Feedback of real-time blood glucose data to an insulin pump for bolus control

CGMS feedback to control the insulin pump will work great to control the blood glucose levels when the user is not eating. However at mealtime, the problem becomes very difficult. If the human body were a predictable machine, then the PID control scheme would be fairly straight forward. It's not of course, but that's not the only reason holding back the established research teams from closing the loop with a pure feedback control scheme: the fact is the predictive approach (open loop) works pretty well when everything is "tuned up," so the perception is that we need the artificial pancreas to work better than the predictive approach. For this reason, I believe the algorithm is going to be "semi-automatic," meaning it is automatic when not eating (closed loop) but you must perform a manual bolus (counting carbs and manually delivering the bolus with the insulin pump) at a meal time.

That said, the desire is still there to study an automatic bolus. When eating, the blood sugar goes up, and a control algorithm determines the right amount of insulin to deliver for the meal. An occasional missed bolus seems to be a part of life with insulin pumps. In this case, an inaccurate or "best try" automatic bolus is better than no bolus. After evaluating the bolus for a while, the closed loop scheme can correct the insulin balance anyway.

There are two approaches to consider. The first employs the standard PID controller (i.e. the proportional, integral, and derivative control algorithm) from controls engineering. This allows the control and adjustment of:
 * the rate of glucose increase (i.e the derivative function would deliver more insulin for a rapid increase in blood sugar);
 * the peak of the glucose curve (i.e. the proportional function would deliver more insulin for a higher peak in the blood sugar); and
 * the duration of elevated glucose (i.e. the integral function would deliver more insulin for a long duration of high blood sugar).



By "tuning" the three constants in the PID controller algorithm, the controller can provide control action designed for specific process requirements. The dynamic nature of the response of the PID controller can be characterized by the response to a "step" in blood glucose data. The object is to achieve the optimal balance of quick response without excessive overshoot or oscillation while driving the blood glucose values back to a target value.

The second approach is an "adaptive bolus" or a self-adjusting bolus that also benefits from closed-loop data from a CGMS to control the bolus shape and duration. The adaptive bolus is really a combination of other standard boluses from today’s insulin pump technology, including:


 * the spike bolus, which delivers a larger dose of insulin in a short duration of time similar to a subcutaneous injection. Eating foods with fast-acting sugars such as breakfast cereal is best “covered” with a spike bolus.


 * the extended bolus, or square wave bolus, which delivers a larger dose of insulin spread over a longer period of time, such as an hour or two. The extended bolus is a good choice when eating foods high in protein. The delay in the delivery of carbohydrates from the digestive system when eating and digesting protein can approach the insulin duration-of-action, so in these cases the blood sugar level is better controlled by a slow extended release of insulin that matches the profile of carbs hitting the bloodstream.


 * the combination bolus, which is both a spike and a square wave together, is great for complex foods such as pizza and chocolate cake.
 * the super-bolus, which considers the basal rate delivery of insulin following the bolus as part of the bolus and can be borrowed ahead and given together with the bolus. The super bolus is a simultaneous spike bolus equal to both the insulin to cover a meal and the borrowed basil insulin.



The adaptive bolus is so-named because different parts of the bolus, as controlled by feedback from a CGMS, will adapt to the needs of the user as follows:
 * the magnitude of the spike will be proportional to the spike in the blood sugar of a missed bolus (i.e. the proportional mode of the PID),


 * the delivery rate of the extended bolus will be proportional to the rate of change of the falling blood sugar after the bolus (i.e. the derivative function of the PID),


 * the duration of the extended bolus will be adjusted by the algorithm in response to the blood sugar level relative to the total insulin-on-board until an insulin balance is observed (i.e. the integration mode of the PID).

The automatic bolus may someday be the standard mode of the closed loop artificial pancreas with future implanted systems that eliminate the 15 minute delay between infusing insulin under the skin and the impact that has for lowering the blood sugar value.

See also: What does it REALLY MEAN to close the loop from the CGMS to the insulin pump?