PG30

These pages are best viewed with Netscape version 3.0 or higher or Internet Explorer version 3.0 or higher. When viewed with other browsers, some characters or attributes may not be rendered correctly.

Since the discovery of insulin in 1921, many improvements in the treatment of diabetes mellitus have occurred. Controlled clinical trials have convincingly demonstrated that normalization of blood glucose will prevent or delay the appearance and progression of microvascular complications.

To this end, major efforts are being made to improve the delivery of insulin into patients with diabetes and to increase its ease of administration. An attractive approach to this problem is the use of an implantable insulin pump that is remotely controlled by an external hand-held device. Potential advantages of this approach include the omission of insulin injections, the ability to provide insulin directly into the portal vein, and the improved kinetics of intraperitoneal insulin administration.

With the development of microelectronics in the 1970s, a remotely controlled implantable pump became theoretically feasible. The first totally implantable device for drug delivery was a nonprogrammable, constant flow, gas-powered pump conceived for the continuous infusion of heparin in patients with severe thromboembolic disease. The clinical use of this pump was then extended to the continuous infusion of chemotherapeutic agents into the hepatic artery in the treatment of liver cancer, and to the continuous intraspinal infusion of morphine for the relief of severe chronic pain.

In the earlier 1990s, this pump was used for the continuous intravenous delivery of insulin in patients with non-insulin-dependent diabetes mellitus (NIDDM) with encouraging results. A constant flow pump did not prove ideal in patients with insulin-dependent diabetes mellitus (IDDM), since there still was the need for subcutaneous insulin injections at mealtime.

Subsequently, the joint efforts of academics and industry generated the first two prototypes of an implantable pump for insulin delivery that could be programmed to deliver different basal rates and meal boluses using an external hand-held programmer communicating to the pump via telemetry. These two pumps were successfully used for short-term intraperitoneal insulin treatment of patients with IDDM, reaching normalization of blood glucose profiles and correction of peripheral hyperinsulinemia.

The precipitation of insulin inside these pumps with the progressive decrease of insulin flow delayed any further development in the field until the pharmaceutical company, Hoechst AG, developed a semisynthetic human insulin at neutral pH, which was stabilized with the addition of the surface agent polyethilenpolypropylenglycol (genapol), and which proved to be stable in a pump environment. The availability of a stable insulin preparation led the way to multicenter, large-scale clinical trials to test safety, feasibility, and efficacy of long-term insulin treatment using three different programmable implantable pumps.

IN BRIEF

Implantable insulin pumps with external, hand-held controls are now being tested for safety and effectiveness in large-scale clinical trials. This article explains how these pumps work, summarizes the clinical experience to date, and offers an outlook for the future of this potentially revolutionary form of insulin delivery.

These initial implantations raised many questions. Where should the pump be implanted? Where should the insulin be delivered? Will the insulin be stable at body temperature? Will the electronics function properly for a prolonged period of time in vivo? These were just a few of the issues raised by investigators. The overriding issue, however, was related to safety. Could a pump malfunction release lethal quantities of insulin in the patient? Furthermore, to be economically feasible, an implantable pump had to have not only minimal risk, but also major benefits to patients with diabetes. If a pump meeting these requirements were developed, its integration with an implanted glucose sensor would provide a potential "cure" for diabetes. This article provides a status report on the efforts that many investigators have made during the last two decades toward achieving this goal.

In people without diabetes, insulin is secreted by the pancreatic §-cells into the hepatic portal vein. This route of insulin delivery has several theoretical advantages including 1) a rapid onset of action, 2) direct hepatic insulization, and 3) a reduction of peripheral insulin levels. The benefits of each of these potential advantages may depend upon the specific physiological conditions present when insulin secretion occurs. In any event, investigators have not used the hepatic portal vein directly to deliver insulin from an implantable pump in man. The primary reason to avoid this delivery route is the danger of catheter-induced venous thrombosis and subsequent emboli to the liver, resulting in hepatic portal hypertension.

