**Prevention of Hospital Hypoglycemia by Algorithm Design: A Programming Pathway for Electronic Order Entry**

Susan S. Braithwaite1, Lisa Clark1, Lydia Dacenko-Grawe1, Radha Devi1, Josefina Diaz2, Mehran Javadi1 and Harley Salinas1 *1University of Illinois-Chicago; Saint Francis Hospital, Resurrection Health Care 2University of Illinois-Chicago; Saint Joseph Hospital, Resurrection Health Care United States of America* 

### **1. Introduction**

182 Diabetes – Damages and Treatments

[113] Rosenstock J, Bergenstal R, Defronzo RA, Hirsch IB, Klonoff D, Boss AH, Kramer D,

[114] Tack CJ, Christov V, de Galan BE, Derwahl KM, Klausmann G, Pelikánová T,

[116] Peyrot M, Rubin RR. Effect of technosphere inhaled insulin on quality of life and

[117] Bernstein G. Delivery of insulin to the buccal mucosa utilizing the RapidMist system.

[118] Heinemann L, Jacques Y. Oral insulin and buccal insulin: a critical reappraisal. J

[119] Cernea S, Kidron M, Wohlgelernter J, Raz I. Dose-response relationship of an oral

[122] Pozzilli P, Manfrini S, Costanza F et al. Biokinetics of buccal spray insulin in patients

[123] Guevara-Aguirre J, Guevara M, Saavedra J, Mihic M, Modi P. Oral spray insulin in

[124] Guevara-Aguirre J, Guevara M, Saavedra J, Mihic M, Modi P. Beneficial effects of

single-blind, 5-way crossover study. Clin Ther 2005; 27(10): 1562-1570. [120] Cernea S, Kidron M, Wohlgelernter J, Modi P, Raz I. Dose-response relationship of oral insulin spray in healthy subjects. Diabetes Care 2005; 28(6): 1353-1357. [121] Cernea S, Kidron M, Wohlgelernter J, Modi P, Raz I. Comparison of pharmacokinetic

insulin spray in six patients with type 1 diabetes: a single-center, randomized,

and pharmacodynamic properties of single-dose oral insulin spray and subcutaneous insulin injection in healthy subjects using the euglycemic clamp

treatment of type 2 diabetes: a comparison of efficacy of the oral spray insulin (Oralin) with subcutaneous (SC) insulin injection, a proof of concept study.

addition of oral spray insulin (Oralin) on insulin secretion and metabolic control in subjects with type 2 diabetes mellitus suboptimally controlled on oral

patients with type 2 diabetes. J Diabetes Sci Technol 2008; 2(1): 47-57. [115] Rosenstock J, Lorber DL, Gnudi L, Howard CP, Bilheimer DW, Chang PC, Petrucci RE,

treatment satisfaction. Diabetes Technol Ther 2010; 12(1): 49-55.

trial. Lancet 2010; 375(9733): 2244-2253.

Diabetes Sci Technol 2009;3(3): 568-584.

Expert Opin Drug Deliv 2008; 5(9): 1047-1055.

technique. Clin Ther 2004; 26(12): 2084-2091.

Diabetes Metab Res Rev 2004; 20: 472–478.

with type 1 diabetes. Metabolism 2005; 54: 930–934.

hypoglycemic agents. Diabetes Technol Ther 2004; 6: 1–8.

2177-2182.

Petrucci R, Yu W, Levy B; 0008 Study Group. Efficacy and safety of Technosphere inhaled insulin compared with Technosphere powder placebo in insulin-naive type 2 diabetes suboptimally controlled with oral agents. Diabetes Care 2008; 31(11):

Perusicová J, Boss AH, Amin N, Kramer D, Petrucci R, Yu W; 005 Study Group. Randomized forced titration to different doses of technosphere insulin demonstrates reduction in postprandial glucose excursions and hemoglobin A1c in

Boss AH, Richardson PC. Prandial inhaled insulin plus basal insulin glargine versus twice daily biaspart insulin for type 2 diabetes: a multicentre randomised

> Caregivers treating hospitalized patients are confronted with the necessity both to control hyperglycemia and also to avoid iatrogenic hypoglycemia. Despite controversy about optimal glycemic targets, a large body of evidence associates uncontrolled hyperglycemia with adverse outcomes, both in the intensive care unit and also on general hospital wards (American Diabetes Association, 2011; Moghissi et al., 2009). On general wards, glycemic control during use of scheduled subcutaneous insulin is superior to that seen during use of sliding scale regimens (Baldwin et al., 2005; Umpierrez et al., 2007). When scheduled insulin was compared to sliding scale treatment among general surgical patients, glycemic control was improved (mean blood glucose 145 ± 32 mg/dL vs. 172 ± 47 mg/dL, p < 0.01), and a composite outcome of complications was reduced from 24.3 to 8.6%with odds ratio 3.39 (95% CI 1.50-7.65), p = 0.003 (Umpierrez et al., 2011). Nevertheless, the problem of hypoglycemia is a barrier to successful control of hospital hyperglycemia. Among 1718 adult patients admitted at academic medical centers and having hyperglycemia or receiving insulin therapy, hypoglycemia occurred on 2.8% of all hospital days (Boord et al., 2009). Predisposing factors and adverse outcomes associated with hypoglycemia have been examined in observational studies and in clinical trials studying the effect of glycemic control upon nonglycemic outcomes (Bagshaw et al., 2009; Fischer et al., 1986; Finfer et al., 2009; Kagansky et al., 2003; Krinsley et al., 2007; Maynard et al., 2008; Smith et al., 2005; Stagnaro-Green et al., 1995; Turchin et al., 2009; Van den Berghe et al., 2006; Varghese et al., 2007; Vriesendorp et al., 2006; Wexler et al., 2007). Mortality of patients having myocardial infarction is higher at the lowest as well as the highest ranges glucose, such that the relationship between mortality and glucose is described by a J-shaped curve (Kosiborod et al., 2008). Outcomes of hospitalized patients that have been linked to hypoglycemia include increased ICU mortality or hospital mortality rates, adverse events such as seizures, and increased length of stay. In the intensive care unit and on general wards, associated factors

Prevention of Hospital Hypoglycemia by Algorithm

**2.1 Intravenous insulin infusion** 

**2. Algorithms for glycemic management in the hospital** 

Design: A Programming Pathway for Electronic Order Entry 185

Any algorithm must identify a target blood glucose or a target range and define a safe and effective method for attainment and homeostatic maintenance of target range control. Provisions must be in place regardless of algorithm design to make anticipatory adjustments to prevent hypoglycemia in case of sudden change of any of the usual determinants of insulin requirement, such as carbohydrate exposure or concomitant medications. After making brief reference to an algorithmic method for protection against hypoglycemia during intravenous insulin infusion, the literature on electronic order entry of glycemic management plans is briefly referenced, and idealized examples of programmable

Computerization of intravenous insulin algorithms may be successfully accomplished either through a free standing electronic decision support system or as part of a hospital computer system (Dortch et al, 2008; Hermayer et al. 2007; Junega et al, 2007). The method of computerized order entry that we will use is under construction and will not be presented here, except to say that the algorithms are related to those previously published (Bellam and Braithwaite, 2010; Devi at al, in press 2011). The choice of intravenous insulin protocol depends upon the population treated. The protocols will be designed according to a mathematical rule having population-specific parameters. The protocols are related to column-based tabular protocols in which each column of the table is associated with an assumed maintenance rate of insulin infusion that is thought to be the rate necessary to maintain target range control. Each row of the table represents a range of blood glucose values. The assumed maintenance rate (column assignment) is determined with knowledge of the previous assumed maintenance rate (column assignment) together with the rate of change of blood glucose, at the previous insulin infusion rate. The next insulin infusion rate, at each nursing interaction, depends upon the blood glucose (the row) and the reassigned maintenance rate (column re-assignment, if any). The conservative protocol differs from the standard critical care protocol for intravenous insulin infusion not in the target range blood glucose values, but in the column change rules that result in changing from a lower to a higher maintenance rate (column), or from a higher to a lower assumed maintenance rate (column). That is to say, by analogy with a paper protocol, the column change rules based on rate-of-change of glucose are more conservative under the conservative protocol. We believe that two design features of the intravenous insulin infusion protocols will be shown to be protective against hypoglycemia, namely (1) the column change rules based on rate of change of blood glucose and (2) the near-sigmoidal relationship, at given maintenance rate

branching pathways for patients who are eating and not eating will be presented.

(within-column), between the insulin infusion rate and the blood glucose.

The protocols for diabetic ketoacidosis and hyperglycemic hyperosmolar state each differ from each other and from the critical care protocols for intravenous insulin infusion by having different column change rules and additionally different target ranges for blood glucose. The target range for blood glucose in treatment of diabetic ketoacidosis or hyperglycemia hyperosmolar coma is higher than for other patients likely to be treated with intravenous insulin infusion. An initially fixed-dose weight-based method for assigning insulin infusion rate during the initial hours of treatment is advocated in the consensus statement of the American Diabetes Association for use during the first several hours of treatment of DKA (Kitabchi et al., 2009). In contrast, a dynamic rule to assign the insulin

identified among patients having hypoglycemia include use of bicarbonate-based substitution fluid during continuous venovenous hemofiltration, need for inotropic support, greater severity of illness, co-administration of octreotide with insulin, comorbidities including chronic kidney disease, sepsis, advanced age, history of diabetes or severe diabetes, and history of prior episodes of hypoglycemia. Of special importance, because of the implication for prevention, is the frequency with which the literature on hospital hypoglycemia describes interruption of normal feedings attributable to hospital routine, or episodes of reduced enteral intake without adjustment of insulin therapy, as a factors predisposing to hypoglycemia.

Mechanisms of potential harm from hypoglycemia are partially understood, but the causal relationships between hypoglycemia or iatrogenic hypoglycemia and outcomes is unclear. Concerning the association of hypoglycemia with adverse outcomes, it remains a tenable explanation at least in part that severity of illness may predispose both to adverse outcomes and to hypoglycemia (Kosiborod et al., 2009; van den Berghe et al., 2006). Proof of permanent injury ascribed to a hypoglycemic episode in the hospital setting sometimes is available, but the case ascertainment rate is low. Although proof of direct harm from identifiable hypoglycemic events within large studies may be not discerned by statistical analysis, yet the harm is uniquely damaging to the individual suffering the hypoglycemic event, so that the reporting of isolated cases remains important (Bhatia et al., 2006; Scalea et al., 2007). Life-changing morbidity or mortality may result from a severe hypoglycemic reaction. It is also suspected that some harms may result from hypoglycemia that are not directly traceable to immediate consequences of a specific hypoglycemic event. Since hospital patients will not be randomized to hypoglycemia or non-hypoglycemia, our understanding of the causes and clinical impact of hypoglycemia will be observational, resulting from analyses of hypoglycemia as a secondary outcome, within trials of therapeutic strategies or interventions aiming at targets other than hypoglycemia, or resulting from analysis of hypoglycemia within cohort studies. For the present, the association of hypoglycemia with adverse outcomes justifies development of strategies for prevention of hypoglycemia in the hospital.

The goal of this chapter is to describe attributes of a programming pathway for computerized order entry that may incorporate the best elements of paper protocols for subcutaneous insulin and that may help prevent hospital hypoglycemia. In designing treatment for the hyperglycemic patient, it is necessary to anticipate events that create risk for hypoglycemia and to meet those events with appropriate revisions of nutritional therapy and scheduled insulin. When insulin orders are in place but patient risk for hypoglycemia is predicted to increase, the components of insulin therapy that might be withheld or reduced may differ, depending upon co-morbidities, anticipated disruption of carbohydrate exposure, alteration of other medical therapies, and classification of diabetes. By juxtaposing elements of care within a checklist of orders, paper protocols present reminders to the prescriber about strategies for hypoglycemia prevention. We believe the main opportunity for improvement within computerized order entry systems is the need to present several different packages of orders at the user interface that match differing patterns of carbohydrate exposure. For each pattern of carbohydrate exposure, the package must include default and acceptable alternative orders that encompass monitoring of blood glucose, scheduled and correction-dose insulin orders, and menus of additional directions associated with the insulin orders.

### **2. Algorithms for glycemic management in the hospital**

Any algorithm must identify a target blood glucose or a target range and define a safe and effective method for attainment and homeostatic maintenance of target range control. Provisions must be in place regardless of algorithm design to make anticipatory adjustments to prevent hypoglycemia in case of sudden change of any of the usual determinants of insulin requirement, such as carbohydrate exposure or concomitant medications. After making brief reference to an algorithmic method for protection against hypoglycemia during intravenous insulin infusion, the literature on electronic order entry of glycemic management plans is briefly referenced, and idealized examples of programmable branching pathways for patients who are eating and not eating will be presented.

### **2.1 Intravenous insulin infusion**

184 Diabetes – Damages and Treatments

identified among patients having hypoglycemia include use of bicarbonate-based substitution fluid during continuous venovenous hemofiltration, need for inotropic support, greater severity of illness, co-administration of octreotide with insulin, comorbidities including chronic kidney disease, sepsis, advanced age, history of diabetes or severe diabetes, and history of prior episodes of hypoglycemia. Of special importance, because of the implication for prevention, is the frequency with which the literature on hospital hypoglycemia describes interruption of normal feedings attributable to hospital routine, or episodes of reduced enteral intake without adjustment of insulin therapy, as a factors

Mechanisms of potential harm from hypoglycemia are partially understood, but the causal relationships between hypoglycemia or iatrogenic hypoglycemia and outcomes is unclear. Concerning the association of hypoglycemia with adverse outcomes, it remains a tenable explanation at least in part that severity of illness may predispose both to adverse outcomes and to hypoglycemia (Kosiborod et al., 2009; van den Berghe et al., 2006). Proof of permanent injury ascribed to a hypoglycemic episode in the hospital setting sometimes is available, but the case ascertainment rate is low. Although proof of direct harm from identifiable hypoglycemic events within large studies may be not discerned by statistical analysis, yet the harm is uniquely damaging to the individual suffering the hypoglycemic event, so that the reporting of isolated cases remains important (Bhatia et al., 2006; Scalea et al., 2007). Life-changing morbidity or mortality may result from a severe hypoglycemic reaction. It is also suspected that some harms may result from hypoglycemia that are not directly traceable to immediate consequences of a specific hypoglycemic event. Since hospital patients will not be randomized to hypoglycemia or non-hypoglycemia, our understanding of the causes and clinical impact of hypoglycemia will be observational, resulting from analyses of hypoglycemia as a secondary outcome, within trials of therapeutic strategies or interventions aiming at targets other than hypoglycemia, or resulting from analysis of hypoglycemia within cohort studies. For the present, the association of hypoglycemia with adverse outcomes justifies development of strategies for

The goal of this chapter is to describe attributes of a programming pathway for computerized order entry that may incorporate the best elements of paper protocols for subcutaneous insulin and that may help prevent hospital hypoglycemia. In designing treatment for the hyperglycemic patient, it is necessary to anticipate events that create risk for hypoglycemia and to meet those events with appropriate revisions of nutritional therapy and scheduled insulin. When insulin orders are in place but patient risk for hypoglycemia is predicted to increase, the components of insulin therapy that might be withheld or reduced may differ, depending upon co-morbidities, anticipated disruption of carbohydrate exposure, alteration of other medical therapies, and classification of diabetes. By juxtaposing elements of care within a checklist of orders, paper protocols present reminders to the prescriber about strategies for hypoglycemia prevention. We believe the main opportunity for improvement within computerized order entry systems is the need to present several different packages of orders at the user interface that match differing patterns of carbohydrate exposure. For each pattern of carbohydrate exposure, the package must include default and acceptable alternative orders that encompass monitoring of blood glucose, scheduled and correction-dose insulin orders, and menus of additional directions

predisposing to hypoglycemia.

prevention of hypoglycemia in the hospital.

associated with the insulin orders.

Computerization of intravenous insulin algorithms may be successfully accomplished either through a free standing electronic decision support system or as part of a hospital computer system (Dortch et al, 2008; Hermayer et al. 2007; Junega et al, 2007). The method of computerized order entry that we will use is under construction and will not be presented here, except to say that the algorithms are related to those previously published (Bellam and Braithwaite, 2010; Devi at al, in press 2011). The choice of intravenous insulin protocol depends upon the population treated. The protocols will be designed according to a mathematical rule having population-specific parameters. The protocols are related to column-based tabular protocols in which each column of the table is associated with an assumed maintenance rate of insulin infusion that is thought to be the rate necessary to maintain target range control. Each row of the table represents a range of blood glucose values. The assumed maintenance rate (column assignment) is determined with knowledge of the previous assumed maintenance rate (column assignment) together with the rate of change of blood glucose, at the previous insulin infusion rate. The next insulin infusion rate, at each nursing interaction, depends upon the blood glucose (the row) and the reassigned maintenance rate (column re-assignment, if any). The conservative protocol differs from the standard critical care protocol for intravenous insulin infusion not in the target range blood glucose values, but in the column change rules that result in changing from a lower to a higher maintenance rate (column), or from a higher to a lower assumed maintenance rate (column). That is to say, by analogy with a paper protocol, the column change rules based on rate-of-change of glucose are more conservative under the conservative protocol. We believe that two design features of the intravenous insulin infusion protocols will be shown to be protective against hypoglycemia, namely (1) the column change rules based on rate of change of blood glucose and (2) the near-sigmoidal relationship, at given maintenance rate (within-column), between the insulin infusion rate and the blood glucose.

