peripheral insulin gradient for glucose metabolism is still not completely clear.19 Consequently, the rationales for the portal route of insulin delivery in the treatment of diabetes patients has not been established. The present study was designed to examine the advantages and disadvantages of portal insulin delivery to establish the ideal insulin delivery route for the AEP. For this purpose, we compared blood glucose levels, plasma insulin proles, and the net hepatic glucose balance (NHGB) achieved with closed-loop portal insulin delivery and peripheral insulin delivery by using an AEP. Furthermore, we performed glycemic control after a 2-g/kg oral glucose load in dogs under pancreatic clamp conditions with the closedloop intraportal insulin infusion algorithm and evaluated NHGB after the oral glucose load.

only on dogs that had (1) a leukocyte count lower than 16 000/mm3, (2) a hematocrit value greater than 38%, (3) a good appetite (consuming all of the daily ration), and (4) normal stools. Effects of route of insulin delivery on NHGB We investigated the inuence of insulin delivery route on NHGB by using the arteriovenous difference technique in dogs (n 5). Under general anesthesia, a catheter for blood sampling was inserted in the hepatic vein via the femoral vein under uoroscopy. The tip of the hepatic venous catheter was placed 1 cm inside the left common hepatic vein, which carries almost half of the venous outow of the liver, the largest portion carried by any hepatic vein.20 Indwelling catheters were introduced into the antecubital veins for indocyanine green (ICG) infusion (Diagnogreen, Daiichi, Tokyo, Japan) and somatostatin infusion (Somatostatin; Calbiochem, Darmstadt, Germany). A catheter for blood sampling was inserted into the femoral artery. Each experiment consisted of a hormonal-adjustment and a dyeequilibration period (from 120 to 30 min), a basal period (from 30 to 0 min), and an experimental period (from 0 to 180 min) (Fig. 1). At 120 min, a constant infusion of ICG (0.07 mg/min) was commenced to measure hepatic blood ow (HBF). A constant infusion of somatostatin (0.8 g kg 1 min 1) was started via the antecubital vein route to inhibit endogenous secretion of insulin and glucagon, and constant intraportal infusions of basal glucagon (Glucagon G Novo; Novo Nordisk, Copenhagen, Denmark, 0.5 ng kg 1 min 1) and insulin (Novolin R, U-40; Novo Nordisk, 0.2 mU kg 1 min 1) were commenced to replace endogenous secretion of these hormones (pancreatic clamp). In addition, a constant peripheral infusion of paminohippuric acid (PAH) (sodium para-aminohippurate, Daiichi, 1.7 mol kg 1 min 1) was started and continued until the intraportal infusion of glucose mixed with PAH was commenced at 0 min. At 0 min, suprabasal insulin (0.8 mU kg 1 min 1) was infused via the portal vein (PI group) or peripheral vein (VI group), and glucose (8 mg kg 1 min 1)/PAH (PAH mixed with glucose in such a way that the PAH infusion rate was 1.7 mol kg 1 min 1) was infused into the portal vein for 180 min. Blood samples were collected for the measurement of blood glucose concentrations from the portal vein, hepatic vein, and femoral artery at 30, 20, 10, 0, 20, 40, 60, 90, 120, and 180 min. Closed-loop intraportal insulin infusion algorithm Plasma insulin responses and glucose infusion rates (GIR) against pulsatile insulin injection Each dog was anesthetized and placed in the Fowler position after a 42-h fast. At 120 min, a pancreatic clamp was started. To analyze insulin kinetics after portal or peripheral venous injection, 0.1 U/kg regular insulin was administered to the dogs (n 10) through the portal or peripheral

Materials and methods

Animals and surgical procedures All experiments in this study were designed according to the Guidelines for Animal Experimentation of Kumamoto University. Experiments were conducted in beagle dogs weighing from 9 to 12 kg. These dogs were fed twice daily (morning and evening) with a total caloric intake of 80 kcal kg 1 day 1 (Pedigree Principal Adult, 41.7% carbohydrate, 24.8% protein, 33.5% fat; Master Foods, Kanagawa, Japan). The body weight of the dogs did not change signicantly during the experimental period. Each dog underwent laparotomy under general anesthesia with pentobarbital sodium (Nembutal; Dainabot, Tokyo, Japan), and two catheters, one for insulin, glucose, and glucagon infusion and the other for portal blood sampling, were inserted through the jejunal veins. The tips of these catheters were placed into the portal vein under uoroscopy. The catheters were tunneled under the skin to outlet sites on the back and connected to infusion pumps. They were lled with heparinized saline to avoid catheter obstruction. Ultrasonic ow probes (Transonic, Ithaca, NY, USA) were placed around the portal vein and the hepatic artery. The ends of the ow probe leads were placed in subcutaneous pockets. Two weeks after the surgery, blood was drawn to determine the leukocyte count and hematocrit values of the animals. Studies were conducted