Two routes of insulin delivery by an implantable pump have been extensively used in humans: the peripheral intravenous and the intraperitoneal. Both routes have specific advantages and disadvantages, which may dictate their choice in a specific patient. The intravenous route is favored when access to the abdomen is limited or if the peritoneal membrane is scarred from recurrent episodes of peritonitis or multiple peritoneal adhesions. The intraperitoneal route has the advantage that some of the insulin is directly absorbed into the hepatic portal vein, and thus, peripheral circulating insulin levels more nearly reflect those present in the nondiabetic state. Both routes of insulin delivery have experienced insulin delivery catheter occlusion, usually from fibrin clots or fibrous tissue at the tip of the catheter.

The goal of insulin delivery is to match circulating insulin to ambient glycemia. Although this is relatively easy in the fasting state when blood glucose changes very slowly, it is difficult in the postprandial state when absorption of foodstuffs results in rapid fluctuations in circulating glucose levels. When there is a major mismatch between glucose and insulin, severe hyperglycemia or hypoglycemia can rapidly occur. Therefore, programmable pumps provide users with a series of insulin delivery algorithms in an attempt to match the rate of insulin delivery to the expected rate of gut-derived glucose absorption.

In addition, there is a diurnal change in insulin sensitivity such that more insulin is required beginning at approximately 6:00 a.m. and continuing through the morning hours. This "dawn phenomenon" may be related to nocturnal growth hormone secretion. Different insulin delivery algorithms used in implantable pumps can automatically infuse more basal insulin during this period to compensate for this increasing need for insulin.

The main features of the three pumps currently undergoing clinical trials are reported in Table 1. Both the MIP 2001 device and the Promedos ID3 device are piston pumps, in which the piston movement aspirates insulin from the reservoir into the piston chamber and then pushes it through the insulin delivery catheter. In these devices, electrical energy is required to operate the piston, piston chamber inlet and outlet valves, and the pump electronics. In contrast, the Infusaid Model 1000 (Fig. 1) device is a gas-powered pump, in which insulin proceeds from the reservoir into a valved accumulator and then into the catheter because of the positive pressure maintained in the reservoir by Freon gas. In this device, electrical energy is required to operate the accumulator inlet and outlet valves and the pump electronics.

Figure 1. The Infusaid implantable pump. The pump's controller is shown on the left. The paitent carries this electronic device to remotely program the implanted pump. The implanted pump on the right is surgically placed under the skin, and the small catheter shown below the pump usually ends in the peritoneal space. The quarter in the lower left-hand area of the figure gives the relative size of the controller and the pump.

The minimum operating life of the pump battery is reported to be 3 years in all devices, with an average pump function of nearly 4 years. The reservoir volume and the concentration of the insulin preparation used determines the average frequency of the refill procedure (every 4-6 weeks in the Infusaid Model 1000 and in the Promedos ID3 device; every 3 months in the MIP 2001 device).

Table 1. Main features of the implantable programmable insulin pumps presently in clinical trails
	INFUSAID Model 1000	MIP 2001	SIEMENS ID3
Manufacturer	Strato/Infusaid Norwood, MA	Minimed Technologies Sylmar, CA	Siemens-Elema Solna, Sweden
Weight (g)/size (cm)	275 g / 9 x 2.7 cm	145 g/ 8 x 2 cm	145 g / 8 x 2.1 cm
Pumping principle	Freon	piston	piston
Stroke volume	1 ą l	0.5 ą l	1 ą l
Reservoir pressure	positive	negative	negative
Battery duration	> 3 years	> 3 years	> 3 years
Reservoir volume	25 ml	15 ml	20 ml
Insulin concentration	100 I.U.	400 I.U.	100 I.U.
Frequency of refilling	4-6 weeks	8-12 weeks	4-6 weeks
Basal rate U/h	0-50	0.13-30	0.3-5
Programmable rates	6	2	6
Bolus	0-99.9 I.U.	0.2-32 I.U.	0.1-20 I.U.
Bolus delivery mode	square wave	instant bolus	square wave
Exchangeable catheter	yes	yes	yes
Sideport	yes	no	yes
Diagnostic procedures	catheter/pump	no	catheter

The Infusaid Model 1000 and the MIP 2001 devices can deliver boluses and basal rates within a wide range of units, covering the need of virtually all insulin-treated patients with diabetes, while the delivery range for basal rates and bolus of the Promedos ID3 device may not be sufficient in patients with high daily insulin requirements (e.g., severe insulin resistance). The feature of programming different basal rates within a 24-hour period is very convenient for patients with a different day/night insulin requirement or in patients with the dawn phenomenon (i.e., a rise in blood glucose in the early morning). The instant delivery of prandial boluses should result in earlier and higher insulin peaks in the peripheral circulation than the square wave delivery, therefore approaching the first phase of insulin secretion in individuals without diabetes. However, no data are available comparing the postprandial control of blood glucose between the two modes of bolus delivery.