The protocols for diabetic ketoacidosis and hyperglycemic hyperosmolar state each differ from each other and from the critical care protocols for intravenous insulin infusion by having different column change rules and additionally different target ranges for blood glucose. The target range for blood glucose in treatment of diabetic ketoacidosis or hyperglycemia hyperosmolar coma is higher than for other patients likely to be treated with intravenous insulin infusion. An initially fixed-dose weight-based method for assigning insulin infusion rate during the initial hours of treatment is advocated in the consensus statement of the American Diabetes Association for use during the first several hours of treatment of DKA (Kitabchi et al., 2009). In contrast, a dynamic rule to assign the insulin

Prevention of Hospital Hypoglycemia by Algorithm

receive continuous enteral tube feedings.

management of hospitalized patients.

Design: A Programming Pathway for Electronic Order Entry 187

75% in the intervention group and 71% in the usual care group [adjusted RR 1.36 (1.02- 1.80)]. With the intervention, there were a lower patient-day weighted mean glucose (148 vs 158, p = 0.04); less use of sliding scale (25% vs 58%, p = 0.01); and no difference in hypoglycemia < 40 mg/dL (0.5% vs 0.3%, p = 0.58). Wexler and colleagues at a single site randomized medical teams to availability of an electronic insulin order template versus usual insulin ordering. Intervention group patients (n=65) had mean glucose of 195 +/- 66 mg/dl. Control group patients (n=63) had mean glucose of 224 +/- 57 mg/dl (P=0.004). In

With electronic order entry, there is a risk that some of the integration between the components of care might be lost that had been achievable with paper order sets. Under some electronic systems, juxtaposition of related orders is lost. Users might have to navigate between screens to complete a package of orders relating to diabetes encompassing such necessities as a nutrition plan, point-of-care tests, insulin doses, and a treatment plan for hypoglycemia. A plan for continuous enteral tube feedings might be entered on one screen, followed by insulin orders on another screen, and finally orders for point-of-care glucose monitoring and call parameters on a third screen. Orders that are preselected as the likeliest choice, based on absolute rate of utilization, could be programmed as defaults but might be misapplied to subgroups through user failure to deselect and replace the order for the patient at hand. As an example, if the choice "ACHS" appears at the top of a list of possible orders for glucose monitoring as the default (ante cibum and hora somni, before meals and at bedtime), then an order for ACHS timing could be accepted by default, rather than timing more appropriate to the carbohydrate exposure actually planned for a patient who might

the intervention group, there was no increase in hypoglycemia (Wexler et al., 2010).

**3. Programming pathway for glycemic management in the hospital** 

In the United States, in coming years hospitals will strive to comply with "meaningful use" regulations for electronic health records, described in the Health Information Technology for Economic and Clinical Health Act (Blumenthal & Tavenner, 2010). Electronic order entry will gradually replace handwriting of orders. Some systems will sharply restrict the use of free-text entries, creating necessity for a system that will link orders to preprogrammed comments that may be selected by the user. A template similar to that of electronic order entry might be used to facilitate communication among caregivers at the time of patient transfers and discharge. The remainder of this chapter will describe the design of an idealized proposal for a programming pathway for electronic order entry for glycemic

Within the figures showing the programming pathway, orders that are members of a category or subcategory have the same level of indentation. Pre-assignment of a default choice withincategory sometimes is justified either based on frequency of use or medical indications. The user may select or deselect an order by clicking on a button associated with an order at the user interface. In some cases, selection by a provider of one order results in de-selection of another order within the same subcategory or category. In other cases, choices within-category or within sub-category are not mutually exclusive; selection of one order does not result in deselection of another order (Figure 1). It is envisioned that the user will move through a sequence of those screens within the programming pathway that are determined by having made an early commitment to one branch of the algorithm. When the provider is satisfied that

no modifications are required, the provider enters an electronic signature.

infusion rate during treatment of hyperglycemic crisis is employed at our institutions (Devi et al., 2011). The deactivation time for intravenous insulin infusion may be as long as 90 minutes (Mudaliar et al., 2008). A rationale for a dynamic insulin infusion rate in the early hours of treatment for hyperglycemic emergency is that under conventional management late hypoglycemia sometimes complicates the treatment course.

### **2.2 Algorithms for subcutaneous insulin**

Many protocols for hospital care were developed in the era of handwritten order entry. Order sets were developed that served as a checklist to prevent omissions of elements of care. For example, a reminder to have a standing "prn" order for intravenous dextrose under selected conditions or to order an A1C may be part of the order set. Order sets help integrate the components of care with each other. Timing of testing, insulin, and meals may be coordinated by justaposition of related orders on a paper order set. A lynchpin of successful order writing is the coordination of the patterns of glucose monitoring and insulin administration with carbohydrate exposure (Bellam and Braithwaite, 2010; Braithwaite et al., 2007; Campbell et al., 2004; Thompson et al., 2005). If an order set is well designed, by checking boxes and entering numbers the prescriber creates orders that are familiar to and readily interpreted by pharmacy and nursing staff. Lengthy narrative is reduced. Standardization of order entry protects patient safety. A well designed order set facilitates individualization of patient care. Guidelines may be appended to or embedded within order sets, together with references to supportive medical literature (Donaldson et al. 2006; Hermayer et al, 2009; Lee et al. 2008; Maynard et al., 2009; Schnipper et al., 2010; Trujillo et al., 2008; Wexler et al., 2010). Protocols executed through order sets were thought to reduce medical errors, improve safety, and increase adherence to those guidelines that were supported by medical evidence.

As electronic order entry began to gain widespread use, a body of descriptive studies developed concerning the use of structured order sets for electronic order entry for subcutaneous insulin therapy in the hospital. Hermeyer and colleagues described a comprehensive program, including a web-based calculator for the intravenous insulin protocol (Hermeyer et al., 2009). Maynard and colleagues, in a published study of computerized order entry with paper guidelines used on the side, defined time periods 1, 2 and 3 (TP1, TP2, and TP3) during rolling out of the program. Paper statements of guidelines adjunctive to computerized order entry were developed (Lee et al., 2008). The relative risk (RR) of an uncontrolled patient-stay was reduced from baseline to 0.91 (CI 0.85-0.96) in TP2, and to 0.84 (CI 0.77-0.89) in TP3, with more marked effects in the secondary analysis limited to patients with at least 8 point-of-care glucose values (Maynard et al., 2009). The percent of patient-days with hypoglycemia was 3.8%, 2.9%, and 2.6% in the 3 time periods, representing a RR for hypoglycemic day in TP3:TP1 of 0.68 (CI 0.59-0.78). Similar reductions were seen in risk for hypoglycemic patient-stays.

Evidence from cluster randomized studies supports the use of structured order sets to improve glycemic outcomes. Schnipper and colleagues in several stages developed a computerized version of their order entry system for glycemic control (Schnipper et al., 2009; Schnipper et al., 2010; Trujillo et al., 2008). In a cluster randomized design of 179 patients at a single site, two of the four medical services were chosen randomly to receive the intervention using a computerized order set built into the proprietary computer at Brigham and Women's Hospital. The mean percent of glucose readings between 60-180 was

infusion rate during treatment of hyperglycemic crisis is employed at our institutions (Devi et al., 2011). The deactivation time for intravenous insulin infusion may be as long as 90 minutes (Mudaliar et al., 2008). A rationale for a dynamic insulin infusion rate in the early hours of treatment for hyperglycemic emergency is that under conventional management

Many protocols for hospital care were developed in the era of handwritten order entry. Order sets were developed that served as a checklist to prevent omissions of elements of care. For example, a reminder to have a standing "prn" order for intravenous dextrose under selected conditions or to order an A1C may be part of the order set. Order sets help integrate the components of care with each other. Timing of testing, insulin, and meals may be coordinated by justaposition of related orders on a paper order set. A lynchpin of successful order writing is the coordination of the patterns of glucose monitoring and insulin administration with carbohydrate exposure (Bellam and Braithwaite, 2010; Braithwaite et al., 2007; Campbell et al., 2004; Thompson et al., 2005). If an order set is well designed, by checking boxes and entering numbers the prescriber creates orders that are familiar to and readily interpreted by pharmacy and nursing staff. Lengthy narrative is reduced. Standardization of order entry protects patient safety. A well designed order set facilitates individualization of patient care. Guidelines may be appended to or embedded within order sets, together with references to supportive medical literature (Donaldson et al. 2006; Hermayer et al, 2009; Lee et al. 2008; Maynard et al., 2009; Schnipper et al., 2010; Trujillo et al., 2008; Wexler et al., 2010). Protocols executed through order sets were thought to reduce medical errors, improve safety, and increase adherence to those guidelines that were

As electronic order entry began to gain widespread use, a body of descriptive studies developed concerning the use of structured order sets for electronic order entry for subcutaneous insulin therapy in the hospital. Hermeyer and colleagues described a comprehensive program, including a web-based calculator for the intravenous insulin protocol (Hermeyer et al., 2009). Maynard and colleagues, in a published study of computerized order entry with paper guidelines used on the side, defined time periods 1, 2 and 3 (TP1, TP2, and TP3) during rolling out of the program. Paper statements of guidelines adjunctive to computerized order entry were developed (Lee et al., 2008). The relative risk (RR) of an uncontrolled patient-stay was reduced from baseline to 0.91 (CI 0.85-0.96) in TP2, and to 0.84 (CI 0.77-0.89) in TP3, with more marked effects in the secondary analysis limited to patients with at least 8 point-of-care glucose values (Maynard et al., 2009). The percent of patient-days with hypoglycemia was 3.8%, 2.9%, and 2.6% in the 3 time periods, representing a RR for hypoglycemic day in TP3:TP1 of 0.68 (CI 0.59-0.78). Similar reductions

Evidence from cluster randomized studies supports the use of structured order sets to improve glycemic outcomes. Schnipper and colleagues in several stages developed a computerized version of their order entry system for glycemic control (Schnipper et al., 2009; Schnipper et al., 2010; Trujillo et al., 2008). In a cluster randomized design of 179 patients at a single site, two of the four medical services were chosen randomly to receive the intervention using a computerized order set built into the proprietary computer at Brigham and Women's Hospital. The mean percent of glucose readings between 60-180 was

late hypoglycemia sometimes complicates the treatment course.

**2.2 Algorithms for subcutaneous insulin** 

supported by medical evidence.

were seen in risk for hypoglycemic patient-stays.

75% in the intervention group and 71% in the usual care group [adjusted RR 1.36 (1.02- 1.80)]. With the intervention, there were a lower patient-day weighted mean glucose (148 vs 158, p = 0.04); less use of sliding scale (25% vs 58%, p = 0.01); and no difference in hypoglycemia < 40 mg/dL (0.5% vs 0.3%, p = 0.58). Wexler and colleagues at a single site randomized medical teams to availability of an electronic insulin order template versus usual insulin ordering. Intervention group patients (n=65) had mean glucose of 195 +/- 66 mg/dl. Control group patients (n=63) had mean glucose of 224 +/- 57 mg/dl (P=0.004). In the intervention group, there was no increase in hypoglycemia (Wexler et al., 2010). With electronic order entry, there is a risk that some of the integration between the components of care might be lost that had been achievable with paper order sets. Under some electronic systems, juxtaposition of related orders is lost. Users might have to navigate between screens to complete a package of orders relating to diabetes encompassing such necessities as a nutrition plan, point-of-care tests, insulin doses, and a treatment plan for hypoglycemia. A plan for continuous enteral tube feedings might be entered on one screen, followed by insulin orders on another screen, and finally orders for point-of-care glucose monitoring and call parameters on a third screen. Orders that are preselected as the likeliest choice, based on absolute rate of utilization, could be programmed as defaults but might be misapplied to subgroups through user failure to deselect and replace the order for the patient at hand. As an example, if the choice "ACHS" appears at the top of a list of possible orders for glucose monitoring as the default (ante cibum and hora somni, before meals and at bedtime), then an order for ACHS timing could be accepted by default, rather than timing more appropriate to the carbohydrate exposure actually planned for a patient who might

receive continuous enteral tube feedings.

### **3. Programming pathway for glycemic management in the hospital**

In the United States, in coming years hospitals will strive to comply with "meaningful use" regulations for electronic health records, described in the Health Information Technology for Economic and Clinical Health Act (Blumenthal & Tavenner, 2010). Electronic order entry will gradually replace handwriting of orders. Some systems will sharply restrict the use of free-text entries, creating necessity for a system that will link orders to preprogrammed comments that may be selected by the user. A template similar to that of electronic order entry might be used to facilitate communication among caregivers at the time of patient transfers and discharge. The remainder of this chapter will describe the design of an idealized proposal for a programming pathway for electronic order entry for glycemic management of hospitalized patients.

Within the figures showing the programming pathway, orders that are members of a category or subcategory have the same level of indentation. Pre-assignment of a default choice withincategory sometimes is justified either based on frequency of use or medical indications. The user may select or deselect an order by clicking on a button associated with an order at the user interface. In some cases, selection by a provider of one order results in de-selection of another order within the same subcategory or category. In other cases, choices within-category or within sub-category are not mutually exclusive; selection of one order does not result in deselection of another order (Figure 1). It is envisioned that the user will move through a sequence of those screens within the programming pathway that are determined by having made an early commitment to one branch of the algorithm. When the provider is satisfied that no modifications are required, the provider enters an electronic signature.

Prevention of Hospital Hypoglycemia by Algorithm

care for each patient.

Fig. 2. Opening screen.

(Figure 3).

**3.3 Intravenous insulin algorithm selection under the pathway** 

The programming pathway specifies options for four different intravenous insulin infusion protocols. A full discussion of these pathways is beyond the scope of the present discussion. For critical care patients requiring an intravenous insulin infusion, to help the user decide whether to order the conservative critical care intravenous insulin infusion protocol or the standard one, the user may find a link to a drop-down guideline for indications for the conservative IV insulin protocol. This states that the conservative IV insulin protocol will be appropriate for patients with renal failure, malnutrition, hepatic failure, sepsis, severe congestive heart failure, adrenal insufficiency, and other conditions that the caregiver judges to create high-risk for hypoglycemia. The conservative protocol also is the protocol to which the prescriber might default, in case a patient already has demonstrated hypoglycemia while on the standard protocol but still requires intravenous insulin infusion therapy

An American Diabetes Association consensus statement provides a summary of diagnostic criteria for diabetic ketoacidosis or hyperosmolar hyperglycemic state (Kitabchi et al., 2009). The criteria for each can be summarized in a link to a drop-down guideline for diagnosis,

Design: A Programming Pathway for Electronic Order Entry 189

numbers. The programming pathway accommodates the spectrum of reasonable provider treatment preferences. By offering a menu of treatment alternatives and additional directions to the insulin orders, the programming pathway facilitates individualization of


Fig. 1. Instructions to programmer. Symbols signify the function of buttons at the user interface and define the structure of the program.

#### **3.1 Programming pathway as checklist**

Just as structured paper order sets for subcutaneous insulin therapy may protect the patient from omissions of needed elements of care by presenting a checklist, similarly a checklist of reminders for glycemic management may appear within a branching programming pathway. For example, the main trunk of the branching pathway may call for elements of care that are considered to be potentially universally appropriate, such as a standing "prn" order for concentrated intravenous dextrose for treatment of hypoglycemia, a nutrition consult, or an A1C (Figure 2). The programming pathway that we will present goes on to branch into 8 different treatment plans, each having preventive measures related to hypoglycemia that are specific to the components of the treatment plan, embedded as checklist options for selection, such as "reduce" orders for basal insulin for type 2 diabetes or "hold" parameters for prandial insulin (Figure 3).

#### **3.2 Individualization facilitated under the programming pathway**

It is necessary to specify precautions against hypoglycemia, but manual entry can be burdensome. Under the branches of the pathway, measures for hypoglycemia prevention could take different forms depending upon the carbohydrate exposure of the patient, including but not limited to the scheduling the monitoring of blood glucose, assignment of call parameters at alert levels of glucose, pattern of insulin administration, or use of the classification of hyperglycemia or diabetes to determine "hold" parameters for specific components of insulin therapy. We believe the ordering of these protective additional directions is more likely to occur when a complete menu of options is presented to the prescriber than when reliance is placed upon provider initiative and recall. To a large extent, manual entry of such safety provisions can be replaced by checking boxes and entering

Fig. 1. Instructions to programmer. Symbols signify the function of buttons at the user

Just as structured paper order sets for subcutaneous insulin therapy may protect the patient from omissions of needed elements of care by presenting a checklist, similarly a checklist of reminders for glycemic management may appear within a branching programming pathway. For example, the main trunk of the branching pathway may call for elements of care that are considered to be potentially universally appropriate, such as a standing "prn" order for concentrated intravenous dextrose for treatment of hypoglycemia, a nutrition consult, or an A1C (Figure 2). The programming pathway that we will present goes on to branch into 8 different treatment plans, each having preventive measures related to hypoglycemia that are specific to the components of the treatment plan, embedded as checklist options for selection, such as "reduce" orders for basal insulin for type 2 diabetes

It is necessary to specify precautions against hypoglycemia, but manual entry can be burdensome. Under the branches of the pathway, measures for hypoglycemia prevention could take different forms depending upon the carbohydrate exposure of the patient, including but not limited to the scheduling the monitoring of blood glucose, assignment of call parameters at alert levels of glucose, pattern of insulin administration, or use of the classification of hyperglycemia or diabetes to determine "hold" parameters for specific components of insulin therapy. We believe the ordering of these protective additional directions is more likely to occur when a complete menu of options is presented to the prescriber than when reliance is placed upon provider initiative and recall. To a large extent, manual entry of such safety provisions can be replaced by checking boxes and entering

interface and define the structure of the program.

or "hold" parameters for prandial insulin (Figure 3).

**3.2 Individualization facilitated under the programming pathway** 

**3.1 Programming pathway as checklist** 

numbers. The programming pathway accommodates the spectrum of reasonable provider treatment preferences. By offering a menu of treatment alternatives and additional directions to the insulin orders, the programming pathway facilitates individualization of care for each patient.