93 Fig. 1. Experimental protocol for the effects of different insulin delivery routes on net hepatic glucose balance (NHGB). PI group, suprabasal insulin was infused via the portal vein; VI group, suprabasal insulin was infused via a peripheral vein; BG, blood glucose; ICG, indocyanine green

mean plasma insulin concentrations after portal insulin injection in the dogs by the nonlinear least-squares method. Derivation of the closed-loop intraportal insulin infusion (PO) algorithm The biphasic insulin response observed in glucose perifusion experiments with pancreatic islets was simulated by using a transfer function with a rst-order delay in both the proportional and derivative actions to blood glucose concentrations.22 By applying this principle, the PO algorithm was developed as for the closed-loop intravenous insulin infusion (IV) algorithm. The relationship between plasma insulin concentration and blood glucose concentration is shown by Eq. 3, as described in detail elsewhere.11
IRI(t ) aG(t )

Fig. 2. A mathematical model for pharmacokinetic analysis of intraportal insulin injection. X(t) is the plasma insulin dose (mU/kg) and IIR(t) and IRI(t) are the insulin infusion rate (mU kg 1 min 1) and plasma insulin level (pmol/l). p is the insulin degradation rate (/min)

venous route at 0 min. Blood glucose was clamped from 30 min at the fasting blood glucose level for each dog for the glucose clamp study by portal glucose infusion. GIR values were calculated by using a bedside-type AEP (model STG-22; Nikkiso, Tokyo, Japan)21 that was operated with a glucose infusion algorithm.8 Blood samples were collected from the femoral vein at 30, 20, 10, 0, 5, 10, 20, 30, 40, 50, 60, 80, 100, 120, 140, and 160 min. Pharmacokinetic analysis by a compartmental model The one compartment model was applied to analyze insulin kinetics after portal insulin injection (Fig. 2). In this model, the insulin dose in the compartment was expressed as shown in Eqs. 1 and 2. dX (t ) dt IRI(t ) IIR(t ) X (t ) p X (t ) (1) (2)

bdG(t ) dt

(3)

where G(t) is the concentration of blood glucose at time t (mmol/l); dG(t)/dt is the rate of change in blood glucose concentration computed over 1 min at time t (mmol l 1 min 1); a, b, and c are the individual parameters for insulin secretion. By substituting Eqs. 1 and 2 into Eq. 3, the PO algorithm dependent on the glycemic level was obtained as Eq. 4. IIR(t )
Kp Kc Kd Kp

KpG(t )
1 p

KddG(t ) dt
ba

Kc

(4)

ap 0.225 Kp c a

where X(t) is the plasma insulin dose (mU/kg), IIR(t) and IRI(t) are the insulin infusion rate (mU kg 1 min 1) and plasma insulin concentration (pmol/l). and p are plasma insulin volume (l/kg) and insulin degradation rate (/min). (pmol/mU) is a conversion factor ( 6). To calculate the pharmacokinetic parameters ( and p), we simulated the

where Kp (mU l mmol 1 kg 1 min 1) is the coefcient for proportional action, Kd (mU l mmol 1 kg 1) is the coefcient for derivative action, and Kc (mU kg 1 min 1) is the constant for basal insulin secretion. The intravenous basal insulin infusion rate is dened as 0.225 mU kg 1 min 1.8 In the equation,

94

the second-order derivative of G(t) was omitted to avoid instability of glycemic control caused by noise generated from the glucose monitoring unit. Thus, we derived the PO algorithm as the summation of the proportional and rstorder derivative actions to changes blood glucose concentration. We calculated the parameters (Kp, Kd, and Kc) by substituting pharmacokinetic parameters and individual parameters. Time course of NHGB during glycemic control after oral glucose load by using a bedside-type AEP To evaluate the appropriateness of the PO algorithm, we performed glycemic control after a 2-g/kg oral glucose load in 42-h-fasted dogs (n 5) under pancreatic clamp conditions using the catheters as mentioned above by applying the PO or IV algorithm. Dogs were randomly assigned to the two protocols on two separate days within a one-week period (n 5). The degree of glycemic control, the total amount of insulin infused, and the plasma insulin proles obtained with the PO algorithm were compared with the corresponding values obtained with the IV algorithm. This experiment consisted of a hormonal-adjustment and dyeequilibration period (from 120 to 30 min), a basal period (from 30 to 0 min), and an experimental period (from 0 to 240 min) (Fig. 3). Constant infusions of ICG, somatostatin, basal insulin, and glucagon were introduced as indicated above. Basal insulin infusion was stopped and the bedsidetype AEP was started at 0 min and glucose (2-g/kg) was administered through the gastric tube after intramuscular (5 mg) and intravenous (5 mg) injections of metoclopramide (Fujisawa, Tokyo, Japan). Blood samples were taken from the femoral vein for the measurement of plasma insulin concentrations initially, every 10 min from 0 min to 180 min, and every 20 min from 180 to 240 min. Samples were also