All three devices have a silicone rubber polyethylene inner-lined detachable catheter that may be exchanged in case of catheter occlusion. The Infusaid Model 1000 device and the Promedos ID3 feature an access port, i.e., a direct access to the insulin path between the outlet valve of the pumping unit and the catheter hub. Through the access port in the Infusaid Model 1000 device, it is possible to perform minimally invasive procedures to diagnose and manage catheter occlusions and pump low flow because of deposition of insulin aggregates inside the valved accumulator. In the Promedos ID3 device, minimally invasive procedures allow only diagnosis and management of catheter occlusions.

The pumps are implanted in a pocket created in the abdominal wall and secured to the muscular fascia. In case of intraperitoneal delivery, the distal portion (10 cm) of the catheter is inserted in the peritoneal cavity through a 1-2 cm transmuscular incision in the pocket area, and secured by suturing the catheter flange to the fascia. In case of intravenous delivery, the catheter is tunneled subcutaneously in the lateral abdomen, inserted into the left subclavian vein, and secured by suturing the catheter flange to the fascia. Pumps for intravenous insulin delivery may also be implanted in the infraclavicular fossa with the catheter directly inserted in the subclavian vein, even if the aesthetic result is quite poor.

The implantation is performed under either local or general anesthesia. The surgery requires 30-90 minutes, has minimal blood loss, and leaves a linear scar of 10-15 cm. In the postoperative period, the patient is usually required to wear a binder around his or her abdomen, and may experience some pain at the site of implantation. Recovery is usually prompt, and the patient may be discharged from the hospital within 48 hours after surgery. Strenuous physical activity should be avoided for the month following the implantation to reduce the risk of pump pocket complications. Seromas or hematomas of the pump pocket are rare and usually disappear spontaneously within a few weeks.

The pump reservoir has to be refilled every 1-3 months, depending on the patient's daily insulin requirements, the pump reservoir volume, and the concentration of the insulin preparation used (i.e., 100 or 400 IU/ml). All devices require insulin preparations stabilized with the addition of surface agents to minimize the formation of insulin aggregates within the pump. Presently, the only insulin preparation suitable for pump use is a semisynthetic neutral human insulin stabilized with genapol (HOE 21 PH), produced by Hoechst AG (Frankfurt AM, Germany).

After cleansing the skin over the pump, the reservoir is reached through a transcutaneous puncture of the central septum of the pump. The insulin remaining in the reservoir is removed, and the reservoir is refilled with fresh insulin. The refilling procedure may be performed in a physician's office by a trained nurse, requires approximately 20 minutes, and does not need local anesthesia, since the patient feels only minor pain when the skin is punctured.

A hand-held programmer communicating to the pump by telemetry is used to program the basal flow rates and meal boluses. The use of the programmer is simple, and the necessary patient training can be provided by a nurse with the support of audiovisual material.

According to the International Study Group on Implantable Insulin Devices (ISGIID) Registry, more than 780 pumps for long-term insulin delivery have been implanted in patients with diabetes since 1980. A summary of the results of the most recent clinical trials with implantable programmable pumps for insulin delivery in IDDM is reported in Table 2. These studies provide answers to most of the questions raised a decade earlier, after the first pilot implantations in patients with diabetes. The studies were designed as multicenter, long-term, nonrandomized safety, feasibility, and efficacy studies, using adult nonobese patients with long duration IDDM and already on intensive insulin treatment before pump implantation. More than half of the patients had long-term complications of diabetes.