Fig. 2. Opening screen.

### **3.3 Intravenous insulin algorithm selection under the pathway**

The programming pathway specifies options for four different intravenous insulin infusion protocols. A full discussion of these pathways is beyond the scope of the present discussion. For critical care patients requiring an intravenous insulin infusion, to help the user decide whether to order the conservative critical care intravenous insulin infusion protocol or the standard one, the user may find a link to a drop-down guideline for indications for the conservative IV insulin protocol. This states that the conservative IV insulin protocol will be appropriate for patients with renal failure, malnutrition, hepatic failure, sepsis, severe congestive heart failure, adrenal insufficiency, and other conditions that the caregiver judges to create high-risk for hypoglycemia. The conservative protocol also is the protocol to which the prescriber might default, in case a patient already has demonstrated hypoglycemia while on the standard protocol but still requires intravenous insulin infusion therapy (Figure 3).

An American Diabetes Association consensus statement provides a summary of diagnostic criteria for diabetic ketoacidosis or hyperosmolar hyperglycemic state (Kitabchi et al., 2009). The criteria for each can be summarized in a link to a drop-down guideline for diagnosis,

Prevention of Hospital Hypoglycemia by Algorithm

appropriate to each subcutaneous pathway.

exposure or omission.

preceding 24 hr.

insulin.

insulin.

carried forward.

treatment.

**orders for subcutaneous insulin** 

**3.5 Subcutaneous insulin dose titration after pathway initiation** 

24 hr to determine total daily dose of insulin actually delivered.

Design: A Programming Pathway for Electronic Order Entry 191

A method for establishing starting doses of insulin is described in sections that will follow. An associated guideline might state that rewriting of the doses of scheduled insulin should be considered daily. Here, a guideline for revision of scheduled insulin is presented that is

• review comorbidities and medications affecting insulin requirement and carbohydrate

• review the medication administration record for confirmation of insulin dosing over the

• add the total amount of scheduled and correction dose insulin delivered in the previous

• *if all blood glucose readings were > 180 mg/dL, add 10%* to the total daily dose actually delivered in the previous 24 hr to determine the new total daily dose of scheduled

• *if any blood glucose was < 80 mg/dL, subtract 20%* from the total daily dose actually delivered in the previous 24 hr to determine the new total daily dose of scheduled

The new dose of scheduled insulin is reapportioned between the components of scheduled therapy. Once the treatment pattern has been entered, changing of dose or correction dose scale can be accomplished outside of the programming pathway. If there are no further specifications, any standing "additional directions" concerning scheduled insulin may be

Therapeutic inertia in changing established insulin regimens is a recognized problem in the care of hospitalized patients. In a study of 52 hospitalized patients treated with 50% dextrose for an episode of hypoglycemia, it was found that subsequent to withholding of insulin at the time of the hypoglycemia, 31% of the patients received no other change in treatment (Garg et al., 2004). The guideline above would give caregivers direction on trouble-shooting of the causes of hypoglycemia and making appropriate revisions of

**3.6 Integration of the components of care under the pathway during placement of** 

A decision support system helps the prescriber to recognize the components of care under the pattern of treatment that is ordered and the relationship between these components. During widespread adoption of computerization of order entry, a distinct computer order is required separately for feedings, intravenous dextrose, glucose monitoring, each component of insulin therapy, treatment of hypoglycemia, and call parameters. The relationship of these elements of care to each other and their timing must be coordinated. The integration of the components of care, achieved by many paper protocols and order sets, must be preserved. The prescriber must be able to accomplish the goals of glycemic control and hypoglycemia prevention without navigation through multiple screens of an electronic order entry system. Whether patients are eating or not, interruption of carbohydrate exposure is a well verified risk factor for hospital hypoglycemia. The risk arises from hospital routine that interrupts feedings or patient factors that result in poor oral intake (Fischer et al., 1986). Restrictions on free-text entries will necessitate preprogramming of additional directions. In each branch of the pathway that will be shown, in case of reduction of carbohydrate exposure, the insulin orders may be accompanied by standardized statements concerning hypoglycemia

accessed from the opening menu (Figure 3). Classification as diabetic ketoacidosis (DKA) is suggested by plasma glucose > 250 mg/dL, arterial pH < 7.3, bicarbonate < 15, anion gap > 12 meq/L, and moderate ketonuria or ketonemia. Classification as hyperglycemic hyperosmolar state (HHS) is suggested by plasma glucose > 600 mg/dl, serum osmolality > 320 mosm/L, arterial pH > 7.3, bicarbonate > 15 meq/L, and minimal ketonuria and ketonemia.

### **3.4 Subcutaneous insulin algorithm selection under the pathway**

For patients who will receive subcutaneous insulin, once the pattern of carbohydrate exposure is determined, then the prescriber can select the appropriate branch of the pathway (Figure 3). Selection of a single branch from the list will launch an appropriate submenu, dependent upon carbohydrate exposure, for the schedule of blood glucose monitoring and the selection and timing of components of insulin administration. One branch of the programming pathway presently under construction will provide for a diabetes hospital patient self-management program. Models for patient self-management in the hospital have been described (Braithwaite et al., 2007; Bailon et al., 2009). The focus of this chapter is on orders for subcutaneous insulin therapy for patients who are not candidates for hospital self-management.



Fig. 3. Selection of branch of the pathway.

accessed from the opening menu (Figure 3). Classification as diabetic ketoacidosis (DKA) is suggested by plasma glucose > 250 mg/dL, arterial pH < 7.3, bicarbonate < 15, anion gap > 12 meq/L, and moderate ketonuria or ketonemia. Classification as hyperglycemic hyperosmolar state (HHS) is suggested by plasma glucose > 600 mg/dl, serum osmolality > 320 mosm/L, arterial pH > 7.3, bicarbonate > 15 meq/L, and minimal ketonuria and

For patients who will receive subcutaneous insulin, once the pattern of carbohydrate exposure is determined, then the prescriber can select the appropriate branch of the pathway (Figure 3). Selection of a single branch from the list will launch an appropriate submenu, dependent upon carbohydrate exposure, for the schedule of blood glucose monitoring and the selection and timing of components of insulin administration. One branch of the programming pathway presently under construction will provide for a diabetes hospital patient self-management program. Models for patient self-management in the hospital have been described (Braithwaite et al., 2007; Bailon et al., 2009). The focus of this chapter is on orders for subcutaneous insulin therapy for patients who are not

**3.4 Subcutaneous insulin algorithm selection under the pathway** 

candidates for hospital self-management.

Fig. 3. Selection of branch of the pathway.

ketonemia.

### **3.5 Subcutaneous insulin dose titration after pathway initiation**

A method for establishing starting doses of insulin is described in sections that will follow. An associated guideline might state that rewriting of the doses of scheduled insulin should be considered daily. Here, a guideline for revision of scheduled insulin is presented that is appropriate to each subcutaneous pathway.


The new dose of scheduled insulin is reapportioned between the components of scheduled therapy. Once the treatment pattern has been entered, changing of dose or correction dose scale can be accomplished outside of the programming pathway. If there are no further specifications, any standing "additional directions" concerning scheduled insulin may be carried forward.

Therapeutic inertia in changing established insulin regimens is a recognized problem in the care of hospitalized patients. In a study of 52 hospitalized patients treated with 50% dextrose for an episode of hypoglycemia, it was found that subsequent to withholding of insulin at the time of the hypoglycemia, 31% of the patients received no other change in treatment (Garg et al., 2004). The guideline above would give caregivers direction on trouble-shooting of the causes of hypoglycemia and making appropriate revisions of treatment.

#### **3.6 Integration of the components of care under the pathway during placement of orders for subcutaneous insulin**

A decision support system helps the prescriber to recognize the components of care under the pattern of treatment that is ordered and the relationship between these components. During widespread adoption of computerization of order entry, a distinct computer order is required separately for feedings, intravenous dextrose, glucose monitoring, each component of insulin therapy, treatment of hypoglycemia, and call parameters. The relationship of these elements of care to each other and their timing must be coordinated. The integration of the components of care, achieved by many paper protocols and order sets, must be preserved. The prescriber must be able to accomplish the goals of glycemic control and hypoglycemia prevention without navigation through multiple screens of an electronic order entry system. Whether patients are eating or not, interruption of carbohydrate exposure is a well verified risk factor for hospital hypoglycemia. The risk arises from hospital routine that interrupts feedings or patient factors that result in poor oral intake (Fischer et al., 1986). Restrictions on free-text entries will necessitate preprogramming of additional directions. In each branch of the pathway that will be shown, in case of reduction of carbohydrate exposure, the insulin orders may be accompanied by standardized statements concerning hypoglycemia

Prevention of Hospital Hypoglycemia by Algorithm

surgery

corticosteroid therapy

Design: A Programming Pathway for Electronic Order Entry 193

• daily prandial requirement 0.15 units/kg for type 2 diabetes or for stress hyperglycemia, and 0.25 units/kg for type 1 diabetes, apportioned between three meals. The total daily dose of scheduled insulin is apportioned between the scheduled basal and prandial insulin. A guideline concerning the initial percentage distribution of total daily dose of scheduled insulin between the components of therapy may suggest 50% basal and

• > 50% basal insulin, < 50%prandial insulin during immediate recovery following heart

• 33% basal insulin , 67%prandial insulin for renal or hepatic failure, malnutrition, or

Fig. 4. Nursing orders for patients who are eating. WMEALS = with meals. WMEALS, HS = with meals and at bedtime. Times of meals may differ according to institutional practices.

50% prandial for most patients, but reapportionment for special conditions:

• 50% basal insulin, 50%prandial insulin for many patients

prevention. Examples include additional directions to "hold" prandial insulin in case of meal omission, "hold" prandial insulin on the mornings of dialysis, or "reduce" insulin in case of poor oral intake.

The order entry system should associate the pattern of blood glucose monitoring, the components of insulin administration together with additional directions, the "call" parameters, and the orders for "prn" oral or intravenous carbohydrate. If an order entry system is well designed, the user will encounter a comprehensive electronic menu for prescribing a glycemic management plan, having internally coordinated components, accessible through a single branch of the pathway of order entry. Under each of the first three branches of the pathway, nursing instructions include an assessment of patient needs, including early attention to patient education and eventual discharge planning.

#### **3.6.1 Subcutaneous Insulin for patients who are eating**

Basal-prandial-correction therapy is a prescribing pattern for insulin, described in previous reviews, that is especially well suited to insulin treatment of the hospitalized patient who is eating (Hirsch, 2005; Clement et al., 2004). The orders for monitoring and insulin are written in association with a meal plan, usually a consistent carbohydrate diet. Other specifications to the diet are preserved that may be required for care of comorbidities. The nursing orders for monitoring of blood glucose provide options for testing postprandially but recommend restriction of scheduled postprandial testing to conditions in which retrospective review of the results might be used to revise scheduled therapy for special populations or conditions, such as pregnancy or cystic fibrosis (Figure 4). Most patients require either testing with meals; with meals and at bedtime; or with meals, at bedtime, and midsleep.

Prandial insulin coverage is the treatment given to cover meals, and basal insulin is the treatment necessary to prevent unchecked gluconeogenesis and ketogenesis, required whether or not nutrition is provided. The long-acting insulin analogs glargine and detemir are designed to provide basal coverage. Glargine may be given once daily for most patients, and detemir may be given once or twice daily. The rapid-acting insulin analogs lispro, aspart and glulisine are designed to provide prandial coverage and to provide rapid correction of hyperglycemia. Biphasic or premixed insulin therapy provides both basal and prandial insulin coverage. In the hospital, since there is a risk of interruption of meals, it is desirable to use an insulin treatment plan under which the prandial component of treatment can be interrupted without compromise to the basal insulin coverage. For patients eating discrete meals, biphasic insulin therapy in the hospital generally is replaced by treatment separately with basal coverage and prandial coverage. For correction of hyperglycemia, the rapid-acting insulin analogs are given with meals, sometimes for coverage of snacks, and sometimes at bedtime or midsleep.

Some patients having type 2 diabetes who normally require insulin may experience reduction of insulin resistance during fasting and may produce endogenous insulin sufficient that under conditions of reduced oral intake the requirements for exogenous insulin may decline. Others, who normally are insulin independent, may experience stressrelated insulin resistance in the hospital sufficient to produce a requirement for exogenous insulin treatment. Dose initiation guidelines for insulin-requiring patients whose dose requirements are not known might be stated conservatively as follows (with reapportionment as indicated for special conditions, as described below):

• daily basal requirement 0.15 units/kg for type 2 diabetes or for stress hyperglycemia, and 0.25 units/kg for type 1 diabetes

prevention. Examples include additional directions to "hold" prandial insulin in case of meal omission, "hold" prandial insulin on the mornings of dialysis, or "reduce" insulin in

The order entry system should associate the pattern of blood glucose monitoring, the components of insulin administration together with additional directions, the "call" parameters, and the orders for "prn" oral or intravenous carbohydrate. If an order entry system is well designed, the user will encounter a comprehensive electronic menu for prescribing a glycemic management plan, having internally coordinated components, accessible through a single branch of the pathway of order entry. Under each of the first three branches of the pathway, nursing instructions include an assessment of patient needs,

Basal-prandial-correction therapy is a prescribing pattern for insulin, described in previous reviews, that is especially well suited to insulin treatment of the hospitalized patient who is eating (Hirsch, 2005; Clement et al., 2004). The orders for monitoring and insulin are written in association with a meal plan, usually a consistent carbohydrate diet. Other specifications to the diet are preserved that may be required for care of comorbidities. The nursing orders for monitoring of blood glucose provide options for testing postprandially but recommend restriction of scheduled postprandial testing to conditions in which retrospective review of the results might be used to revise scheduled therapy for special populations or conditions, such as pregnancy or cystic fibrosis (Figure 4). Most patients require either testing with

Prandial insulin coverage is the treatment given to cover meals, and basal insulin is the treatment necessary to prevent unchecked gluconeogenesis and ketogenesis, required whether or not nutrition is provided. The long-acting insulin analogs glargine and detemir are designed to provide basal coverage. Glargine may be given once daily for most patients, and detemir may be given once or twice daily. The rapid-acting insulin analogs lispro, aspart and glulisine are designed to provide prandial coverage and to provide rapid correction of hyperglycemia. Biphasic or premixed insulin therapy provides both basal and prandial insulin coverage. In the hospital, since there is a risk of interruption of meals, it is desirable to use an insulin treatment plan under which the prandial component of treatment can be interrupted without compromise to the basal insulin coverage. For patients eating discrete meals, biphasic insulin therapy in the hospital generally is replaced by treatment separately with basal coverage and prandial coverage. For correction of hyperglycemia, the rapid-acting insulin analogs are given with meals, sometimes for coverage of snacks, and

Some patients having type 2 diabetes who normally require insulin may experience reduction of insulin resistance during fasting and may produce endogenous insulin sufficient that under conditions of reduced oral intake the requirements for exogenous insulin may decline. Others, who normally are insulin independent, may experience stressrelated insulin resistance in the hospital sufficient to produce a requirement for exogenous insulin treatment. Dose initiation guidelines for insulin-requiring patients whose dose requirements are not known might be stated conservatively as follows (with

• daily basal requirement 0.15 units/kg for type 2 diabetes or for stress hyperglycemia,

reapportionment as indicated for special conditions, as described below):

including early attention to patient education and eventual discharge planning.

meals; with meals and at bedtime; or with meals, at bedtime, and midsleep.

**3.6.1 Subcutaneous Insulin for patients who are eating** 

case of poor oral intake.

sometimes at bedtime or midsleep.

and 0.25 units/kg for type 1 diabetes


Fig. 4. Nursing orders for patients who are eating. WMEALS = with meals. WMEALS, HS = with meals and at bedtime. Times of meals may differ according to institutional practices.

Prevention of Hospital Hypoglycemia by Algorithm

Design: A Programming Pathway for Electronic Order Entry 195

Fig. 5. Basal insulin orders for patients who are eating. The start time and duration for each

recurring medication order are to be programmed, but will not shown. SC = subcutaneously; NPO = nihil per os (nothing by mouth). Abbreviations may differ

according to institutional policy.

Under a treatment plan using basal-prandial-correction dose therapy, typically basal insulin is given once daily as long-acting insulin analog. Before development of insulin analogs, NPH insulin had been used to provide basal coverage, and regular insulin to provide prandial coverage and correction of hyperglycemia. In general, during the treatment of type 2 diabetes, in comparison with NPH-based basal insulin regimens there is less hypoglycemia with use of long-acting insulin analog therapy for basal coverage (Rosenstock et al., 2005; Hermansen et al., 2006). However, some patients prefer to be treated with NPH. Morning dosing with NPH insulin may provide both basal and partial prandial insulin coverage. In the ambulatory setting, some patients use NPH insulin to achieve pattern correction; for example, an evening dose of NPH insulin may cover the dawn phenomenon, correcting a pattern of morning hyperglycemia by meeting predawn insulin resistance with increased insulin levels. Under the branch of the pathway for patients who are eating, prescribers are given the alternatives of using a long-acting insulin analog or NPH for basal insulin coverage (Figure 5).

Not uncommonly, in order to correct fasting hyperglycemia, doses of intermediate or long acting insulin may have been increased during normal dietary intake to a dose higher than true basal requirements. If the basal insulin dose is unchanged during NPO status (nihil per os, nothing by mouth), patients having type 2 diabetes may experience hypoglycemia (Olson et al., 2009). It is important that the programming pathway should present options for basal insulin reduction or interruption in case of planned NPO status. On the other hand, if the basal insulin dose is established correctly in type 1 diabetes, the dose during NPO status usually may be preserved (Mucha et al., 2004). Omission of basal insulin during NPO status in type 1 diabetes may result in ketoacidosis. Therefore, the programming pathway provides options for prescribers to reduce basal insulin in type 2 diabetes but to continue basal insulin in type 1 diabetes, in anticipation of NPO status. A prescriber guideline embedded in the order entry screen warns against interruption of basal insulin for type 1 diabetes (Figure 5).