taken for measurements of blood glucose concentrations and plasma insulin concentrations from the portal vein, hepatic vein, and femoral artery at 30, 20, 10, 0, 20, 40, 60, 90, 120, 180, and 240 min after the oral glucose load. Analytical procedures Blood glucose concentrations in whole blood samples were measured by the glucose oxidase method. In the experiments using the AEP, blood glucose concentrations were also measured by the sensor system based on the glucose oxidase method.21 Plasma insulin concentrations were measured by a radioimmunoassay kit (Insulin Riabead II, Dainabot).23 ICG was measured spectrophotometrically at 810 nm to estimate HBF according to the method of Leevy et al.24 Calculations and data analysis HBF was measured by ultrasonic ow probes and by use of ICG dye extraction. The two methods yielded similar results, but the data reported here were calculated with the ultrasonically determined ows because their measurement did not require any assumptions regarding the relative contribution of arterial and portal ows to total hepatic blood ow. The ICG technique was used as a backup for the ultrasonic ow probes. The recovery of PAH across the liver was measured as reported by Pagliassotti et al.,25 and allowed assessment of mixing of the infusate with the portal blood. When the ratio of recovered PAH to infused PAH in the portal or hepatic vein was greater than 1.4 or less than 0.6 for more than three time points, these data were excluded. NHGB was determined using the formula ([G]H HBF) ([G]A ABF [G]P PBF), where [G]A, [G]P, and

Fig. 3. Experimental protocol for evaluating the time course of NHGB during glycemic control after an oral glucose load. PO algorithm, insulin was infused via the portal vein; IV algorithm, insulin was infused via a peripheral vein; BG, blood glucose; IRI, plasma insulin concentration

95

[G]H are glucose concentrations in the artery, the portal vein, and the hepatic vein, respectively, and ABF and PBF refer to blood ow through the hepatic artery and the portal vein, respectively.25,26 HBF is a summation of ABF and PBF. IRI is the summation of plasma insulin concentrations at 0, 20, 40, 60, 90, 120, and 180 min. Statistical analysis Results were expressed as mean SD. Differences between groups were examined for statistical signicance using Students t test. P 0.05 denoted a statistically signicant difference.

between NHGB for the basal period (from 30 to 0 min) and at the end of the experimental period were 3.54 0.66 mg kg 1 min 1 in the PI group and 2.45 0.82 mg kg 1 min 1 in the VI group (P 0.05). Closed-loop intraportal insulin infusion algorithm Plasma insulin responses and GIR for pulsatile insulin injection Figure 5 shows plasma insulin responses (Fig. 5A) and GIR (Fig. 5B) after insulin injection. The plasma insulin concentration at 5 min after portal insulin injection was 1323.6 58.8 pmol/l, which was signicantly lower than that at 5 min after peripheral venous insulin injection (1656.6 58.2 pmol/l, P 0.05). The plasma insulin concentration after portal insulin injection was also signicantly lower than that after peripheral venous insulin injection until after 20 min (P 0.05). The total amount of glucose infused after portal or peripheral venous insulin injection from 0 to 160 min was 879.9 96.7 and 967.6 101.3 mg/kg, respectively, and the values were not signicantly different (P 0.06). Pharmacokinetic analysis by a compartmental model Figure 6 depicts the plasma insulin curve (solid line) simulated by the nonlinear least-squares method, and mean plasma insulin levels (lled circles) measured in dogs under pancreatic clamp conditions after intraportal injection of regular insulin. The pharmacokinetic parameters ( and p) were obtained from application of the compartmental model to this curve, and calculated to be 0.157 (l/kg) and p 0.214 (/min). Derivation of the PO algorithm The parameters for the PO algorithm were calculated from the pharmacokinetic parameters of intraportally injected insulin and the individual parameters (a, b, and c; 18, 72, and 72, respectively)8 as follows: Kp 0.612, Kd 5.24, and Kc 2.42. The parameters (Kp, Kd, and Kc) for the IV algorithm were (0.504, 4.86 and 2.02) as reported previously.8 Glycemic control and time course of NHGB after oral glucose load The results of glycemic control after a 2-g/kg oral glucose load in ve dogs under pancreatic clamp conditions using the PO or IV algorithm are shown in Figure 7 and Table 1. With the PO or IV algorithm, the blood glucose concentration ranged from 4.91 0.36 or 4.96 1.11 mmol/l at 0 min to a maximum level of 9.69 1.24 or 10.11 1.63 mmol/l at 65 min and decreased to 4.78 0.50 or 4.70 1.32 mmol/l at 240 min, respectively (Fig. 7A). There were no signicant differences between the groups. The total amount of insulin infused with the PO algorithm over 240 min was comparable