Table 2. Summary of the clinical trials on long-term insulin therapy using implantable programmable pumps
	Diabetes Care 18:388, 1995 and Lancet 343:514, 1994	Diabetes 43:164A, 1994	Int J Artif Organs 18:322, 1995
Patients (n)	224	76	31
Implantable device (n)	Minimed MIP 2001 (205) Infusaid Model 1000 (48) Promedos ID3 (7)	Infusaid Model 1000	Promedos ID3
Route of delivery	IP	IP/IV	IP
Pump therapy (pt/yr)	353	213 IP/ 38 IV	51
Age (yr)	39 ą 9	37 ą 8	41
M/F	1.3:1	1:1	1:0.8
Duration of IDDM (yr)	19 ą 8	15 ą 8	23
BMI	24 ą 3	107 ą 11 (% IBW)	24
Complications of diabetes: retinopathy nephropathy neuropathy	111 24 46	n.a. n.a. n.a.	20 4 17
HbA_1C (%) SC	7.4 ą 1.8 (mean)	7.9 ą 1.5 (mean)	7.5 (median)
IP	6.8 ą 1.0 (mean)	7-7.5 over 3 yr (mean)	7.5 (median)
Ketoacidosis (% pt yr) SC IP (pump)	n.a. 3.1	n.a. 0.4	7.8 1.9
Severe hypo (% pt yr) SC IP (pump) IV (pump)	15.2 2.5 n.a.	33.0 2.0 13	39.2 0.0 n.a.
Pocket Events (% pt yr)	8.5	5.3	3.9
Catheter events (% pt yr) occlusions migration breakage venous thrombosis	13.3 0.8 0.8 n.a.	IP IV 14.5 22.2 1.8 16.7 3.9 5.6 n.a. 11.1	3.9 0.0 2.0 n.a.
Pump failures(% pt/yr)	n.a.	n.a.	5.9
Programmer failures (% pt/yr)	n.a.	n.a.	2.0

In all three studies, a decrease in HBA1c values (below 7.5% n.v. 3.5Đ6.0%) was observed during pump treatment, reflecting a decrease in the mean capillary glucose levels and a decrease in the blood glucose excursions. A major benefit was that the improvement of metabolic control was not associated with an increase in the rate of severe hypoglycemia (coma, seizures, external help for resolution), and/or an increase in body weight.

The rate of ketoacidosis was low, and primarily associated to catheter occlusions. Quality of life was assessed with the diabetes-specific questionnaire developed and validated for the Diabetes Control and Complications Trial (DCCT). Pump implantation did not result in any significant change in the score assessing impact on living and worry. However, scores related to satisfaction with treatment improved.

The most frequent complications of pump treatment are catheter occlusions due to deposition of insulin aggregates inside the lumen, or of fibrous tissue encapsulation or clotting at the tip and inside the distal end of the catheter. The rate of intraperitoneal catheter occlusions requiring surgery (either for replacement or for clearing of adhesions) is similar in devices with and without a sideport. The feature of the access port allows the documentation of obstruction before proceeding with surgery through the measurement of an increased catheter resistance not overcome by flushing the catheter with buffer solution. Intravenous catheters have a shorter life span than intraperitoneal catheters because of a higher rate of occlusions and because of the occurrence of venous thrombosis, which usually requires the removal of the catheter.

Catheter replacement is performed under local or general anesthesia through an incision of the pump pocket. Another frequent complication of pump treatment is the deposition of insulin aggregates within the pump, with a progressive decrease of the insulin outflow in positive pressure pumps and of backflow of insulin from the piston to the reservoir in negative pressure pumps. In the Infusaid Model 1000 pump, the use of a basic solution placed into the reservoir and drained through the access port dissolves the aggregates and restores normal flow rates. Pump pocket complications include seromas, hematomas, skin erosions, infections, pump migrations and local pain. The majority of cases require surgery for correction or pump explantation. Strenuous physical activity has been suggested as a predisposing factor. Failure of the pump, including the programmer, are relatively rare events.

Recent observations have raised the possibility that intraperitoneal insulin therapy may be associated with an increase in the levels of insulin antibodies that in a few patients may be associated with nocturnal hypoglycemia and with an increase in the rate of thyroid autoimmune disease.