In the treatment of type 2 diabetes, rapid-acting analogs for prandial coverage may produce less hypoglycemia than regular insulin (Anderson et al., 1997; Raymann et al, 2006; Velussi at al. 2002). The provider may see the need to provide differing doses of prandial insulin at different times of day; the programming pathway permits flexibility in the prescribing of prandial doses, allowing either a fixed dose (best ordered usually with a consistent carbohydrate diet) or a variable dose (Figure 6). This programming pathway is designed for use on the assumption that not all nursing staff are trained on recognition of carbohydrate content of meals; therefore, insulin-to-carbohydrate ratios are not prescribed under the branch of the pathway for subcutaneous insulin for patients who are eating. A modification of the pathway might be used by hospitals that routinely train all nurses on advanced carbohydrate counting so that the provider might order and nurses might use an insulin to carbohydrate ratio to assign prandial insulin doses according to what is on the patient's tray. Patients using the skills of advanced carbohydrate counting and already skilled in self management may best be treated under a different branch of the pathway, for diabetes hospital patient self management.

Several additional directions may be selected in conjunction with orders for prandial use of rapid-acting insulin analog that provide protection against hypoglycemia. Most obviously, the direction "HOLD IF NPO" is intended to reduce the risk of administration of prandial insulin at times when meals might be omitted. The order to hold prandial insulin for

Under a treatment plan using basal-prandial-correction dose therapy, typically basal insulin is given once daily as long-acting insulin analog. Before development of insulin analogs, NPH insulin had been used to provide basal coverage, and regular insulin to provide prandial coverage and correction of hyperglycemia. In general, during the treatment of type 2 diabetes, in comparison with NPH-based basal insulin regimens there is less hypoglycemia with use of long-acting insulin analog therapy for basal coverage (Rosenstock et al., 2005; Hermansen et al., 2006). However, some patients prefer to be treated with NPH. Morning dosing with NPH insulin may provide both basal and partial prandial insulin coverage. In the ambulatory setting, some patients use NPH insulin to achieve pattern correction; for example, an evening dose of NPH insulin may cover the dawn phenomenon, correcting a pattern of morning hyperglycemia by meeting predawn insulin resistance with increased insulin levels. Under the branch of the pathway for patients who are eating, prescribers are given the alternatives of using a long-acting insulin analog or NPH for basal

Not uncommonly, in order to correct fasting hyperglycemia, doses of intermediate or long acting insulin may have been increased during normal dietary intake to a dose higher than true basal requirements. If the basal insulin dose is unchanged during NPO status (nihil per os, nothing by mouth), patients having type 2 diabetes may experience hypoglycemia (Olson et al., 2009). It is important that the programming pathway should present options for basal insulin reduction or interruption in case of planned NPO status. On the other hand, if the basal insulin dose is established correctly in type 1 diabetes, the dose during NPO status usually may be preserved (Mucha et al., 2004). Omission of basal insulin during NPO status in type 1 diabetes may result in ketoacidosis. Therefore, the programming pathway provides options for prescribers to reduce basal insulin in type 2 diabetes but to continue basal insulin in type 1 diabetes, in anticipation of NPO status. A prescriber guideline embedded in the order entry screen warns against interruption of basal insulin for type 1

In the treatment of type 2 diabetes, rapid-acting analogs for prandial coverage may produce less hypoglycemia than regular insulin (Anderson et al., 1997; Raymann et al, 2006; Velussi at al. 2002). The provider may see the need to provide differing doses of prandial insulin at different times of day; the programming pathway permits flexibility in the prescribing of prandial doses, allowing either a fixed dose (best ordered usually with a consistent carbohydrate diet) or a variable dose (Figure 6). This programming pathway is designed for use on the assumption that not all nursing staff are trained on recognition of carbohydrate content of meals; therefore, insulin-to-carbohydrate ratios are not prescribed under the branch of the pathway for subcutaneous insulin for patients who are eating. A modification of the pathway might be used by hospitals that routinely train all nurses on advanced carbohydrate counting so that the provider might order and nurses might use an insulin to carbohydrate ratio to assign prandial insulin doses according to what is on the patient's tray. Patients using the skills of advanced carbohydrate counting and already skilled in self management may best be treated under a different branch of the pathway, for diabetes

Several additional directions may be selected in conjunction with orders for prandial use of rapid-acting insulin analog that provide protection against hypoglycemia. Most obviously, the direction "HOLD IF NPO" is intended to reduce the risk of administration of prandial insulin at times when meals might be omitted. The order to hold prandial insulin for

insulin coverage (Figure 5).

diabetes (Figure 5).

hospital patient self management.


Fig. 5. Basal insulin orders for patients who are eating. The start time and duration for each recurring medication order are to be programmed, but will not shown. SC = subcutaneously; NPO = nihil per os (nothing by mouth). Abbreviations may differ according to institutional policy.

Prevention of Hospital Hypoglycemia by Algorithm

for each of those three time plans (Figure 6).

Design: A Programming Pathway for Electronic Order Entry 197

pre-meal targets, may result in hypoglycemia when applied postprandially, prior to full dissipation of the effects of any earlier correction dose. Therefore, orders for correction doses under the programming pathway are restricted to the following three time plans for administration: with meals; HS (bedtime); 0200. The orders may provide a different scale

Fig. 7. Bridging doses of insulin for patients who are eating, at the time of transition from

Transitioning guidelines from intravenous insulin infusion to subcutaneous insulin recommend that the provider should order subcutaneous insulin before interruption of insulin infusion (Osburne et al., 2006). Infrequently small amounts of basal and prandial insulin (but not subcutaneous correction doses) may be started more than 2 - 4 hr prior to interruption of intravenous insulin infusion. In order to transition from intravenous insulin to subcutaneous insulin, the 24-hr requirement for scheduled subcutaneous basal insulin that is to be started or added may be about 80% of the 24-hr amount of basal insulin, extrapolated from observation of insulin requirement during the last 6-8 hr of intravenous insulin infusion. To avoid overestimation of basal dose requirement, observation must be made during a timeframe of medical stability during which there have been no meals, such as midnight to 0800; there must be no change of carbohydrate-containing maintenance fluids, enteral feedings, or total parenteral nutritional at the time of transition to

intravenous insulin infusion to subcutaneous insulin.

glucose below a given threshold replicates a conservative practice pattern that many users of multiple daily insulin injections employ at home. Acceptable control may be achieved by postprandial administration of rapid-acting insulin analog (Jungmann, 2005). For patients whose oral intake is uncertain, the programming pathway provides the option that the use of prandial insulin might be withheld until 50% of the tray has been taken. For patients with stage V chronic kidney disease having hemodialysis, there may be greater risk for hypoglycemia on hemodialysis days (Kazempour-Ardebili et al., 2009). To permit insulin dose reduction by dose omission of rapid-acting analog at breakfast and lunch on dialysis days, a checkbox is provided specifying that the nurse should withhold the scheduled rapidacting analog before breakfast and lunch on hemodialysis days (Figure 6).

Fig. 6. Prandial insulin (nutritional insulin) for patients who are eating. BG = point-of-care blood glucose.

Frequent dosing with rapid-acting analogs for correction of hyperglycemia creates the risk of "stacking" of effect. When a patient has had hyperglycemia prior to a meal, consideration of another correction dose may arise in the postprandial state. The effect of a previously administered correction dose may not have been fully exerted when the blood glucose is retested. Use of a fixed glucose-dependent correction dose rule, designed to meet specific

glucose below a given threshold replicates a conservative practice pattern that many users of multiple daily insulin injections employ at home. Acceptable control may be achieved by postprandial administration of rapid-acting insulin analog (Jungmann, 2005). For patients whose oral intake is uncertain, the programming pathway provides the option that the use of prandial insulin might be withheld until 50% of the tray has been taken. For patients with stage V chronic kidney disease having hemodialysis, there may be greater risk for hypoglycemia on hemodialysis days (Kazempour-Ardebili et al., 2009). To permit insulin dose reduction by dose omission of rapid-acting analog at breakfast and lunch on dialysis days, a checkbox is provided specifying that the nurse should withhold the scheduled rapid-

Fig. 6. Prandial insulin (nutritional insulin) for patients who are eating. BG = point-of-care

Frequent dosing with rapid-acting analogs for correction of hyperglycemia creates the risk of "stacking" of effect. When a patient has had hyperglycemia prior to a meal, consideration of another correction dose may arise in the postprandial state. The effect of a previously administered correction dose may not have been fully exerted when the blood glucose is retested. Use of a fixed glucose-dependent correction dose rule, designed to meet specific

blood glucose.

acting analog before breakfast and lunch on hemodialysis days (Figure 6).

pre-meal targets, may result in hypoglycemia when applied postprandially, prior to full dissipation of the effects of any earlier correction dose. Therefore, orders for correction doses under the programming pathway are restricted to the following three time plans for administration: with meals; HS (bedtime); 0200. The orders may provide a different scale for each of those three time plans (Figure 6).


Fig. 7. Bridging doses of insulin for patients who are eating, at the time of transition from intravenous insulin infusion to subcutaneous insulin.

Transitioning guidelines from intravenous insulin infusion to subcutaneous insulin recommend that the provider should order subcutaneous insulin before interruption of insulin infusion (Osburne et al., 2006). Infrequently small amounts of basal and prandial insulin (but not subcutaneous correction doses) may be started more than 2 - 4 hr prior to interruption of intravenous insulin infusion. In order to transition from intravenous insulin to subcutaneous insulin, the 24-hr requirement for scheduled subcutaneous basal insulin that is to be started or added may be about 80% of the 24-hr amount of basal insulin, extrapolated from observation of insulin requirement during the last 6-8 hr of intravenous insulin infusion. To avoid overestimation of basal dose requirement, observation must be made during a timeframe of medical stability during which there have been no meals, such as midnight to 0800; there must be no change of carbohydrate-containing maintenance fluids, enteral feedings, or total parenteral nutritional at the time of transition to

Prevention of Hospital Hypoglycemia by Algorithm

Design: A Programming Pathway for Electronic Order Entry 199

Fig. 8. Nursing orders and scheduled insulin orders for patients who are not eating,

enteral feedings, or no carbohydrate exposure.

including patients receiving continuous dextrose-containing maintenance fluids, continuous

The algorithm we use does not provide a protocolized rule for changing the NPH dose based on glycemic response, but rather alters insulin delivery below a given glucose

subcutaneous insulin therapy; there must have been independence from pressors and continuous veno-venous hemodialysis; and there must be no change of corticosteroid dose. If the time of transition occurs at a time of day that differs from the usual time of administration of long-acting insulin analog, then a bridging dose of regular or intermediate-acting insulin may be given (Figure 7).

#### **3.6.2 Subcutaneous insulin for patients who are not eating**

The patient receiving continuous exposure to carbohydrate as intravenous dextrose or enteral feedings, or the patient receiving no carbohydrate, generally should have glucose monitoring at time intervals that are equally spaced (Figure 8). The order "ACHS" for a patient who has been made "NPO" is meaningless. Therefore, the branch of the programming pathway for patients who are not eating starts with the default order for monitoring of point-of-care blood glucose every 6 hr.

During NPO status, the dose of insulin required to cover dextrose-containing maintenance fluids, total parenteral nutrition (TPN), or enteral tube feedings is described as nutritional insulin. Outside of the programming pathway, the provider may include insulin among the TPN additives. long-acting insulin analog sometimes is used for coverage of continuous enteral feedings. Under such a regimen, safety precautions must be in place for dextrose infusion in case of interruption of enteral feedings. A barrier to creating a universal rule is that patient tolerance for intravenous fluids differs according to condition. Safety data about use of basal insulin during enteral feedings, conducted with careful definition of insulin dose, has been generated in the context of a clinical trial, such that close supervision of the patients can be assumed to have occurred (Koryotkoski et al., 2009).

Personal observation of isolated cases of severe hypoglycemia outside of the research context has led to concern that safety of covering enteral feedings with long-acting analog, demonstrated under controlled research conditions, is not generalizable. In actual practice, use of long-acting analog to cover enteral feedings can be complicated by protracted hypoglycemia, a special risk in case of unforeseen interruption of enteral feedings. When NPH and regular insulin are used every 6 hours, each insulin dose is smaller than under a once-daily glargine program, and the frequency of insulin administration provides deliberate stacking of effect. The use of more frequent and smaller doses of intermediate acting insulin achieves control superior to that of sliding scale insulin; such therapy is intended to reduce the risk of prolonged exposure to high doses of long-acting insulin in case of sudden interruption of enteral feedings, and to reduce the importance of reliance upon the antidote of intravenous dextrose, in case of feeding interruptions.

A "sliding scale" regimen of NPH insulin every 4 hr or every 6 hr has been examined, compared to sliding scale aspart insulin alone for treatment of patients receiving enteral feedings (Cook, A. et al., 2009). Amber Cook and colleagues use a standardized rule for altering the NPH dose based on response of blood glucose. In our programming pathway, and on the antecedent paper order sets, in contrast to the closely related regimen of Amber Cook at al., we specify an option for use of mixtures of NPH and regular insulin every 6 hours. The prescribing style is intended to achieve flat-line coverage of insulin effect. The method described in our guideline is to administer equal doses of insulin every 6 hours, apportioned as 2/3 NPH and 1/3 regular insulin, with instructions to withhold the regular insulin in case of glucose below a given threshold.

subcutaneous insulin therapy; there must have been independence from pressors and continuous veno-venous hemodialysis; and there must be no change of corticosteroid dose. If the time of transition occurs at a time of day that differs from the usual time of administration of long-acting insulin analog, then a bridging dose of regular or

The patient receiving continuous exposure to carbohydrate as intravenous dextrose or enteral feedings, or the patient receiving no carbohydrate, generally should have glucose monitoring at time intervals that are equally spaced (Figure 8). The order "ACHS" for a patient who has been made "NPO" is meaningless. Therefore, the branch of the programming pathway for patients who are not eating starts with the default order for

During NPO status, the dose of insulin required to cover dextrose-containing maintenance fluids, total parenteral nutrition (TPN), or enteral tube feedings is described as nutritional insulin. Outside of the programming pathway, the provider may include insulin among the TPN additives. long-acting insulin analog sometimes is used for coverage of continuous enteral feedings. Under such a regimen, safety precautions must be in place for dextrose infusion in case of interruption of enteral feedings. A barrier to creating a universal rule is that patient tolerance for intravenous fluids differs according to condition. Safety data about use of basal insulin during enteral feedings, conducted with careful definition of insulin dose, has been generated in the context of a clinical trial, such that close supervision of the

Personal observation of isolated cases of severe hypoglycemia outside of the research context has led to concern that safety of covering enteral feedings with long-acting analog, demonstrated under controlled research conditions, is not generalizable. In actual practice, use of long-acting analog to cover enteral feedings can be complicated by protracted hypoglycemia, a special risk in case of unforeseen interruption of enteral feedings. When NPH and regular insulin are used every 6 hours, each insulin dose is smaller than under a once-daily glargine program, and the frequency of insulin administration provides deliberate stacking of effect. The use of more frequent and smaller doses of intermediate acting insulin achieves control superior to that of sliding scale insulin; such therapy is intended to reduce the risk of prolonged exposure to high doses of long-acting insulin in case of sudden interruption of enteral feedings, and to reduce the importance of reliance

A "sliding scale" regimen of NPH insulin every 4 hr or every 6 hr has been examined, compared to sliding scale aspart insulin alone for treatment of patients receiving enteral feedings (Cook, A. et al., 2009). Amber Cook and colleagues use a standardized rule for altering the NPH dose based on response of blood glucose. In our programming pathway, and on the antecedent paper order sets, in contrast to the closely related regimen of Amber Cook at al., we specify an option for use of mixtures of NPH and regular insulin every 6 hours. The prescribing style is intended to achieve flat-line coverage of insulin effect. The method described in our guideline is to administer equal doses of insulin every 6 hours, apportioned as 2/3 NPH and 1/3 regular insulin, with instructions to withhold the regular

intermediate-acting insulin may be given (Figure 7).

monitoring of point-of-care blood glucose every 6 hr.

**3.6.2 Subcutaneous insulin for patients who are not eating** 

patients can be assumed to have occurred (Koryotkoski et al., 2009).

upon the antidote of intravenous dextrose, in case of feeding interruptions.

insulin in case of glucose below a given threshold.

Fig. 8. Nursing orders and scheduled insulin orders for patients who are not eating, including patients receiving continuous dextrose-containing maintenance fluids, continuous enteral feedings, or no carbohydrate exposure.

The algorithm we use does not provide a protocolized rule for changing the NPH dose based on glycemic response, but rather alters insulin delivery below a given glucose

Prevention of Hospital Hypoglycemia by Algorithm

**3.6.4 Diabetes hospital patient self management** 

regimen may be required.

doses for treatment of hyperglycemia.