Results
Effects of route of insulin delivery on NHGB under pancreatic clamp conditions The PI and VI groups exhibited a similar rate of net hepatic glucose production during the basal period (1.25 0.09 and 1.20 0.20 mg kg 1 min 1, respectively, P 0.60) (Fig. 4). In the PI group, NHGB switched from net production to net uptake as early as 20 min after the suprabasal insulin infusion. On the other hand, the switch of NHGB from net production to net uptake occurred more slowly and to a lesser extent in the VI group. The mean net hepatic glucose uptake (NHGU) in the PI group was signicantly greater than in the VI group at 20, 40, 90, and 120 min (NHGB: 2.21 0.58 and 1.07 0.43 mg kg 1 min 1 in the PI and VI groups at 120 min, respectively, P 0.05). Differences

Fig. 4. Effects of the different insulin delivery routes on NHGB. Intraportal glucose infusion was performed in 42-h-fasted dogs maintained under pancreatic clamp conditions (somatostatin plus portal vein insulin and glucagon infusions) with suprabasal insulin via the portal or a peripheral vein. NHGB was calculated as described in the text. The solid line depicts data from the PI group, in which suprabasal insulin was infused into the portal vein; the broken line depicts data from the VI group, in which suprabasal insulin was infused into a peripheral vein. Data are mean SD values for the ve dogs in each group. * P 0.05 for the PI group versus the VI group

96

Fig. 5. Plasma insulin responses (A) and glucose infusion rates (B) against pulsatile insulin injection (0.1 U/kg, n 10). Insulin was administered to 42-h-fasted dogs through the portal vein or a peripheral vein under pancreatic and glucose clamp conditions. Plasma insulin concentrations were measured and glucose infusion rates were calculated as

described in the text. The solid line depicts data from intraportal insulin injection and the broken line depicts data from peripheral venous insulin injection. Data are mean SD values for ten dogs in each group. * P 0.05 for intraportal insulin injection versus peripheral venous insulin injection

those with the IV algorithm during the 40-to-90-min period (P 0.05) (Fig. 7C, 7D). On the other hand, the portal plasma insulin concentration with the PO algorithm was signicantly higher than that with the IV algorithm during the 20-to-90-min period (P 0.05) (Fig. 7E). The mean levels of NHGU with the PO algorithm were signicantly greater than those with the IV algorithm at 20, 40, 60, and 90 min (NHGB: 3.82 0.81 or 2.36 0.80 mg kg 1 min 1 with the PO or IV algorithm at 90 min, respectively, P 0.05) (Fig. 7F). As shown in Table 1, the IRI values for the peripheral vein and artery using the PO algorithm were signicantly lower than those using the IV algorithm (P 0.05). In contrast, the IRI value for the portal vein using the PO algorithm was signicantly higher than that using the IV algorithm (P 0.05).

Discussion
Fig. 6. Plasma insulin responses and simulation curve against pulsatile insulin injection (0.1 U/kg, n 10). Plasma insulin concentrations were measured as described in the text. Circles indicate mean plasma insulin concentrations with intraportal insulin injection. The solid line indicates a simulation curve from a model using intraportal insulin injection

to that with the IV algorithm (PO; 536.6 71.6 versus IV; 546.8 59.7 mU/kg, P 0.74) (Fig. 7B). The peripheral venous plasma insulin concentration with the PO or IV algorithm increased gradually to a peak level of 305.1 68.9 or 468.1 66.9 pmol/l at 60 min and then decreased gradually to 46.4 9.6 or 50.1 5.8 pmol/l at 240 min, respectively (Fig. 7C). The arterial insulin concentration with the PO or IV algorithm also increased gradually to a peak level of 305.3 62.9 or 469.6 85.1 pmol/l at 60 min, respectively (Fig. 7D). The portal insulin concentration with the PO or IV algorithm increased gradually to a peak level of 619.9 101.7 or 414.3 79.9 pmol/l at 60 min, respectively (Fig. 7E). Peripheral venous and arterial plasma insulin concentrations with the PO algorithm were signicantly lower than

In previous studies, we developed closed-loop intravenous and subcutaneous insulin infusion algorithms for the AEP and assessed their feasibility for glycemic control.8,9 We developed each algorithm by simulating biphasic insulin secretion in response to a step-wise glucose infusion in perifusion experiments using pancreatic islets. This biphasic insulin response was successfully simulated by using a transfer function with a rst-order delay for both proportional and derivative actions to blood glucose concentrations.22 Excellent glycemic control was achieved in diabetes patients by using an AEP operated by the closed-loop intravenous or subcutaneous insulin algorithm.8,9 However, each method delivered insulin directly into the systemic circulation, which resulted in relatively low insulin concentrations in the portal vein and high insulin concentrations in peripheral veins, compared with those obtained using a direct portal delivery system. Because the liver responds more quickly to changes in insulin concentration in the portal vein than to those in peripheral arteries,27 the insulin delivery route can

