An Artificial Approach to a Very Real Problem: Creating a Pancreas to Treat Type I Diabetes

Nimit Jain
By Nimit Jain February 25, 2010 05:30

Left: The insulin pump (superimposed) has improved treatment for diabetics over the past three decades. However, researchers are now setting their sights higher with the artificial pancreas.

The insulin pump, developed at Yale in the 1970s, revolutionized the management of Type 1 Diabetes. This autoimmune condition occurs when the pancreas fails to produce sufficient amounts of insulin, a hormone necessary for the metabolism of glucose, which is the body’s main energy source.

By substituting the painful injections of insulin with the click of a button, the pump allowed greater flexibility and control because it could be programmed to accom­modate the body’s insulin needs at different times of the day.

A 2002 Yale study of the latest devices found that the pump had cut episodes of hypoglycemia—or low blood sugar—by half. Hypoglycemia is a complication of diabetes that, if severe enough, can poten­tially lead to seizure and coma.

However, manual adjustment of insulin can never be exact. As a result, the fear of hypoglycemia and other long term com­plications associated with diabetes, such as blindness, heart disease and kidney failure, persists.

In response to these fears, researchers are designing an artificial pancreas that aims to imitate the organ’s function of keeping blood sugar levels in a safe range. The pan­creas does this naturally by providing the right amount of insulin at the right time. The medical equipment approach to the artificial pancreas has been the subject of pioneering research by Dr. Stuart Weinz­imer, Associate Professor of Pediatrics at the Yale Medical School.

A common model for the medical equipment approach consists of a glucose sensor, an algorithm for determining insulin dosage, and an insulin pump. The sensor, located under the skin, continuously detects glucose levels in the interstitial fluid between body cells and converts these to plasma glucose levels.

An algorithm processes the sensor sig­nals to determine insulin secretion. The pump component injects the requisite insulin into the subcutaneous tissue. Prog­ress in this field can best be assessed by understanding the characteristics of β-cells (pancreatic cells that secrete insulin) and by contrasting them with present-day glucose sensors and controller algorithms.

β-cell Basics

The mechanism of β-cell sensing essen­tially involves determining glucose concen­tration on the basis of solute movement and diffusion from the blood. However, β-cells may adjust secretion on the basis of antecedent hyperglycemia or high blood sugar, levels of free-fatty acids in the body and plasma insulin levels.

Therefore, the cells must have a sens­ing mechanism for all of these. It has also been found that β-cells secrete insulin in response to neural signals prior to meals and in response to gut hormones during meals. Insulin infusion changes concen­tration of intermediate metabolites, and β-cells have a mechanism to sense the levels of these compounds. Insulin is then delivered in two phases.

Gauging Glucose: A Need for Sensors

Considering that glucose can be mea­sured in more ways than all other blood constituents except pH, it may come as a surprise that we still do not have a perma­nent and fully functional glucose sensor. The different sensing mechanisms are based on various properties of glucose such as oxidation, optical rotation and glycosylation.

The most common sensors use an enzyme-based electrochemical approach, which uses the oxidation of glucose by the enzyme glucose oxidase. Consumption of oxygen, decrease in pH, and generation of hydrogen are three measurements that can be correlated electrochemically with the blood glucose concentration.

The complexity behind developing good sensors can be seen in the fact that many of the problems that were present as early as in 1990 remain unresolved. Some of these problems include noise elimination, longevity, safety of immobilized enzyme and electrodes, immunological response of the bio-fluids to the sensors, site of insertion of the sensors, noninvasiveness for measurement, development of light and small systems, sensing of the intermediate metabolites involved in glucose metabo­lism, and cost.

An additional challenge is posed by the production of a large number of data points, which offsets the accuracy of the sensors to some extent.

As Weinzimer explained, “Present day sensors have an error of about 11-13% as compared to 6% in a standard blood gluco-meter (used presently by diabetics to monitor their blood glucose levels in combination with insulin pump therapy or injections). However, the number of insulin units that depend on a sensor value is much lower than that depending on a gluco-meter reading.”

Development of an efficient glucose sensor would be a huge break-through in a fully operative closed loop system and Yale scientists and engineers are up to the task. Yale is one of the five centers for a project sponsored by the National Institute of Child Health and Human Development that focuses on designing these better glu­cose sensing systems.

Imitation via Algorithm: the PIDs

Models which describe rates of changes of plasma glucose, interstitial insulin and plasma insulin levels are used to develop and test possible controller algorithms. Though many such algorithms have been hypothesized, the Proportional-Integrative-Derivative (PID) Controller algorithm is the tool that has been used in clinical trials of the closed loop system at Yale.

However, the PID suffers some draw­backs, including overestimation of required insulin, absorption delays, system delays and human immune responses. It uses glucose concentration as its only signal for calculating insulin doses.

The delays in absorbing insulin from the subcutaneous tissue to the blood lead to hyperglycemia, causing the PID to secrete more insulin. This hyperinsulinmia affects sensor accuracy and subsequently results in delayed hypoglycemia.

“Clogging and infection at the site of injection and the need for personal fine tuning are also problems associated with this approach,” says Weinzimer.

Figure 1. The PID Algorithm (a) shows the response of the b-cells (circles) and the PID response (solid line) (assump¬tion: one compartment model of insulin clearance). (b), (c) and (d) show the indi¬vidual proportional, integral and derivative responses respectively. (e) shows the com¬bined effect of the three.

Clinical trials offer promise

The Yale-New Haven Children’s Hospital was awarded funding from the Juvenile Diabetes Research Foundation to test the effectiveness of a fully automated closed loop using the PID algorithm, with Weinz­imer as one of the trial leaders.

Seventeen teenagers were tested over the course of three days. Remarkably, 85% of the sensor values were between a normal glucose control range for diabetics, as com­pared to 58% in open loop therapy.

The clinical trials provide hope that insu­lin administration in a closed loop fashion as dictated by the PID may ultimately yield an effective therapy for blood glucose control.

About the Author
NIMIT JAIN is a sophomore in Saybrook College majoring in biomedical engineer­ing. While he loves all aspects of science, he particularly enjoys the application of the scientific method to fields such as economics and linguistics. Nimit currently heads the Engineering Tour Guiding program on campus and is involved in the Saybrook blog and a classic rock society.

Acknowledgements
The author would like to thank Professor Mark Saltzman, Professor Themis Kyria­kides, Dr. Robert Sherwin and Dr. Stuart Weinzimer for their time and help.

Further Reading

  • Weinzimer SA, Steil GM, Swan KL, Dziura J, Kurtz N, Tamborlane WV. “Fully Automated Closed-Loop Insulin Delivery Versus Semiautomated Hybrid Control in Pediatric Patients with Type 1 Diabetes Using an Artificial Pancreas.” Diabetes Care. 2008 May; 31(5): 934-939.
Nimit Jain
By Nimit Jain February 25, 2010 05:30