Design: A Programming Pathway for Electronic Order Entry 201

10). The need for correction dose insulin is likely to occur during and at the end of each feeding. As the patient's intake of oral feedings improves, correction dose insulin during the day may be required. For patients having type 1 diabetes, additionally daily use of long acting insulin analog should be ordered, in an amount restricted to the basal dose, together with the regimen of premixed insulin that is being used for nutritional coverage and regular insulin for correction dose coverage. Once dietary intake is adequate, overnight enteral feedings and the accompanying premedication with 70/30 isophane NPH/regular insulin no longer are required. Once the patient is eating, a new basal-prandial-correction insulin

In the ambulatory setting, skilled use of a flexible insulin program may reduce the frequency of hypoglyemia (Samann et al., 2006). Patients competent at diabetes selfmanagement, for example patients using multiple daily injections or insulin pump therapy, under defined conditions can be treated in the hospital with continuation of their usual program of self-management (Braithwaite et al., 2007; Bailon et al., 2009). A full description of such a program is beyond the scope of this chapter. A hallmark feature is the utilization of the skills of advanced carbohydrate counting to permit matching of mealtime insulin bolus doses to carbohydrate intake, and the use of a rule for establishment of correction

Fig. 9. Basal insulin orders for type 1 diabetes patients who are not eating.

threshold by protocolized omission of scheduled regular insulin. The prescriber within the programming pathway is invited to provide the additional direction for scheduled regular insulin "HOLD FOR BG < XXX" (where BG = point-of-care blood glucose).

Transitioning guidelines from intravenous insulin infusion to subcutaneous insulin for the prescriber include the following, with recognition that the guidelines may not be appropriate for every case, and that individualization is required:


A set of dose initiation guidelines is given for insulin-requiring patients whose dose requirements are not known. The total daily dose of insulin for coverage of enteral feedings or continuous intravenous dextrose exposure may be calculated conservatively as follows:


On a periodic basis, the caregiver then may alter the scheduled NPH and insulin by revising the scheduled insulin orders, using a guideline as shown above for re-establishing total daily dose (see section 3.5), and reapportioning the dose between NPH and regular insulin.

To prevent inadvertent interruption of basal insulin for type 1 diabetes patients who are not eating, a special provision is available, in the branch of the programming pathway for patients who are not eating, to maintain basal dose requirements of long-acting insulin analog treatment when interruption of carbohydrate exposure necessitates interruption of nutritional insulin (Figure 9).

#### **3.6.3 Subcutaneous insulin for patients with overnight enteral feedings and daytime meals**

For patients whose oral intake is temporarily poor but likely to improve, overnight enteral tube feedings may be used during transition from negligible oral intake to a full meal plan. Premixed 70% human insulin isophane suspension/30% human insulin (70/30 NPH / regular insulin) may be used as premedication to cover overnight enteral feedings (Figure

threshold by protocolized omission of scheduled regular insulin. The prescriber within the programming pathway is invited to provide the additional direction for scheduled regular

Transitioning guidelines from intravenous insulin infusion to subcutaneous insulin for the prescriber include the following, with recognition that the guidelines may not be

• Order subcutaneous insulin before interruption of insulin infusion. Infrequently small amounts of scheduled NPH and regular insulin (but not sliding scale) may be started

• Prescribe equal total doses of scheduled insulin every 6 hr, apportioned as 2/3 NPH,

• In order to transition from intravenous insulin to subcutaneous insulin, the 24-hr requirement for scheduled NPH and regular insulin that is to be started or added may be about 80% of the 24-hr amount of intravenous insulin, extrapolated from observation of insulin requirement during the last 6-8 hr of intravenous insulin infusion. To avoid overestimation of dose requirement, observation must be made during a timeframe of medical stability; there must be no change of carbohydrate-containing maintenance fluids, enteral feedings, or total parenteral nutrition at the time of transition to subcutaneous insulin therapy; there must have been independence from pressors and

A set of dose initiation guidelines is given for insulin-requiring patients whose dose requirements are not known. The total daily dose of insulin for coverage of enteral feedings or continuous intravenous dextrose exposure may be calculated conservatively as follows: • 24 hr basal requirement is 0.15 units/kg for type 2 diabetes or for stress hyperglycemia,

• nutritional requirement is 1 unit per 10 gm of carbohydrate per 24 hr, as determined by review of maintenance fluid or enteral tube feeding composition and delivery rate. • for type 2 diabetes or stress hyperglycemia, during continuous carbohydrate exposure, the total daily dose of insulin is the sum of the 24 hr basal and nutritional components of

• the total daily dose of insulin is apportioned between NPH and regular insulin as above. On a periodic basis, the caregiver then may alter the scheduled NPH and insulin by revising the scheduled insulin orders, using a guideline as shown above for re-establishing total daily dose (see section 3.5), and reapportioning the dose between NPH and regular insulin. To prevent inadvertent interruption of basal insulin for type 1 diabetes patients who are not eating, a special provision is available, in the branch of the programming pathway for patients who are not eating, to maintain basal dose requirements of long-acting insulin analog treatment when interruption of carbohydrate exposure necessitates interruption of

**3.6.3 Subcutaneous insulin for patients with overnight enteral feedings and daytime** 

For patients whose oral intake is temporarily poor but likely to improve, overnight enteral tube feedings may be used during transition from negligible oral intake to a full meal plan. Premixed 70% human insulin isophane suspension/30% human insulin (70/30 NPH / regular insulin) may be used as premedication to cover overnight enteral feedings (Figure

insulin "HOLD FOR BG < XXX" (where BG = point-of-care blood glucose).

more than 2 - 4 hr prior to interruption of intravenous insulin infusion.

appropriate for every case, and that individualization is required:

CVVHD; and there must be no change of corticosteroid dose.

and 0.25 units/kg for type 1 diabetes.

1/3 regular insulin.

therapy.

**meals** 

nutritional insulin (Figure 9).

10). The need for correction dose insulin is likely to occur during and at the end of each feeding. As the patient's intake of oral feedings improves, correction dose insulin during the day may be required. For patients having type 1 diabetes, additionally daily use of long acting insulin analog should be ordered, in an amount restricted to the basal dose, together with the regimen of premixed insulin that is being used for nutritional coverage and regular insulin for correction dose coverage. Once dietary intake is adequate, overnight enteral feedings and the accompanying premedication with 70/30 isophane NPH/regular insulin no longer are required. Once the patient is eating, a new basal-prandial-correction insulin regimen may be required.

### **3.6.4 Diabetes hospital patient self management**

In the ambulatory setting, skilled use of a flexible insulin program may reduce the frequency of hypoglyemia (Samann et al., 2006). Patients competent at diabetes selfmanagement, for example patients using multiple daily injections or insulin pump therapy, under defined conditions can be treated in the hospital with continuation of their usual program of self-management (Braithwaite et al., 2007; Bailon et al., 2009). A full description of such a program is beyond the scope of this chapter. A hallmark feature is the utilization of the skills of advanced carbohydrate counting to permit matching of mealtime insulin bolus doses to carbohydrate intake, and the use of a rule for establishment of correction doses for treatment of hyperglycemia.

Fig. 9. Basal insulin orders for type 1 diabetes patients who are not eating.

Prevention of Hospital Hypoglycemia by Algorithm

**5. Acknowledgement** 

Care 34 (Suppl 1):S11-S61.

*Med* 157 (11):1249-55.

28 (5):1008-1011.

discussion 55-7.

Diabetes 22 (2):81-88.

*Insulin Journal* 5 (1):16-36.

**6. References** 

Design: A Programming Pathway for Electronic Order Entry 203

springboard for development and for modification to meet local needs. A user of the electronic order entry system may opt out of the programming pathway. The pathway is intended to both standardize order entry and also to facilitate individualization of care by the provider and for the patient. An opening screen offers default orders that will be universally desirable for glycemic management and then asks the user to choose a branch of the programming pathway based on route of insulin (intravenous or subcutaneous) and, for subcutaneous insulin regimens, based upon carbohydrate exposure. Within each branch of the pathway for subcutaneous insulin, it is possible to complete related orders without navigation between screens and without use of free-text, by entering numbers and selecting additional directions from side menus and drop down menus. User guidelines are displayed or available by computer link. By grouping and prioritizing related orders (especially the plans for nutrition, glucose monitoring, and insulin) and by offering appropriate additional directions within a branch of the pathway, the integration of the components of care, achievable on paper order sets by juxtaposition, is preserved under the electronic order

The authors would like to acknowledge the work of the Diabetes Council of the Resurrection Health Care System and the input of members of the nursing, nutrition services, pharmacy, resident physician, attending physician, and administrative staff.

American Diabetes Association. 2011. Standards of medical care in diabetes - 2011. Diabetes

Anderson, J. H., Jr., R. L. Brunelle, P. Keohane, V. A. Koivistos, M. E. Trautmann, L. Vignati,

Bagshaw, S. M., M. Egi, C. George, and R. Bellomo. 2009. Early blood glucose control and mortality in critically ill patients in Australia. *Crit Care Med* 37 (2):463-70. Bhatia, A., B. Cadman, and I. Mackenzie. 2006. Hypoglycemia and Cardiac Arrest in a Critically Ill Patient on Strict Glycemic Control. *Anesth Analg* 102 (2):549-551. Baldwin, D., G. Villanueva, R. McNutt, and S. Bhatnagar. 2005. Eliminating Inpatient

Bellam, H., and S. S. Braithwaite. 2010. Hospital hypoglycemia: from observation to action.

Blumenthal, D., and M. Tavenner. 2010. The "Meaningful Use" Regulation for Electronic

Boord, J.A. , R.A. Greevy, SS Braithwaite, et al. 2009. Evaluation of hospital glycemic control

Braithwaite, S. S., B. Robertson, H. P. Mehrotra, L. M. McElveen, and C. L. Thompson. 2007.

Campbell, K B, and S S Braithwaite. 2004. Hospital Management of Hyperglycemia. Clin

Managing hyperglycemia in hospitalized patients. *Clin Cornerstone* 8 (2):44-54;

Health Records. *New England Journal of Medicine* 363 (6):501-504.

at US academic medical centers. Hospital Medicine 4:35–44.

and R. DiMarchi. 1997. Mealtime treatment with insulin analog improves postprandial hyperglycemia and hypoglycemia in patients with non-insulindependent diabetes mellitus. Multicenter Insulin Lispro Study Group. *Arch Intern* 

Sliding-Scale Insulin: A reeducation project with medical house staff. *Diabetes Care*

entry system by user choice of a branch of the programming pathway.

Fig. 10. Insulin orders for patients having overnight enteral tube feedings and daytime meals.

## **4. Conclusion**

A programming pathway for computerized order entry is described that will present templates to the prescriber for well-established strategies for control of hyperglycemia and prevention of hospital hypoglycemia. There is not yet an embodiment of the plan within an existing electronic health record, nor is the content specifically yet endorsed by the healthcare system of the authors, but rather the plan is proposed in general terms as a springboard for development and for modification to meet local needs. A user of the electronic order entry system may opt out of the programming pathway. The pathway is intended to both standardize order entry and also to facilitate individualization of care by the provider and for the patient. An opening screen offers default orders that will be universally desirable for glycemic management and then asks the user to choose a branch of the programming pathway based on route of insulin (intravenous or subcutaneous) and, for subcutaneous insulin regimens, based upon carbohydrate exposure. Within each branch of the pathway for subcutaneous insulin, it is possible to complete related orders without navigation between screens and without use of free-text, by entering numbers and selecting additional directions from side menus and drop down menus. User guidelines are displayed or available by computer link. By grouping and prioritizing related orders (especially the plans for nutrition, glucose monitoring, and insulin) and by offering appropriate additional directions within a branch of the pathway, the integration of the components of care, achievable on paper order sets by juxtaposition, is preserved under the electronic order entry system by user choice of a branch of the programming pathway.

## **5. Acknowledgement**

The authors would like to acknowledge the work of the Diabetes Council of the Resurrection Health Care System and the input of members of the nursing, nutrition services, pharmacy, resident physician, attending physician, and administrative staff.

### **6. References**

202 Diabetes – Damages and Treatments

Fig. 10. Insulin orders for patients having overnight enteral tube feedings and daytime meals.

A programming pathway for computerized order entry is described that will present templates to the prescriber for well-established strategies for control of hyperglycemia and prevention of hospital hypoglycemia. There is not yet an embodiment of the plan within an existing electronic health record, nor is the content specifically yet endorsed by the healthcare system of the authors, but rather the plan is proposed in general terms as a

**4. Conclusion** 


Prevention of Hospital Hypoglycemia by Algorithm

Circulation 117(8):1018-27.

*Hosp Med* 4 (1):3-15.

discussion 610-2.

and outcomes. *Crit Care Med* 35 (10):2262-7.

Profiles. *Diabetes Care* 25 (9):1597-1602.

Design: A Programming Pathway for Electronic Order Entry 205

Kazempour-Ardebili, S., V. L. Lecamwasam, T. Dassanyake, A. H. Frankel, F. W. K. Tam, A.

Korytkowski, M. T., R. J. Salata, G. L. Koerbel, F. Selzer, E. Karslioglu, A. M. Idriss, K. K.W.

Kosiborod, M, S.E. Inzucchi, H.M. Krumholz, L. Xiao, P.G. Jones, S. Fiske, F.A. Masoudi, S.P.

Kosiborod, M., S.E. Inzucchi, A. Goyal, H.M. Krumholz, F.A. Masoudi, L. Xiao, and J.A. Spertus.

patients hospitalized with acute myocardial infarction. *JAMA* 301 (15):1556-64. Krinsley, J. S., and A. Grover. 2007. Severe hypoglycemia in critically ill patients: risk factors

Lee, J., B. Clay, Z. Zelazny, and G. Maynard. 2008. Indication-based ordering: a new paradigm for glycemic control in hospitalized inpatients. *J Diabetes Sci Technol* 2 (3):349-56. Maynard, G. A., M. P. Huynh, and M. Renvall. 2008. Iatrogenic Inpatient Hypoglycemia: Risk

Maynard, G., J. Lee, G. Phillips, E. Fink, and M. Renvall. 2009. Improved inpatient use of

Moghissi E.S., M.T. Korytkowski, M. DiNardo, et al. 2009. American Association of Clinical

Mucha, G. T., S. Merkel, W. Thomas, and J. P. Bantle. 2004. Fasting and Insulin Glargine in

Mudaliar, S., P. Mohideen, R. Deutsch, T. P. Ciaraldi, D. Armstrong, B. Kim, X. Sha, and R.

Olson, R. P., M. A. Bethel, and L. Lien. 2009. Preoperative hypoglycemia in a patient

Osburne, R. C., C. B. Cook, L. Stockton, M. Baird, V. Harmon, A. Keddo, T. Pounds, L.

Rayman, G., V. Profozic, M. Middle. 2006. Insulin glulisine imparts effective glycaemic control in patients with Type 2 diabetes. Diabetes Res Clin Pract. 76(2):304-312. Rosenstock, J., G. Dailey, M. Massi-Benedetti, A. Fritsche, Z. Lin, and A. Salzman. 2005.

Inpatient Glycemic Control. Endocrine Practice 15(4):1-17.

receiving insulin detemir. *Anesth Analg* 108 (6):1836-8.

Algorithm. *The Diabetes Educator* 32 (3):394-403.

Individuals With Type 1 Diabetes. *Diabetes Care* 27 (5):1209-1210.

randomized controlled clinical trial. *Diabetes Care* 32 (4):594-596.

Dornhorst, G. Frost, and J. J. O. Turner. 2009. Assessing Glycemic Control in Maintenance Hemodialysis Patients With Type 2 Diabetes. *Diabetes Care* 32 (7):1137.

Lee, A. J. Moser, and F. G.S. Toledo. 2009. Insulin Therapy and Glycemic Control in Hospitalized Patients With Diabetes During Enteral Nutrition Therapy: A

Marso, J.S. Spertus. 2008. Glucometrics in patients hospitalized with acute myocardial infarction: defining the optimal outcomes-based measure of risk.

2009. Relationship between spontaneous and iatrogenic hypoglycemia and mortality in

Factors, Treatment, and Prevention: Analysis of Current Practice at an Academic Medical Center With Implications for Improvement Efforts. *Diabetes Spectr* 21 (4):241-247.

basal insulin, reduced hypoglycemia, and improved glycemic control: effect of structured subcutaneous insulin orders and an insulin management algorithm. *J* 

Endocrinologists and American Diabetes Association Consensus Statement on

R. Henry. 2002. Intravenous Glargine and Regular Insulin Have Similar Effects on Endogenous Glucose Output and Peripheral Activation/Deactivation Kinetic

Lowey, J. Reid, K. A. McGowan, and P. C. Davidson. 2006. Improving Hyperglycemia Management in the Intensive Care Unit: Preliminary Report of a Nurse-Driven Quality Improvement Project Using a Redesigned Insulin Infusion

Reduced hypoglycemia risk with insulin glargine: a meta-analysis comparing insulin glargine with human NPH insulin in type 2 diabetes. *Diabetes Care* 28 (4):950-5. Scalea, T. M., G. V. Bochicchio, K. M. Bochicchio, S. B. Johnson, M. Joshi, and A. Pyle. 2007.

Tight glycemic control in critically injured trauma patients. *Ann Surg* 246 (4):605-10;


Clement S., S.S. Braithwaite, M.F. Magee, et al. 2004. Management of Diabetes and

Cook, A., D. Burkitt, L. McDonald, and L. Sublett. 2009. Evaluation of Glycemic Control

Continuously Tube-Fed Patients. *Nutrition in Clinical Practice* 24 (6):718-722. Bailon, RM, B.J. Partlow, V. Miller-Cage, M.E. Boyle, J.C. Castro, P.B. Bourgeois, C.B. Cook.

and fluid therapy of hyperglycemic crisis. *Diabetes Management* 1 (4):397–412 Donaldson, S., G. Villanuueva, L. Rondinelli, and D. Baldwin. 2006. Rush University

Dortch, M. J., N. T. Mowery, A. Ozdas, L. Dossett, H. Cao, B. Collier, G. Holder, R. A. Miller,

Finfer S., D.R. Chittock, S.Y. Su, et al. 2009. Intensive versus conventional glucose control in

Fischer, K F, J A Lees, and J H Newman. 1986. Hypoglycemia in hospitalized patients. *N* 

Garg, R., H. Bhutani, A. Jarry, and M. Pendergrass. 2007. Provider response to insulininduced hypoglycemia in hospitalized patients. *J Hosp Med* 2 (4):258-60. Hermansen, K., M. Davies, T. Derezinski, G. Martinez Ravn, P. Clauson, and P. Home. 2006.