97

Fig. 7. Peripheral venous blood glucose concentrations (A), insulin infusion rates (B), peripheral venous (C), arterial (D), and portal (E) plasma insulin concentrations, and NHGB (F) after a 2-g/kg oral glucose load. Glycemic control was performed in 42-h-fasted dogs under pancreatic clamp conditions using a bedside-type articial endocrine pancreas operated by the closed-loop intraportal (PO) or intravenous

(IV) algorithm (n 5). Blood glucose and plasma insulin concentrations were measured and NHGB was calculated. The solid line and broken line indicate results using the PO and IV algorithms, respectively. Data are mean SD. * P 0.05 for the PO algorithm versus the IV algorithm

Table 1. Results of glycemic control after 2-g/kg oral glucose load in ve 42-h-fasted dogs under pancreatic clamp conditions using the intraportal (PO) or intravenous (IV) algorithm BG Max (mmol/l) PO algorithm IV algorithm 9.69 10.11 1.24 1.63 IRI (pmol/l) Peripheral venous 1144.9 1590.1 261.4* 339.3 Arterial 1166.9 1701.0 290.7* 402.2 Portal 2226.1 1465.2 430.8* 293.1

Data are mean SD values of ve dogs in each group BG Max, maximum level of blood glucose concentration; IRI, summation of plasma insulin levels after 0, 20, 40, 60, 90, 120, 180, and 240 min * P 0.05, compared with the IV algorithm

affect hepatic glucose handling. In this study, to establish the ideal insulin delivery route for the AEP, we investigated the effects of closed-loop portal insulin delivery on glycemic control, insulin proles, and hepatic glucose handling regulated by both hepatic glucose production (HGP) and hepatic glucose uptake (HGU). To evaluate the appropriateness of the PO algorithm, we performed glycemic control after an oral glucose load. We

chose the oral glucose challenge rather than the regular meal challenge because it imposes a higher load on the glucose regulation system. In fact, in our previous studies, we showed that closed-loop insulin delivery, which can achieve glycemic regulation after an oral glucose load, can also regulate postprandial glycemia after a regular meal load.9,21 For glycemic control after an oral glucose load, blood glucose concentrations and insulin requirements

98

were similar for the PO and IV algorithms, whereas peripheral plasma insulin concentrations for the PO algorithm were signicantly lower than those for the IV algorithm. These results corresponded with the report of Canavan et al.,12 and indicate that fasting peripheral insulin levels after peripheral insulin delivery were signicantly higher than those after portal insulin delivery using an open-loop insulin infusion device. Furthermore, other investigators reported similar results after human pancreas transplantation. These studies showed that portal venous drainage of pancreas grafts eliminated peripheral hyperinsulinemia associated with systemic drainage via the vena cava or common iliac vein.16,17 The recent results of meta-analyses such as the Quebec Cardiovascular Study and the Helsinki Policeman Study have indicated that hyperinsulinemia is a strong predictor of future coronary artery disease and stroke events.2830 Moreover, the Diabetes Prevention Program Group reported a signicant association between fasting insulin concentration and blood pressure.31 Since hyperinsulinemia may be involved in some complications related to diabetes, glycemic control without hyperinsulinemia would be ideal. In normal dogs (n 5), we conrmed that the peak values of both peripheral venous and portal insulin concentrations after a 2-g/kg oral glucose load were 298.1 78.1 pmol/l and 627.2 148.7 pmol/l, respectively (unpublished data). In dogs with an AEP operating under the PO or IV algorithm, the peak values of both peripheral venous and portal insulin concentrations were 305.1 68.9 and 619.9 101.7 pmol/l for (the PO algorithm) and 468.1 66.9 and 414.3 79.9 pmol/l for the IV algorithm, respectively, during glycemic control after an oral glucose load under somatostatin infusion. There were no signicant differences in either peripheral venous or portal insulin concentrations between the normal dogs and dogs controlled with the PO algorithm, whereas peripheral hyperinsulinemia and portal hypoinsulinemia were evident with the IV algorithm. Thus, these ndings conrmed that insulin infusion with the PO algorithm reproduces physiological peripheral and portal insulin proles, which indicates that it may reduce the long-term complications of diabetes by avoiding peripheral hyperinsulinemia. The homeostasis of glucose metabolism fails in diabetes patients not only because of a reduction of glucose utilization in peripheral tissue but also because of insufcient suppression of HGP and insufcient augmentation of HGU after meals. It was reported that almost one-third of the glucose load was retained by the liver after oral glucose intake in normal subjects. Kawamori32 demonstrated that if rst-pass HGU is only 15% of the total glucose delivered to the liver, then post-meal hyperglycemia would go up to 16.7 mmol/l. HGU is dependent on three factors:32,33 the glucose load reaching the liver; the insulin concentration within the liver sinusoids; and signals generated by portal glucose delivery, called the portal signal. Pagliassotti et al.25 reported that the portal signal, rather than sinusoidal hyperinsulinemia caused by portal insulin delivery, rapidly activated NHGU. However, portal insulin delivery may activate NHGU more rapidly than peripheral venous insulin delivery because the PI group reached the maximum level