Since implantable insulin pumps are not yet commercially available, specific price schedules have not been published. However, based on discussions with some manufacturers, it is anticipated that the selling price for each pump will range between $10,000 and $15,000. These figures do not include the expenses for surgery ($2,000 to $5,000) and for subsequent visits for pump insulin refills. Nonaggregating insulin will also need to be purchased. If complications with the pump or insulin delivery catheter occur, additional surgery and/or medical treatment may be necessary. Although competition between pump manufacturers may force the retail cost of the pump down, the other expenses will remain considerable.

Because of the significant costs involved, the vast majority of patients with diabetes will not be able to afford an implantable insulin pump unless health insurance companies agree to reimburse the majority of the expenses. At the present time, we believe that this policy would be unlikely because of the great initial expense involved. However, if the recent results of the DCCT can be extrapolated to patients with diabetes utilizing implantable insulin pumps, then major monetary savings may be realized by reductions in the incidence and severity of diabetic complications. In countries in which medical care costs are assumed by the state, it is possible that in select patients an implantable insulin pump would also be a viable option to reduce hypoglycemia or improve glucose control.

To date, many patients with diabetes have been implanted with a remotely controlled insulin pump. It is a reasonable conclusion that these pumps are sufficiently reliable to last several years and to provide glucose control that is as good as, if not better than, externally delivered intensive insulin therapy. In addition, the incidence of hypoglycemia may be reduced and improved metabolic indices observed.

Unfortunately, the concurrent development of an implantable or noninvasive glucose sensor has not progressed as rapidly as that observed with implantable pumps. We estimate that several years will be required before these glucose sensors are commercially available. In the meantime, implantable pumps will continue to improve and be used on a limited basis, primarily by those who can afford the cost and who wish to be involved in the latest technical treatment approaches.

Irsigler K, Kritz H, Hagmuller G, Franetzki M, Thurow H, Geisen K: Long-term continuous intraperitoneal insulin infusion with an implanted remote-controlled insulin infusion device. Diabetes 30:1072-75, 1981.

Schade DS, Eaton RP, Sterling WE, Doberneck RC, Spencer WJ, Carlson GA, Bair RE, Love JT, Urenda RS, Gaona JI: A remotely programmable insulin delivery system. Successful short-term implantation in man. JAMA 247:1848-53, 1982.

Point Study Group: One-year trial of a remote-controlled implantable insulin infusion system in type I diabetic patients. Lancet II:866-69, 1988.

Saudek CD, Selam JL, Pitt HA, Waxman K, Rubio M, Jeandidier NJ, Turner D, Fischell RE, Charles MA: A preliminary trial of the programmable implantable medication system for insulin delivery. N Engl J Med 321:574-79, 1989.

Selam JL, Micossi P, Dunn FL, and Nathan DM for the Implantable Insulin Pump Trial Study Group: Clinical trial of programmable implantable insulin pump for type I diabetes. Diabetes Care 15:877-85, 1992.

Dunn FL for the Infusaid Multicenter Implantable Insulin Pump Trial: Beneficial effects of implantable insulin pumps for long-term intensive insulin therapy of IDDM (Abstract). Diabetes 43(Suppl.1):164A, 1994.

Broussole C, Jeandidier N, and Hanaire-Broutin H for the EVIADIAC Study Group: French multicentre experience of implantable insulin pumps. Lancet 343:514-15, 1994.

Hanaire-Broutin H, Broussole C, Jeandidier N, Renard E, Guerci B, Haardt MJ, Lassanann-Vague V and the EVADIAC Group: Feasibility of intraperitoneal insulin therapy with programmable implantable pumps in IDDM. Diabetes Care 18:388-92, 1995.

The Point Study II Group: Multicentre trial of a programmable implantable insulin pump in type I diabetes. Int J Artif Organs 18:322-25, 1995.

Veterans Affairs Study Group: Veterans Affairs Implantable Insulin Pump (IIP) Study (Abstract). Diabetes 43(Suppl.1):61A, 1994.

Marina Scavani, MD, is located at the Istituto Scientifico H San Raffaele in Milano, Italy, and David S. Schade is professor of medicine in the Department of Internal Medicine, Division of Endocrinology at the University of New Mexico in Albuquerque.