Hermayer, K.L., D.E. Neal, T.V. Hushion, M.G. Irving, P.C. Arnold, L.Kozlowski, M.R.

Hermayer, K.L., P. Cawley, P. Arnold, A. Sutton, J. Crudup, L. Kozlowski, and T.V.

Juneja, R., C. Roudebush, N. Kumar, A. Macy, A. Golas, D. Wall, C. Wolverton, D. Nelson, J.

Jungmann E. 2005. Intensified insulin therapy of type 2 diabetes mellitus: pre- or

Kagansky, N, S Levy, E Rimon, L Cojocaru, A Fridman, Z Ozer, and H Knobler. 2003.

Kitabchi, A. E., G. E. Umpierrez, J. M. Miles, and J. N. Fisher. 2009. Hyperglycemic crises in

adult patients with diabetes. Diabetes Care 32 (7):1335-43.

hypoglycemia at a university medical center. *J. Hosp. Med.* 4 (6):331-339.

naive people with type 2 diabetes. *Diabetes Care* 29 (6):1269-74.

University Hospital. *Diabetes Technology & Therapeutics* 9:523-534.

Using NPH Insulin Sliding Scale Versus Insulin Aspart Sliding Scale in

2009. Continuous subcutaneous insulin infusion (insulin pump) therapy can be safely used in the hospital in select patients. Endocr Pract. Jan-Feb;15(1):24-9. Devi, R., G. Selvakumar, L. Clark, C. Downer, and S.S. Braithwaite. 2011. A dose-defining algorithm

for attainment and maintenance of glycemic targets during intravenous insulin infusion

guidelines and protocols for the management of hyperglycemia in hospitalized patients: elimination of the sliding scale and improvement of glycemic control

and A. K. May. 2008. A Computerized Insulin Infusion Titration Protocol Improves Glucose Control With Less Hypoglycemia Compared to a Manual Titration Protocol in a Trauma Intensive Care Unit. *JPEN J Parenter Enteral Nutr* 32 (1):18-27.

A 26-week, randomized, parallel, treat-to-target trial comparing insulin detemir with NPH insulin as add-on therapy to oral glucose-lowering drugs in insulin-

Stroud, F.B. Kerr, and J.M. Kratz. 2007. Outcomes of a Cardiothoracic Intensive Care Web-Based Online Intravenous Insulin Infusion Calculator Study at a Medical

Hushion. 2009. Impact of improvement efforts on glycemic control and

Carroll, and S. Flanders. 2007. Utilization of a computerized intravenous insulin infusion program to control blood glucose in the intensive care unit. *Diabetes* 

postprandial injection of aspart insulin? Dtsch Med Wochenschr 130(20):1254-1257.

Hypoglycemia as a predictor of mortality in hospitalized elderly patients. *Arch* 

Hyperglycemia in Hospitals. Diabetes Care 27(2):553-591.

throughout the hospital. *Diabetes Educ* 32 (6):954-62.

critically ill patients. N Engl J Med 360(13):1283-97.

Hirsch, I.B. 2005. Insulin analogues. N Engl J Med. 352(2):174-83.

*Engl J Med* 315:1245-50.

*Technol Ther* 9:232-240.

*Intern Med* 163:1825-1829.


**0**

**11**

**Hypoglycemia Prevention in Closed-Loop**

<sup>1</sup>*Research Group on Statistics, Applied Economics and Health (GRECS),*

The current chapter addresses the problem of hypoglycemia in type 1 diabetes from biomedical and control engineering points of view. It gives a general introduction to the artificial pancreas system, and the risk of hypoglycemia in closed-loop insulin treatment. Then, it provides a review on the state of the art in hypoglycemia control, and the recent approaches in dealing with hypoglycemia in closed-loop artificial pancreas systems. Next, different control techniques that can be used to minimize the risk of hypoglycemia and

Since the Diabetes Control and Complications Trial (DCCT), tight glycemic control has been established as the control objective in the treatment of patients with type 1 diabetes mellitus (T1DM) (DCCT Research Group (1993)), except if some contraindication exists. However, there still lacks a universal, efficient and safe system able to normalize the glucose levels of patients. The intensive insulin therapy required to achieve the tight glycemic control, based on the injection of basal and bolus insulin to reproduce its physiological secretion, has as counteraction an increase in the risk of significant and severe hypoglycemia with all their consequences. Therefore, hypoglycemia is considered as one of the major limiting factors in

With the inability of conventional therapy to achieve satisfactory glycemic control, and the development in continuous glucose monitoring (CGM) systems and the increasing use of insulin pumps, the idea of developing an artificial pancreas is viewed as the ideal solution for glycemic control in T1DM (Bequette (2005); Hovorka et al. (2006); Kumareswaran et al. (2009)). The artificial pancreas is an automated closed-loop system that maintains blood glucose levels within the desired range and prevents hypoglycemia, while minimizing or eliminating the need for patient intervention. The artificial pancreas replaces the *β*-cells functions in glucose sensing and insulin delivery. It consists of three main components (Figure 1): a glucose sensor to measure glucose concentration, a pump for insulin delivery, and a closed-loop control algorithm to bridge between the glucose measurements and the dose of insulin to be delivered. As other medical devices, the architecture of closed-loop

**1. Introduction**

improve the control outputs are presented.

achieving tight glycemic control in T1DM (Cryer (2008)).

<sup>2</sup>*Automatic Control Laboratory, KTH Royal Institute of Technology*

**Artificial Pancreas for Patients**

Amjad Abu-Rmileh1 and Winston Garcia-Gabin<sup>2</sup>

**with Type 1 Diabetes**

*University of Girona*

<sup>1</sup>*Spain* <sup>2</sup>*Sweden*


## **Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes**

Amjad Abu-Rmileh1 and Winston Garcia-Gabin<sup>2</sup>

<sup>1</sup>*Research Group on Statistics, Applied Economics and Health (GRECS), University of Girona* <sup>2</sup>*Automatic Control Laboratory, KTH Royal Institute of Technology* <sup>1</sup>*Spain* <sup>2</sup>*Sweden*

#### **1. Introduction**

206 Diabetes – Damages and Treatments

Schnipper, J. L., C. D. Ndumele, C. L. Liang, and M. L. Pendergrass. 2009. Effects of a

Schnipper, J. L., C. L. Liang, C. D. Ndumele, and M. L. Pendergrass. 2010. Effects of a

Smith, W. D., A. G. Winterstein, T. Johns, E.Rosenberg, and B. C. Sauer. 2005. Causes of

Stagnaro-Green, A, M K Barton, P L Linekin, E Corkery, K deBeer, and S H Roman. 1995.

Thompson, C. L., K. C. Dunn, M. C. Menon, L. E. Kearns, and S. S. Braithwaite. 2005.

Turchin, A., M. E. Matheny, M. Shubina, J. V. Scanlon, B. Greenwood, and M. L.

Trujillo, J. M., E. E. Barsky, B. C. Greenwood, S. A. Wahlstrom, S. Shaykevich, M. L.

Umpierrez, G. E., J. P. Kelly, J. E. Navarrete, M. M. C. Casals, and A. E. Kitabchi. 1997.

Umpierrez, G. E., D. Smiley, A. Zisman, L. M. Prieto, A. Palacio, M. Ceron, A. Puig, and R. Mejia.

Varghese, P., V. Gleason, R. Sorokin, C. Senholzi, S. Jabbour, and J. E. Gottlieb. 2007.

Velussi, M. 2002. Lispro insulin treatment in comparison with regular human insulin in type 2 diabetic patients living in nursing homes. *Diabetes Nutr Metab* 15 (2):96-100. Vriesendorp, T., S. van Santen, H. De Vries, E. de Jonge, F. Rosendaal, M. Schultz, and J.

Wexler, D. J., J. B. Meigs, E. Cagliero, D. M. Nathan, and R. W. Grant. 2007. Prevalence of

Wexler, Deborah J., Peter Shrader, Sean M. Burns, and Enrico Cagliero. 2010. Effectiveness

randomized controlled trial. *Endocr Pract* 16 (2):209-18.

Hyperglycemia in the Hospital. *Diabetes Spectr* 18 (1):20-27.

hospitalized in the general ward. *Diabetes Care* 32 (7):1153-7.

Care Units: Benefit Versus Harm. *Diabetes* 55 (11):3151-3159.

*Mount Sinai Journal of Medicine* 62 (6):422-426.

inpatients: a pilot study. *J Hosp Med* 3 (1):55-63.

*Hosp Med* 4 (1):16-27.

*Hosp Med* 2 (4):234-40.

*Crit Care Med* 34 (1):96-101.

44 U.S. hospitals. *Diabetes Care* 30 (2):367-369.

Type 2 Diabetes. *Diabetes Care* 33 (10):2181-2183.

(7):714-719.

subcutaneous insulin protocol, clinical education, and computerized order set on the quality of inpatient management of hyperglycemia: results of a clinical trial. *J* 

computerized order set on the inpatient management of hyperglycemia: a cluster-

hyperglycemia and hypoglycemia in adult inpatients. *Am J Health Syst Pharm* 62

Mortalilty in hospitalized patients with hypoglycemia and severe hyperglycemia.

Pendergrass. 2009. Hypoglycemia and clinical outcomes in patients with diabetes

Pendergrass, and J. L. Schnipper. 2008. Improving glycemic control in medical

Hyperglycemic Crises in Urban Blacks. *Archives of Internal Medicine* 157 (6):669-675.

2007. Randomized Study of Basal-Bolus Insulin Therapy in the Inpatient Management of Patients With Type 2 Diabetes (RABBIT 2 Trial). *Diabetes Care* 30 (9):2181-2186. Umpierrez, G. E., D. Smiley, S. Jacobs, L. Peng, A. Temponi, P. Mulligan, D. Umpierrez, C.

Newton, D. Olson, and M. Rizzo. 2011. Randomized Study of Basal Bolus Insulin Therapy in the Inpatient Management of Patients With Type 2 Diabetes Undergoing General Surgery (RABBIT 2 Surgery). *Diabetes Care* 34 (2):256-261. Van den Berghe, G., A. Wilmer, I. Milants, P. J. Wouters, B. Bouckaert, F. Bruyninckx, R.Bouillon,

and M. Schetz. 2006. Intensive Insulin Therapy in Mixed Medical/Surgical Intensive

Hypoglycemia in hospitalized patients treated with antihyperglycemic agents. *J* 

Hoekstra. 2006. Predisposing factors for hypoglycemia in the intensive care unit.

Hyper- and Hypoglycemia Among Inpatients With Diabetes: A national survey of

of a Computerized Insulin Order Template in General Medical Inpatients With

The current chapter addresses the problem of hypoglycemia in type 1 diabetes from biomedical and control engineering points of view. It gives a general introduction to the artificial pancreas system, and the risk of hypoglycemia in closed-loop insulin treatment. Then, it provides a review on the state of the art in hypoglycemia control, and the recent approaches in dealing with hypoglycemia in closed-loop artificial pancreas systems. Next, different control techniques that can be used to minimize the risk of hypoglycemia and improve the control outputs are presented.

Since the Diabetes Control and Complications Trial (DCCT), tight glycemic control has been established as the control objective in the treatment of patients with type 1 diabetes mellitus (T1DM) (DCCT Research Group (1993)), except if some contraindication exists. However, there still lacks a universal, efficient and safe system able to normalize the glucose levels of patients. The intensive insulin therapy required to achieve the tight glycemic control, based on the injection of basal and bolus insulin to reproduce its physiological secretion, has as counteraction an increase in the risk of significant and severe hypoglycemia with all their consequences. Therefore, hypoglycemia is considered as one of the major limiting factors in achieving tight glycemic control in T1DM (Cryer (2008)).

With the inability of conventional therapy to achieve satisfactory glycemic control, and the development in continuous glucose monitoring (CGM) systems and the increasing use of insulin pumps, the idea of developing an artificial pancreas is viewed as the ideal solution for glycemic control in T1DM (Bequette (2005); Hovorka et al. (2006); Kumareswaran et al. (2009)). The artificial pancreas is an automated closed-loop system that maintains blood glucose levels within the desired range and prevents hypoglycemia, while minimizing or eliminating the need for patient intervention. The artificial pancreas replaces the *β*-cells functions in glucose sensing and insulin delivery. It consists of three main components (Figure 1): a glucose sensor to measure glucose concentration, a pump for insulin delivery, and a closed-loop control algorithm to bridge between the glucose measurements and the dose of insulin to be delivered. As other medical devices, the architecture of closed-loop

diabetic patient in artificial pancreas studies are: the *Meal model* (Dalla Man et al. (2006; 2007)), *Hovorka model* (Hovorka et al. (2004; 2002)), the minimal model (Bergman et al. (1979)), and *Sorensen model* (Sorensen (1985)). Extensive reviews on available models can be found in Chee & Fernando (2007) and Cobelli et al. (2009). Some of these models have been implemented in simulation environments designed to support the development of the

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 209

Due to the complex nature of the insulin-glucose system, different empirical models have been proposed to relate insulin input to glucose response (see for example Eren-Oruklu et al. (2009b); Finan et al. (2009)). Empirical models develop a functional relationship between insulin and glucose based on empirical observations (i.e. collected patient data). These models do not describe the physiological model, but they explicitly address inter-patient variability since the data-driven model is specific to individual patient dynamics. Empirical models are more suitable for real-time parameter estimation and updating due to their simple structure

The feasibility of closed-loop artificial pancreas systems and their advantage over conventional treatment has been proved in several clinical studies (Atlas et al. (2010); Clarke et al. (2009); Hovorka et al. (2010); Steil et al. (2011; 2006); Weinzimer et al. (2008)), and a wide spectrum of control algorithms has been proposed to close the control loop, including classical and modern control strategies. Many reviews on closed-loop algorithms are available, see for example (Bequette (2005); Chee & Fernando (2007); Doyle III et al. (2007);

However, blood glucose control in T1DM is still one of the difficult control problems to be solved in biomedical engineering. In addition to the inherent complexity of glucoregulatory system, which includes the presence of nonlinearities, and time-varying and patient-specific dynamics, there exist other problems, such as noisy measurements, limitations of the models used to develop the control algorithms, as well as the limitations of the subcutaneous route used for glucose sensing and insulin delivery (e.g. technological and physiological delays and subcutaneous tissues dynamics). The aforementioned challenges make it very difficult to find a general and reliable solution to the nonlinear problem of glycemic control. Therefore, the design of a robust closed-loop control algorithm is an essential step for the progress of the

For closed-loop artificial pancreas system to be optimal and replicate the normal insulin secretion, the insulin therapy should respect the fact that hypoglycemia is not a naturally occurring episode in T1DM. Also, hypoglycemia is believed to be more dangerous in short term than hyperglycemia. Therefore, in order to achieve tight control while not substituting the problem of hyperglycemia for the life-threatening hypoglycemia, the insulin therapy in T1DM should be optimized so that it reduces the risk of hyperglycemic events in both frequency and magnitude, without provoking significant or severe hypoglycemia as a result

Hypoglycemia is the most common complication of insulin therapy in T1DM and continuously limits the efforts to improve glycemic control. Therefore, hypoglycemia prevention should be unavoidably considered among the main objectives in the development of the closed-loop artificial pancreas systems. Severe hypoglycemia episodes are a

closed-loop artificial pancreas (Kovatchev et al. (2009); Wilinska et al. (2010)).

in comparison with complex first order models.

El-Youssef et al. (2009); Takahashi et al. (2008)).

of excessive or ill-timed insulin infusion.

**2. Hypoglycemia in closed-loop artificial pancreas**

**1.2 Control problems**

artificial pancreas.

artificial pancreas should include strict safety measures implemented as safety module or supervision system, to evaluate the performance of the control algorithm and apply fault detection techniques (Doyle III et al. (2007)).

Fig. 1. Artificial pancreas components with patient in the loop. Control algorithm may use feedback or feedforward-feedback control loops

Closed-loop control of blood glucose has been a subject of continuous research for more than 40 years, however, till now no commercially available product does exist. The continuous subcutaneous insulin infusion (CSII) pumps are being widely used, and a number of CGM systems have received regulatory approval (Kumareswaran et al. (2009)). Although the sensors and pumps systems still have some limitations, their use in an open-loop combination resulted in better clinical outcomes over conventional injections therapy (Klonoff (2005); Kumareswaran et al. (2009)). Thus, the primary limitations to develop such an artificial pancreas are the development of reliable closed-loop control algorithms, and the availability of robust and precise glucose sensors. However, recent research in the development of the artificial pancreas suggests that types of the automatic glucose control system are likely to come to market in the near future.

#### **1.1 Patient modeling**

The artificial pancreas automatically regulates the blood glucose level based on the glucose measurements, the insulin infusions and in model-based control approaches, on the mathematical insulin-glucose model (diabetic patient model) used to design the controller. Also, these models are essential for testing and validating the artificial pancreas in simulation studies (i.e. *in-silico*) before putting it into clinical use with real patients. Thus, one essential task in the development of artificial pancreas is to obtain a model of T1DM patient, which can help in the development of a closed-loop control system.