of NHGU earlier than the VI group in our study (Fig. 4). Furthermore, because the contributions of the portal vein and hepatic artery to total hepatic blood ow are almost 75% and 25%, respectively, portal insulin delivery is superior to any other insulin delivery route in increasing the liver sinusoidal insulin level and NHGU. HGP is suppressed by both the direct and indirect actions of insulin.27 The direct action can inhibit glycogenolysis and gluconeogenesis by directly controlling some key regulatory enzymes in the liver, and the indirect action decreases the supply of free fatty acids and glycerol from peripheral tissues to the liver and reduces the output of gluconeogenic substrate amino acids from muscle. Regarding the effects of insulin delivery route on HGP, some reports have indicated that the suppression of HGP is superior with the peripheral route,1315 whereas another report showed similar effects in both the peripheral and portal routes.10 Above all, Lewis et al.13,15 demonstrated that HGP was suppressed to a greater extent with the peripheral insulin infusion than portal insulin infusion in nondiabetic humans and type 2 diabetes patients because of an increase in the indirect effect that decreases plasma gluconeogenic precursor levels. However, it may be difcult to prove this mechanism when the insulin sensitivity of adipose tissue is lower and the suppression of lipolysis is not sufcient. Some reports12,27 have indicated that peripheral insulin delivery did not achieve complete normalization of glucose metabolism because the immediate suppression of HGP reected a rapid response of the liver to sinusoidal insulin level. The sinusoidal insulin level in portal insulin delivery was higher than that in peripheral venous insulin delivery. In this study, suppression of HGP with portal insulin delivery was considered to be greater than that with peripheral venous insulin delivery because the magnitude of NHGU after portal insulin delivery was signicantly greater than that after peripheral venous insulin delivery. Therefore, this nding indicates that closed-loop intraportal insulin delivery enhances the effect of insulin on the liver and improves glucose metabolism by reproducing physiological insulin proles in the peripheral and portal veins. Further studies by means of a primed tracer infusion are planned to investigate the total rate of glucose production. Insulin administered in the peritoneal cavity is dominantly absorbed by the portal circulation.3437 There are several potential advantages to intraperitoneal insulin therapy. This therapy reduces peripheral hyperinsulinemia and hypoglycemic events, and results in metabolic efcacy and improvement of glycemic stability compared with peripheral insulin therapy.3437 In addition, intraperitoneal insulin delivery is superior to intraportal insulin therapy in terms of management and safety. There are several reports on the long-term application of intraperitoneal insulin therapy to diabetes patients.37,38 On the other hand, in attempting to control glucose levels in diabetic dogs by using an AEP operated by the closed-loop intraperitoneal insulin infusion algorithm, physiological peripheral insulin levels were not achieved.21 With regard to the metabolic efcacy and insulin requirement, intraportal insulin therapy is expected to be

99

more effective than intraperitoneal insulin therapy. However, because the technique of placing an insulin catheter into the portal vein in humans is associated with severe invasion and high risks, such as infection and catheter thrombosis, there are few reports of applying intraportal insulin therapy to diabetes patients. Recently there have been several reports of new methods that overcome these problems. Liang et al.39 reported that percutaneous transsplenic portal venous catheterization under ultrasonic guidance could be safely performed in humans. Moreover, Shishko et al.40 reported that umbilical venous catheterization was performed in diabetes patients. They demonstrated that intraportal insulin administration via the umbilical vein reduced levels of three carbon precursors (lactate, pyruvate, and alanine, which serve predominantly as substrates for gluconeogenesis); cortisol; and growth hormone more signicantly than did subcutaneous insulin administration.40 Thus, with technical improvement, it should be possible to safely infuse insulin intraportally. In conclusion, the present study suggested that although closed-loop portal and peripheral venous insulin delivery systems were equally effective in terms of blood glucose control and insulin requirements, portal insulin delivery was superior to peripheral delivery in maintaining more appropriate hepatic glucose handling and physiological insulin proles. These results indicate that the portal vein is the most suitable insulin delivery route for the AEP. Extrapolation of the present results to humans suggests that delivery of insulin into the portal vein can achieve optimal glucose regulation and complete amelioration of impaired metabolism in patients with diabetes mellitus.
Acknowledgments This work was supported by a research grant from the New Energy and Industrial Technology Development Organization (NEDO) of Japan as part of a project on an implantable insulin infusion system utilizing optical blood glucose monitoring, and from the Ministry of Education, Culture, Sports, Science and Technology of Japan (Grant-in-Aid for Development of Highly Advanced Medical Technology to Dr. E. Araki et al. and Scientic Research Grant 15500318 to Dr. K. Nishida et al.). We thank Dr. Shousuke Satake from the National Chubu Hospital for excellent technical suggestions. We also thank Panapharm Laboratories Co., Ltd. for laboratory assistance.