Several models with different structures and degrees of complexity are being used to describe the glucoregulatory system - mainly as insulin-glucose and meal-glucose relationships - in T1DM. Most of these are first principle models represented by differential and algebraic equations and based on existing knowledge and hypotheses regarding the underlying physiological system. Among the models that have been frequently used to represent the diabetic patient in artificial pancreas studies are: the *Meal model* (Dalla Man et al. (2006; 2007)), *Hovorka model* (Hovorka et al. (2004; 2002)), the minimal model (Bergman et al. (1979)), and *Sorensen model* (Sorensen (1985)). Extensive reviews on available models can be found in Chee & Fernando (2007) and Cobelli et al. (2009). Some of these models have been implemented in simulation environments designed to support the development of the closed-loop artificial pancreas (Kovatchev et al. (2009); Wilinska et al. (2010)).

Due to the complex nature of the insulin-glucose system, different empirical models have been proposed to relate insulin input to glucose response (see for example Eren-Oruklu et al. (2009b); Finan et al. (2009)). Empirical models develop a functional relationship between insulin and glucose based on empirical observations (i.e. collected patient data). These models do not describe the physiological model, but they explicitly address inter-patient variability since the data-driven model is specific to individual patient dynamics. Empirical models are more suitable for real-time parameter estimation and updating due to their simple structure in comparison with complex first order models.

#### **1.2 Control problems**

2 Will-be-set-by-IN-TECH

artificial pancreas should include strict safety measures implemented as safety module or supervision system, to evaluate the performance of the control algorithm and apply fault

F eedforward

controller P atient

CGM

Fig. 1. Artificial pancreas components with patient in the loop. Control algorithm may use

Closed-loop control of blood glucose has been a subject of continuous research for more than 40 years, however, till now no commercially available product does exist. The continuous subcutaneous insulin infusion (CSII) pumps are being widely used, and a number of CGM systems have received regulatory approval (Kumareswaran et al. (2009)). Although the sensors and pumps systems still have some limitations, their use in an open-loop combination resulted in better clinical outcomes over conventional injections therapy (Klonoff (2005); Kumareswaran et al. (2009)). Thus, the primary limitations to develop such an artificial pancreas are the development of reliable closed-loop control algorithms, and the availability of robust and precise glucose sensors. However, recent research in the development of the artificial pancreas suggests that types of the automatic glucose control system are likely to

The artificial pancreas automatically regulates the blood glucose level based on the glucose measurements, the insulin infusions and in model-based control approaches, on the mathematical insulin-glucose model (diabetic patient model) used to design the controller. Also, these models are essential for testing and validating the artificial pancreas in simulation studies (i.e. *in-silico*) before putting it into clinical use with real patients. Thus, one essential task in the development of artificial pancreas is to obtain a model of T1DM patient, which can

Several models with different structures and degrees of complexity are being used to describe the glucoregulatory system - mainly as insulin-glucose and meal-glucose relationships - in T1DM. Most of these are first principle models represented by differential and algebraic equations and based on existing knowledge and hypotheses regarding the underlying physiological system. Among the models that have been frequently used to represent the

Pump

insulin

controller Meal

Blood glucose

detection techniques (Doyle III et al. (2007)).

Supervision System

G lucose target

Feedback

feedback or feedforward-feedback control loops

help in the development of a closed-loop control system.

come to market in the near future.

**1.1 Patient modeling**

The feasibility of closed-loop artificial pancreas systems and their advantage over conventional treatment has been proved in several clinical studies (Atlas et al. (2010); Clarke et al. (2009); Hovorka et al. (2010); Steil et al. (2011; 2006); Weinzimer et al. (2008)), and a wide spectrum of control algorithms has been proposed to close the control loop, including classical and modern control strategies. Many reviews on closed-loop algorithms are available, see for example (Bequette (2005); Chee & Fernando (2007); Doyle III et al. (2007); El-Youssef et al. (2009); Takahashi et al. (2008)).

However, blood glucose control in T1DM is still one of the difficult control problems to be solved in biomedical engineering. In addition to the inherent complexity of glucoregulatory system, which includes the presence of nonlinearities, and time-varying and patient-specific dynamics, there exist other problems, such as noisy measurements, limitations of the models used to develop the control algorithms, as well as the limitations of the subcutaneous route used for glucose sensing and insulin delivery (e.g. technological and physiological delays and subcutaneous tissues dynamics). The aforementioned challenges make it very difficult to find a general and reliable solution to the nonlinear problem of glycemic control. Therefore, the design of a robust closed-loop control algorithm is an essential step for the progress of the artificial pancreas.

For closed-loop artificial pancreas system to be optimal and replicate the normal insulin secretion, the insulin therapy should respect the fact that hypoglycemia is not a naturally occurring episode in T1DM. Also, hypoglycemia is believed to be more dangerous in short term than hyperglycemia. Therefore, in order to achieve tight control while not substituting the problem of hyperglycemia for the life-threatening hypoglycemia, the insulin therapy in T1DM should be optimized so that it reduces the risk of hyperglycemic events in both frequency and magnitude, without provoking significant or severe hypoglycemia as a result of excessive or ill-timed insulin infusion.

### **2. Hypoglycemia in closed-loop artificial pancreas**

Hypoglycemia is the most common complication of insulin therapy in T1DM and continuously limits the efforts to improve glycemic control. Therefore, hypoglycemia prevention should be unavoidably considered among the main objectives in the development of the closed-loop artificial pancreas systems. Severe hypoglycemia episodes are a

reduction in overnight hypoglycemia episodes was observed with closed-loop control in comparison with standard therapy. Also, during closed-loop period, the blood glucose level was within the target glycemic range for a longer time period, and the frequency of low

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 211

Beside control algorithms, several algorithms for hypoglycemia detection and prediction are proposed as alarm systems to avoid hypoglycemia. The progress in CGM systems has made it possible to develop such real-time algorithms to reduce the hypoglycemic risk. These algorithms can be used to detect occurring hypoglycemia or warn about a pending hypoglycemic episode. The algorithms are based mainly on a combination of CGM data and a set of defined threshold of glucose and glucose rate of change. Different estimation and prediction approaches (e.g. linear and statistical prediction, Kalman filter optimal estimation, time series, etc.) have been proposed to develop these algorithms (Buckingham et al. (2009); Cameron et al. (2008); Hughes et al. (2010); Palerm et al. (2005); Sparacino et al. (2007)). Nguyen et al. (2009) used a specialized sensor (*Hypoglycemia monitor*) for nocturnal hypoglycemia detection, based on bayesian neural networks approach. The sensor measures specific physiological parameters continuously trying to detect the hypoglycemic events. In Skladnev et al. (2010), a data fusion approach was used to enhance the hypoglycemia alarm of CGM systems. The CGM information (data and alarms) was fused with autonomic nervous system responses that were detected by the specialized *Hypoglycemia monitor*. The data fusion method was able to improve nocturnal hypoglycemia alarms, and

Hypoglycemia prediction/detection algorithms are usually coupled with specific supporting actions to improve their efficiency in preventing hypoglycemia. Different actions have been proposed, such as gradual insulin attenuation (Hughes et al. (2010)), pump suspension (Buckingham et al. (2009); Lee & Bequette (2009)), glucose infusion (Choleau et al. (2002)), and audible (Buckingham et al. (2009); Weinzimer et al. (2008)) or visual (Hughes et al. (2010)) alarms to alert the patient about actual or impending hypoglycemia. The statistical and linear hypoglycemia predictors with pump suspension algorithm proposed in (Buckingham et al. (2009)) were used in a clinical study, and proved to be effective in preventing hypoglycemia

To improve the performance of the closed-loop system, and significantly reduce the risk of hypoglycemia, the control system of the artificial pancreas can be augmented with different control techniques. Such techniques can be introduced either by modifying the controller structure (i.e. internal), or by implementing the additional technique separately (i.e. external component). The increased cost or complexity that could be added to the system by incorporating such techniques can be justified by the improved performance of the system in dealing with life-threatening hypoglycemia. Both external and internal techniques have been tested and proved to provide satisfactory results, and to outperform the stand-alone

Several studies have concluded that model predictive control (widely known as MPC) is expected to be the core of closed-loop control algorithm in the near future artificial pancreas.

without provoking rebound hyperglycemia after the suspension of the pump.

**3. Hypoglycemia prevention by control algorithm improvement**

glucose values was reduced.

closed-loop controllers.

**3.1 Model predictive control**

**2.2 Hypoglycemia alarm systems**

reduced the number of undetected hypoglycemic events.

well-known cause of death in diabetic patients, and are more commonly seen during the night than during the day. Given that the first generations of the artificial pancreas are not expected to achieve complete regulation of the glucose levels during the 24 hours period, first generations of the artificial pancreas might be focusing on critical aspects like preventing hypoglycemia episodes during night (Hovorka et al. (2010)).

Currently, the vast majority of closed-loop artificial pancreas works focuses on the achievement of tight control during daily life conditions (i.e. 24 hours control), and therefore addresses both hyper- and hypoglycemia in fasting and postprandial conditions. Various strategies are employed in these works to avoid fasting, postprandial and nocturnal hypoglycemia. Mostly, the control algorithms use changes in the target blood glucose to adjust the doses of insulin to prevent hypoglycemia (i.e. higher target glucose level during night and postprandial periods) (Eren-Oruklu et al. (2009a); Marchetti et al. (2008); Weinzimer et al. (2008)). In other works, hypoglycemia prediction algorithms were tested, and short-term suspension of insulin pump was used as safety approach when hypoglycemia is predicted (Lee & Bequette (2009)). Also, variations in insulin sensitivity during the day (due to the 24 hours circadian cycle in insulin sensitivity), have been considered in the design of artificial pancreas control algorithms, and used to adjust the basal insulin requirements during the day (Garcia-Gabin et al. (2009); Steil et al. (2003); Wang et al. (2009)).

Another strategy used to avoid hypoglycemia is the double hormone closed-loop system, which uses glucagon infusion in response to low glucose levels. In T1DM, insulin deficiency is often accompanied by the loss of glucagon secretory response to hypoglycemia. Furthermore, insulin therapy causes even more degradation in the functionality of other counterregulatory hormones (Briscoe & Davis (2006)), and consequently, results in higher possibility for hypoglycemic risk. Different artificial pancreas studies have demonstrated that glucagon infusion significantly reduces the risk of insulin-induced hypoglycemia in T1DM (Castle et al. (2010); El-Khatib et al. (2009; 2010); Ward et al. (2008)).

#### **2.1 Overnight hypoglycemia control**

Overnight closed-loop insulin delivery has received great interest because it addresses the critical problem of nocturnal hypoglycemia. Furthermore, prevention of nocturnal hypoglycemia and achieving good control overnight can help in improving the quality of glycemic control during the day (Hovorka et al. (2010)) (e.g. starting the day with acceptable glucose levels). A number of clinical and *in-silico* studies attempts to deal with the hypoglycemia prevention - mainly nocturnal hypoglycemia - as the primary control objective. In (Wilinska et al. (2009)), a manual closed-loop insulin delivery system was employed during night period using model predictive control (MPC) algorithm and CGM measurements (CGM readings were provided to the MPC by medical staff), and aimed at regulating glucose level overnight to avoid nocturnal hypoglycemia. In (Hovorka et al. (2010)), the system was tested in a clinical study with children and adolescents. Earlier version of this MPC algorithm was tested in previous clinical study to evaluate its control and prediction performance during fasting conditions (Shaller et al. (2006)). An automated closed-loop insulin delivery system was tested in a multinational clinical trial (Bruttomesso et al. (2009); Clarke et al. (2009)). The system used a personalized MPC algorithm developed in (Magni et al. (2007)). The system was developed completely *in-silico* and then tested in the clinical trial.

The studies concluded that the MPC algorithm is well suited for glucose control under fasting and overnight conditions in T1DM patients. The studies showed that the artificial pancreas is superior to open-loop control in preventing overnight hypoglycemia where significant reduction in overnight hypoglycemia episodes was observed with closed-loop control in comparison with standard therapy. Also, during closed-loop period, the blood glucose level was within the target glycemic range for a longer time period, and the frequency of low glucose values was reduced.

#### **2.2 Hypoglycemia alarm systems**

4 Will-be-set-by-IN-TECH

well-known cause of death in diabetic patients, and are more commonly seen during the night than during the day. Given that the first generations of the artificial pancreas are not expected to achieve complete regulation of the glucose levels during the 24 hours period, first generations of the artificial pancreas might be focusing on critical aspects like preventing

Currently, the vast majority of closed-loop artificial pancreas works focuses on the achievement of tight control during daily life conditions (i.e. 24 hours control), and therefore addresses both hyper- and hypoglycemia in fasting and postprandial conditions. Various strategies are employed in these works to avoid fasting, postprandial and nocturnal hypoglycemia. Mostly, the control algorithms use changes in the target blood glucose to adjust the doses of insulin to prevent hypoglycemia (i.e. higher target glucose level during night and postprandial periods) (Eren-Oruklu et al. (2009a); Marchetti et al. (2008); Weinzimer et al. (2008)). In other works, hypoglycemia prediction algorithms were tested, and short-term suspension of insulin pump was used as safety approach when hypoglycemia is predicted (Lee & Bequette (2009)). Also, variations in insulin sensitivity during the day (due to the 24 hours circadian cycle in insulin sensitivity), have been considered in the design of artificial pancreas control algorithms, and used to adjust the basal insulin requirements during the day

Another strategy used to avoid hypoglycemia is the double hormone closed-loop system, which uses glucagon infusion in response to low glucose levels. In T1DM, insulin deficiency is often accompanied by the loss of glucagon secretory response to hypoglycemia. Furthermore, insulin therapy causes even more degradation in the functionality of other counterregulatory hormones (Briscoe & Davis (2006)), and consequently, results in higher possibility for hypoglycemic risk. Different artificial pancreas studies have demonstrated that glucagon infusion significantly reduces the risk of insulin-induced hypoglycemia in T1DM

Overnight closed-loop insulin delivery has received great interest because it addresses the critical problem of nocturnal hypoglycemia. Furthermore, prevention of nocturnal hypoglycemia and achieving good control overnight can help in improving the quality of glycemic control during the day (Hovorka et al. (2010)) (e.g. starting the day with acceptable glucose levels). A number of clinical and *in-silico* studies attempts to deal with the hypoglycemia prevention - mainly nocturnal hypoglycemia - as the primary control objective. In (Wilinska et al. (2009)), a manual closed-loop insulin delivery system was employed during night period using model predictive control (MPC) algorithm and CGM measurements (CGM readings were provided to the MPC by medical staff), and aimed at regulating glucose level overnight to avoid nocturnal hypoglycemia. In (Hovorka et al. (2010)), the system was tested in a clinical study with children and adolescents. Earlier version of this MPC algorithm was tested in previous clinical study to evaluate its control and prediction performance during fasting conditions (Shaller et al. (2006)). An automated closed-loop insulin delivery system was tested in a multinational clinical trial (Bruttomesso et al. (2009); Clarke et al. (2009)). The system used a personalized MPC algorithm developed in (Magni et al. (2007)). The system

The studies concluded that the MPC algorithm is well suited for glucose control under fasting and overnight conditions in T1DM patients. The studies showed that the artificial pancreas is superior to open-loop control in preventing overnight hypoglycemia where significant

hypoglycemia episodes during night (Hovorka et al. (2010)).

(Garcia-Gabin et al. (2009); Steil et al. (2003); Wang et al. (2009)).

(Castle et al. (2010); El-Khatib et al. (2009; 2010); Ward et al. (2008)).

was developed completely *in-silico* and then tested in the clinical trial.

**2.1 Overnight hypoglycemia control**

Beside control algorithms, several algorithms for hypoglycemia detection and prediction are proposed as alarm systems to avoid hypoglycemia. The progress in CGM systems has made it possible to develop such real-time algorithms to reduce the hypoglycemic risk. These algorithms can be used to detect occurring hypoglycemia or warn about a pending hypoglycemic episode. The algorithms are based mainly on a combination of CGM data and a set of defined threshold of glucose and glucose rate of change. Different estimation and prediction approaches (e.g. linear and statistical prediction, Kalman filter optimal estimation, time series, etc.) have been proposed to develop these algorithms (Buckingham et al. (2009); Cameron et al. (2008); Hughes et al. (2010); Palerm et al. (2005); Sparacino et al. (2007)). Nguyen et al. (2009) used a specialized sensor (*Hypoglycemia monitor*) for nocturnal hypoglycemia detection, based on bayesian neural networks approach. The sensor measures specific physiological parameters continuously trying to detect the hypoglycemic events. In Skladnev et al. (2010), a data fusion approach was used to enhance the hypoglycemia alarm of CGM systems. The CGM information (data and alarms) was fused with autonomic nervous system responses that were detected by the specialized *Hypoglycemia monitor*. The data fusion method was able to improve nocturnal hypoglycemia alarms, and reduced the number of undetected hypoglycemic events.

Hypoglycemia prediction/detection algorithms are usually coupled with specific supporting actions to improve their efficiency in preventing hypoglycemia. Different actions have been proposed, such as gradual insulin attenuation (Hughes et al. (2010)), pump suspension (Buckingham et al. (2009); Lee & Bequette (2009)), glucose infusion (Choleau et al. (2002)), and audible (Buckingham et al. (2009); Weinzimer et al. (2008)) or visual (Hughes et al. (2010)) alarms to alert the patient about actual or impending hypoglycemia. The statistical and linear hypoglycemia predictors with pump suspension algorithm proposed in (Buckingham et al. (2009)) were used in a clinical study, and proved to be effective in preventing hypoglycemia without provoking rebound hyperglycemia after the suspension of the pump.

### **3. Hypoglycemia prevention by control algorithm improvement**

To improve the performance of the closed-loop system, and significantly reduce the risk of hypoglycemia, the control system of the artificial pancreas can be augmented with different control techniques. Such techniques can be introduced either by modifying the controller structure (i.e. internal), or by implementing the additional technique separately (i.e. external component). The increased cost or complexity that could be added to the system by incorporating such techniques can be justified by the improved performance of the system in dealing with life-threatening hypoglycemia. Both external and internal techniques have been tested and proved to provide satisfactory results, and to outperform the stand-alone closed-loop controllers.