5.

6. 7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

References
1. The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329:977986 2. Ohkubo Y, Kishikawa H, Araki E, Miyata T, Isami S, Motoyoshi S, Kojima Y, Furuyoshi N, Shichiri M. Intensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study. Diabetes Res Clin Pract 1995;28:103117 3. United Kingdom Prospective Diabetes Study Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 1998;352:837853 4. Wake N, Hisashige A, Katayama T, Kishikawa H, Ohkubo Y, Sakai M, Araki E, Shichiri M. Cost-effectiveness of intensive

18.

19.

20. 21.

22.

insulin therapy for type 2 diabetes. a 10-year follow-up of the Kumamoto study. Diabetes Res Clin Pract 2000;48:201210 Nishikawa T, Sasahara T, Kiritoshi S, Sonoda K, Senokuchi T, Matsuo T, Kukidome D, Wake N, Matsumura N, Sakakida M, Kishikawa H, Araki E. Evaluation of urinary 8-hydroxydeoxyguanosine as a novel biomarker of macrovascular complications in type 2 diabetes. Diabetes Care 2003;26:15071512 Shichiri M, Kawamori R, Yamasaki Y, Hakui N, Abe H. Wearable articial endocrine pancrease with needle-type glucose sensor. Lancet 1982;2:11291131 Nishida K, Sakakida M, Ichinose K, Uemura T, Uehara M, Kajiwara K, Miyata T, Shichiri M, Ishihara K, Nakabayashi N. Development of a ferrocene-mediated needle-type glucose sensor covered with newly designed biocompatible membrane, 2methacryloyloxyethyl phosphorylcholine-co-n-butyl methacrylate. Med Prog Technol 1995;21:91103 Shichiri M, Kawamori R. Optimized algorithms for closed-loop glycemic control systems. In: Albisser JBM, Schrezenmeir J, Lehmann L (eds) Computer systems for insulin adjustment in diabetes mellitus. Hedingen, Switzerland: Pancscienta, 1985;171183 Shimoda S, Nishida K, Sakakida M, Konno Y, Ichinose K, Uehara M, Nowak T, Shichiri M. Closed-loop subcutaneous insulin infusion algorithm with a short-acting insulin analog for long-term clinical application of a wearable articial endocrine pancreas. Front Med Biol Eng 1997;8:197211 Ishida T, Chap Z, Chou J, Lewis RM, Hartley CJ, Entman ML, Field JB. Effects of portal and peripheral venous insulin infusion on glucose production and utilization in depancreatized, conscious dogs. Diabetes 1984;33:984990 Kubota M, Yamasaki Y, Sekiya M, Morishima T, Kishimoto M, Kawamori R, Shichiri M, Kamada T. Portal insulin delivery is superior to peripheral delivery in handling of portally delivered glucose. Metabolism 1996;45:150154 Canavan JP, Flecknell PA, New JP, Alberti KG, Home PD. The effect of portal and peripheral insulin delivery on carbohydrate and lipid metabolism in a miniature pig model of human IDDM. Diabetologia 1997;40:11251134 Lewis GF, Zinman B, Groenewoud Y, Vranic M, Giacca A. Hepatic glucose production is regulated both by direct hepatic and extrahepatic effects of insulin in humans. Diabetes 1996;45:454 462 Giacca A, Fisher SJ, McCall RH, Shi ZQ, Vranic M. Direct and indirect effects of insulin in suppressing glucose production in depancreatized dogs: role of glucagon. Endocrinology 1997;138: 9991007 Lewis GF, Carpentier A, Vranic M, Giacca A. Resistance to insulins acute direct hepatic effect in suppressing steady-state glucose production in individuals with type 2 diabetes. Diabetes 1999;48:570576 Gaber AO, Shokouh-Amiri MH, Hathaway DK, Hammontree L, Kitabchi AE, Gaber LW, Saad MF, Britt LG. Results of pancreas transplantation with portal venous and enteric drainage. Ann Surg 1995;221:613622 Bartlett ST, Kuo PC, Johnson LB, Lim JW, Schweitzer EJ. Pancreas transplantation at the University of Maryland. Clin Transplant 1996;271280 Martin X, Petruzzo P, Dawahra M, Feitosa Tajra LC, Da Silva M, Pibiri L, Chapuis F, Dubernard JM, Lefrancois N. Effects of portal versus systemic venous drainage in kidneypancreas recipients. Transpl Int 2000;13:6468 Mason TM, Chan B, El-Bahrani B, Goh T, Gupta N, Gamble J, Qing Shi Z, Prentki M, Steiner G, Giacca A. The effect of chronic insulin delivery via the intraperitoneal versus the subcutaneous route on hepatic triglyceride secretion rate in streptozotocin diabetic rats. Atherosclerosis 2002;161:345352 Shoemaker W, Walker W, Van Italie T, Moore F. A method for simultaneous catheterization of major hepatic vessels in a chronic canine preparation. Am J Physiol 1948;196:311314 Matsuo Y, Shimoda S, Sakakida M, Nishida K, Sekigami T, Ichimori S, Ichinose K, Shichiri M, Araki E. Strict glycemic control in diabetic dogs with a closed-loop intraperitoneal insulin infusion algorithm designed for an implantable articial endocrine pancreas. J Artif Organs 2003;6:5563 Nomura M, Shichiri M, Kawamori R, Yamasaki Y, Iwama N, Abe H. A mathematical insulin-secretion model and its validation in

100 isolated rat pancreatic islets perifusion. Comput Biomed Res 1984;17:570579 Beck P, Hales CN. Immunoassay of serum polypeptide hormones by using 125I-labeled anti-(Immunoglobulin G) antibodies. Biochem J 1975;145:607616 Leevy CM, Mendenhall CL, Lesko W, Howard MM. Estimation of hepatic blood ow with indocyanine green. J Clin Invest 1962;41: 11691179 Pagliassotti MJ, Holste LC, Moore MC, Neal DW, Cherrington AD. Comparison of the time courses of insulin and the portal signal on hepatic glucose and glycogen metabolism in the conscious dog. J Clin Invest 1996;97:8191 Satake S, Moore MC, Igawa K, Converse M, Farmer B, Neal DW, Cherrington AD. Direct and indirect effects of insulin on glucose uptake and storage by the liver. Diabetes 2002;51:1663 1671 Sindelar DK, Balcom JH, Chu CA, Neal DW, Cherrington AD. A comparison of the effects of selective increases in peripheral or portal insulin on hepatic glucose production in the conscious dog. Diabetes 1996;45:15941604 Despres JP, Lamarche B, Mauriege P, Cantin B, Dagenais GR, Moorjani S, Lupien PJ. Hyperinsulinemia as an independent risk factor for ischemic heart disease. N Engl J Med 1996;334:952957 Pyorala M, Miettinen H, Laakso M, Pyorala K. Hyperinsulinemia predicts coronary heart disease risk in healthy middle-aged men: the 22-year follow-up results of the Helsinki Policeman Study. Circulation 1998;98:398404 Pyorala M, Miettinen H, Laakso M, Pyorala K. Hyperinsulinemia and the risk of stroke in healthy middle-aged men: the 22-year follow-up results of the Helsinki Policeman Study. Stroke 1998;29:18601866 31. The Diabetes Prevention Program Research Group. Hypertension, insulin, and proinsulin in participants with impaired glucose tolerance. Hypertension 2002;40:679686 32. Kawamori R. The role of the liver on post-prandial glycaemia emphasis on hepatic glucose uptake. Int J Clin Pract (Suppl) 2000;1922 33. Cherrington AD. Banting Lecture 1997. Control of glucose uptake and release by the liver in vivo. Diabetes 1999;48:1198 1214 34. Selam JL, Bergman RN, Raccah D, Jean-Didier N, Lozano J, Charles MA. Determination of portal insulin absorption from peritoneum via novel nonisotopic method. Diabetes 1990;39: 13611365 35. Thompson JS, Duckworth WC. Insulin pumps and glucose regulation. World J Surg 2001;25:523526 36. Schade DS, Eaton RP. Insulin delivery: how, when, and where. N Engl J Med 1985;312:11201121 37. Dunn FL, Nathan DM, Scavini M Selam JL, Wingrove TG. Longterm therapy of IDDM with an implantable insulin pump. The Implantable Insulin Pump Trial Study Group. Diabetes Care 1997;20:5963 38. Selam JL. External and implantable insulin pumps: current place in the treatment of diabetes. Exp Clin Endocrinol Diabetes 2001; 109(Suppl 2):S333340 39. Liang HL, Yang CF, Pan HB, Chen CK, Chang JM. Percutaneous transsplenic catheterization of the portal venous system. Acta Radiol 1997;38:292295 40. Shishko PI, Kovalev PA, Goncharov VG, Zajarny IU. Comparison of peripheral and portal (via the umbilical vein) routes of insulin infusion in IDDM patients. Diabetes 1992;41:1042 1049

23. 24. 25.

26.

27.

28. 29.

30.