### **3.1 Model predictive control**

Several studies have concluded that model predictive control (widely known as MPC) is expected to be the core of closed-loop control algorithm in the near future artificial pancreas.

Subject to the following constraints:

*umin* ≤ *uk* ≤ *umax* Δ*umin* ≤ Δ*uk* ≤ Δ*umax*

where *y*ˆ(*k* + *j*|*k*) is the j-step prediction of the output on data up to instant k, *r*(*k* + *j*) is the target glucose level, Δ*u* is the insulin input increment, *Np* and *Nu* are the prediction and control horizons, and *w*Δ*u*, *w<sup>y</sup>* are weights on the insulin increments and the error between *y*(*k*) and *r*(*k*) respectively, *ε* is a slack variable used for output constraints softening (to avoid infeasibility problems in the optimization), *q* is the weight on the slack variable *ε*, *umin*/*max*, Δ*umin*/*max* and *ymin*/*max* are the constraints imposed on the input, input increments, and

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 213

The cost function in equation (1) is asymmetric in the sense that the lower and upper output constraints are subjected to unequal relaxation bands and therefore, the constraints have different levels of softness. The unequal softness levels could be achieved by introducing the nonnegative relaxation variables Φ*min*, Φ*max* which represent the concern for relaxing the corresponding constraint; the larger Φ, the softer the constraint. MPC with asymmetric cost function was tested with different diabetic patient models, and showed an excellent ability to minimize the hypoglycemic events, especially in postprandial period (Abu-Rmileh & Garcia-Gabin (2010a;b); Kirchsteiger & Del Re (2009)). Kirchsteiger & Del Re (2009) give a comparison between symmetric and asymmetric cost function MPC's, where the

In Dua et al. (2009), a multi-programming MPC is used, and provided with different techniques to avoid hypoglycemia. In the multi-programming approach, the optimization problem in MPC is solved by searching for optimal solution within some valid regions (search regions) defined by the constraints and the parameters of the cost function. The main advantage of the multi-parametric MPC is that it provides the same performance as traditional MPC with lower computational load. The controller is provided with asymmetric cost function, and higher priority is given to the satisfaction of constraints imposed on hypoglycemia. Another type of asymmetric performance is presented in Grosman et al. (2010) to minimize the undesirable hypoglycemic and hyperglycemic events. The proposed MPC uses a glycemic zone rather than a fixed glucose level as a target (Zone-MPC). Three different zones are defined (permitted, lower, and upper zones), where the control objective is adjusting

Gain scheduling (GS) is a well-known technique for controlling nonlinear systems by linear controllers. Briefly, GS is one of the simplest forms of adaptive control that employs different control structures in the different operating ranges of the nonlinear system. In glucose control, GS was inspired from the natural pancreas where the level of insulin activity varies between different glycemic ranges; being dominant in the hyperglycemic range, in balance with glucagon action in normoglycemia, and almost inactive in the hypoglycemic range where

From an engineering perspective, a simple nonlinearity test (e.g. steady state insulin-glucose relationship) can be used to show that insulin has a nonlinear effect on blood glucose in different glycemic ranges (see Figure 2). Linear control algorithms are intended to control

output respectively, and Φ*min*, Φ*max* are the relaxation variables.

latter shows superior performance in avoiding hypoglycemia.

the insulin input to maintain glucose level within the permitted zone.

**3.3 Gain scheduling**

glucagon is dominant.

*ymin* − *ε*Φ*min* ≤ *yk* ≤ *ymax* + *ε*Φ*max* (2)

Therefore, MPC is discussed in some details in this chapter. MPC is a control strategy that has developed considerably over the past few decades. Basically, MPC is based on a model of the system to be controlled. The model is used to predict the future system outputs, based on the past and current values and on the proposed optimal future control actions. These actions are calculated by optimizing a cost function where the future tracking error is considered, as well as the system constraints if any (Maciejowski (2002)). MPC employs a receding horizon strategy; repeated displacement of the time horizon, while only applying the first control signal in the calculated sequence at each time step, with the rest of the sequence being discarded.

MPC has many virtues that make it a competitive candidate for the blood glucose control problem: (1) The prediction nature of MPC allows for anticipatory and careful insulin delivery to avoid large fluctuations in glucose levels. Such feature is important for avoiding overdosing and hypoglycemic risk. (2) The ability of MPC to handle constraints on system inputs and outputs is a major advantage of MPC over other control strategies. These constraints are very critical when dealing with the human body, and allow to satisfy hardware specifications of the insulin pump. (3) The applicability of MPC to systems with time delays can be useful to overcome the physiological and technological delays associated with the subcutaneous route. (4) MPC allows the introduction of feedforward control action to compensate for known sources of disturbance affecting the system, such as meal intake. These advantages of MPC over other control strategies have promoted the use of MPC in the field of insulin delivery. Different MPC schemes are being used in artificial pancreas research, where the applicability of such control strategy has been demonstrated in *in-silico* studies (see for instance (Abu-Rmileh et al., 2010a; Dua et al., 2009; Grosman et al., 2010; Hovorka et al., 2004; Lee & Bequette, 2009; Magni et al., 2007; Parker et al., 1999)), and clinical trials as mentioned earlier.

#### **3.2 Unequal penalization**

Closed-loop control schemes can be designed so that unequal penalties are used upon hyperglycemia and hypoglycemia. The reason for such unequal penalties is that in diabetes therapy, the performance requirement of a controller has asymmetric nature, as hypoglycemic events are much less tolerable than hyperglycemia. Since hypoglycemia is believed to be more life-threatening in the short term, the control algorithm should be more aggressive in avoiding hypoglycemic episodes than in correcting hyperglycemic events.

MPC is one control strategy that permits to incorporate this kind of unequal penalization. To achieve such requirements of asymmetrical response, an asymmetric cost function is used in the optimization algorithm in MPC. The asymmetric cost function imposes different weight on hypoglycemia than on hyperglycemia, in contrast to conventional cost functions that impose the same weight on hypoglycemic and hyperglycemic events. As stated before, MPC calculates the insulin control action *uk*, by optimizing a quadratic cost function, penalizing predicted output deviations and control signal along some prediction horizons. The asymmetric cost function has the form:

$$\min\_{\Delta u} f = \sum\_{j=1}^{N\_p} \left\| w^y (\mathcal{Y}(k+j|k) - r(k+j)) \right\|^2 + \sum\_{j=1}^{N\_u} \left\| w^{\Delta u} (\Delta u(k+j|k)) \right\|^2 + q\varepsilon^2 \tag{1}$$

Subject to the following constraints:

6 Will-be-set-by-IN-TECH

Therefore, MPC is discussed in some details in this chapter. MPC is a control strategy that has developed considerably over the past few decades. Basically, MPC is based on a model of the system to be controlled. The model is used to predict the future system outputs, based on the past and current values and on the proposed optimal future control actions. These actions are calculated by optimizing a cost function where the future tracking error is considered, as well as the system constraints if any (Maciejowski (2002)). MPC employs a receding horizon strategy; repeated displacement of the time horizon, while only applying the first control signal in the calculated sequence at each time step, with the rest of the sequence

MPC has many virtues that make it a competitive candidate for the blood glucose control problem: (1) The prediction nature of MPC allows for anticipatory and careful insulin delivery to avoid large fluctuations in glucose levels. Such feature is important for avoiding overdosing and hypoglycemic risk. (2) The ability of MPC to handle constraints on system inputs and outputs is a major advantage of MPC over other control strategies. These constraints are very critical when dealing with the human body, and allow to satisfy hardware specifications of the insulin pump. (3) The applicability of MPC to systems with time delays can be useful to overcome the physiological and technological delays associated with the subcutaneous route. (4) MPC allows the introduction of feedforward control action to compensate for known sources of disturbance affecting the system, such as meal intake. These advantages of MPC over other control strategies have promoted the use of MPC in the field of insulin delivery. Different MPC schemes are being used in artificial pancreas research, where the applicability of such control strategy has been demonstrated in *in-silico* studies (see for instance (Abu-Rmileh et al., 2010a; Dua et al., 2009; Grosman et al., 2010; Hovorka et al., 2004; Lee & Bequette, 2009; Magni et al., 2007; Parker et al., 1999)), and clinical trials as mentioned

Closed-loop control schemes can be designed so that unequal penalties are used upon hyperglycemia and hypoglycemia. The reason for such unequal penalties is that in diabetes therapy, the performance requirement of a controller has asymmetric nature, as hypoglycemic events are much less tolerable than hyperglycemia. Since hypoglycemia is believed to be more life-threatening in the short term, the control algorithm should be more aggressive in avoiding

MPC is one control strategy that permits to incorporate this kind of unequal penalization. To achieve such requirements of asymmetrical response, an asymmetric cost function is used in the optimization algorithm in MPC. The asymmetric cost function imposes different weight on hypoglycemia than on hyperglycemia, in contrast to conventional cost functions that impose the same weight on hypoglycemic and hyperglycemic events. As stated before, MPC calculates the insulin control action *uk*, by optimizing a quadratic cost function, penalizing predicted output deviations and control signal along some prediction horizons.

> *Nu* ∑ *j*=1

�*w*Δ*u*(Δ*u*(*<sup>k</sup>* <sup>+</sup> *<sup>j</sup>*|*k*)�<sup>2</sup> <sup>+</sup> *<sup>q</sup><sup>ε</sup>*

<sup>2</sup> (1)

hypoglycemic episodes than in correcting hyperglycemic events.

�*wy*(*y*ˆ(*<sup>k</sup>* <sup>+</sup> *<sup>j</sup>*|*k*) <sup>−</sup> *<sup>r</sup>*(*<sup>k</sup>* <sup>+</sup> *<sup>j</sup>*))�<sup>2</sup> <sup>+</sup>

The asymmetric cost function has the form:

*Np* ∑ *j*=1

being discarded.

earlier.

**3.2 Unequal penalization**

min <sup>Δ</sup>*<sup>u</sup> <sup>J</sup>* <sup>=</sup>

$$\begin{aligned} \mathfrak{u}\_{\min} \le \mathfrak{u}\_k \le \mathfrak{u}\_{\max} \\ \Delta \mathfrak{u}\_{\min} \le \Delta \mathfrak{u}\_k \le \Delta \mathfrak{u}\_{\max} \\ \mathfrak{y}\_{\min} - \varepsilon \Phi\_{\min} \le \mathfrak{y}\_k \le \mathfrak{y}\_{\max} + \varepsilon \Phi\_{\max} \end{aligned} \tag{2}$$

where *y*ˆ(*k* + *j*|*k*) is the j-step prediction of the output on data up to instant k, *r*(*k* + *j*) is the target glucose level, Δ*u* is the insulin input increment, *Np* and *Nu* are the prediction and control horizons, and *w*Δ*u*, *w<sup>y</sup>* are weights on the insulin increments and the error between *y*(*k*) and *r*(*k*) respectively, *ε* is a slack variable used for output constraints softening (to avoid infeasibility problems in the optimization), *q* is the weight on the slack variable *ε*, *umin*/*max*, Δ*umin*/*max* and *ymin*/*max* are the constraints imposed on the input, input increments, and output respectively, and Φ*min*, Φ*max* are the relaxation variables.

The cost function in equation (1) is asymmetric in the sense that the lower and upper output constraints are subjected to unequal relaxation bands and therefore, the constraints have different levels of softness. The unequal softness levels could be achieved by introducing the nonnegative relaxation variables Φ*min*, Φ*max* which represent the concern for relaxing the corresponding constraint; the larger Φ, the softer the constraint. MPC with asymmetric cost function was tested with different diabetic patient models, and showed an excellent ability to minimize the hypoglycemic events, especially in postprandial period (Abu-Rmileh & Garcia-Gabin (2010a;b); Kirchsteiger & Del Re (2009)). Kirchsteiger & Del Re (2009) give a comparison between symmetric and asymmetric cost function MPC's, where the latter shows superior performance in avoiding hypoglycemia.

In Dua et al. (2009), a multi-programming MPC is used, and provided with different techniques to avoid hypoglycemia. In the multi-programming approach, the optimization problem in MPC is solved by searching for optimal solution within some valid regions (search regions) defined by the constraints and the parameters of the cost function. The main advantage of the multi-parametric MPC is that it provides the same performance as traditional MPC with lower computational load. The controller is provided with asymmetric cost function, and higher priority is given to the satisfaction of constraints imposed on hypoglycemia. Another type of asymmetric performance is presented in Grosman et al. (2010) to minimize the undesirable hypoglycemic and hyperglycemic events. The proposed MPC uses a glycemic zone rather than a fixed glucose level as a target (Zone-MPC). Three different zones are defined (permitted, lower, and upper zones), where the control objective is adjusting the insulin input to maintain glucose level within the permitted zone.

#### **3.3 Gain scheduling**

Gain scheduling (GS) is a well-known technique for controlling nonlinear systems by linear controllers. Briefly, GS is one of the simplest forms of adaptive control that employs different control structures in the different operating ranges of the nonlinear system. In glucose control, GS was inspired from the natural pancreas where the level of insulin activity varies between different glycemic ranges; being dominant in the hyperglycemic range, in balance with glucagon action in normoglycemia, and almost inactive in the hypoglycemic range where glucagon is dominant.

From an engineering perspective, a simple nonlinearity test (e.g. steady state insulin-glucose relationship) can be used to show that insulin has a nonlinear effect on blood glucose in different glycemic ranges (see Figure 2). Linear control algorithms are intended to control

M ultiple linear controllers

target u(t)

a feedback signal, and to GS to select the controller to be used.

additional meal-time insulin bolus (Figure 1).

*u f f* is calculated as:

compensation.

G lucose

G ain Scheduling

Selectcontroller

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 215

Fig. 3. Gain scheduling control scheme; the CGM output is delivered to the controllers box as

To avoid the limitation of purely reactive feedback control action and improve the controller response against meal effect, feedforward control (i.e. meal announcement) can be used. Feedforward is a well-known control technique used to eliminate the disturbance effect when the source of disturbance can be measured. In blood glucose control, the meal intake can be viewed as a known source of disturbance, and feedforward control can be used for meal announcement. In case information is given to the artificial pancreas system about the upcoming meal (size and time), a feedforward scheme may be implemented to deliver

For the design of the feedforward controller, the effect of meal on blood glucose level should be modeled. The system model (insulin-glucose) in the feedforward element describes or predicts how each change in insulin will affect glucose, while the disturbance model (meal-glucose) is used to describe or predict how each change in meal will affect glucose. Let *Gs* and *Gd* be the system and disturbance models respectively, the feedforward control

*Gs*

Feedforward controllers can range from simple scaling multipliers (static feedforward) to sophisticated differential equations (dynamic feedforward). Dynamic models give a better description of actual system and disturbance behaviors, often achieving improved disturbance rejection performance. However, the dynamic feedforward can be difficult to obtain and implement. In specific control algorithms such as MPC, the feedforward control signal can be calculated by the controller itself rather than using a separate feedforward controller. If the meal effect is included in the prediction model of the MPC, the controller predicts the future glucose levels as a function of insulin-glucose dynamics, CGM measurements, and meal information. Consequently, the meal effect on blood glucose will be considered in calculating the future insulin dose (i.e. predictive feedforward). In this controller configuration, the insulin dose has two parts: feedback insulin delivered in fasting conditions, and feedforward insulin bolus used at meal time to obtain better meal

*<sup>u</sup> f f* <sup>=</sup> <sup>−</sup> *Gd*

Patient

Blood glucose

× *Meal* (3)

CGM

linear systems, and they usually offer poor results when used to control nonlinear systems in regions far from where the linear model used was obtained. Therefore, nonlinear control or multiple linear controllers should be applied to handle each glycemic range separately and mimic the natural pancreas secretions. The use of multiple linear controllers by gain scheduling approach is discussed here, while nonlinear control is addressed later in this chapter.

Fig. 2. Nonlinear steady-state insulin-glucose behavior in different models of diabetic patients

The idea behind using the GS strategy in artificial pancreas is to use multiple linear controllers to deal with the system nonlinear behavior and maintain the ability of handling each glycemic range separately according to its dynamics. Since most of the closed-loop control strategies use insulin only, the control algorithm should provide the different levels of insulin activity in different glycemic ranges by employing the GS technique. GS scheme requires the assignation of scheduling parameters that can be used to select the suitable linear controller for each range. The GS strategy overcomes the limitations of the linear control approach which is only valid in the neighborhood of a single operating point, and provides a performance similar to nonlinear controllers with lower complexity.

A simplified diagram of the GS control is shown in Figure 3. As it can be seen in the figure, the measured glucose level is used as a scheduling variable, and also delivered to the controllers box as feedback signal. The controllers receive the desired glucose level (glucose target) to calculate the required insulin based on the difference between target glucose and CGM measurements, and the glycemic range defined by the GS selection. A control approach combining linear MPC with GS was tested in (Abu-Rmileh & Garcia-Gabin (2010a;b)), and proved to enhance the performance of the closed-loop controller in avoiding hypoglycemia.

#### **3.4 Meal announcement**

Regulation of blood glucose level after a meal is one of the main challenges for the fully developed artificial pancreas. Meals usually lead to a significant glucose flux into the blood stream. If feedback control is used to eliminate the meal effect, the controller reacts only after a rise in glucose has occurred and been detected by the CGM sensor. Elevated glucose level can lead to insulin overdosing, resulting in postprandial hypoglycemia (Steil et al. (2006)).

8 Will-be-set-by-IN-TECH

linear systems, and they usually offer poor results when used to control nonlinear systems in regions far from where the linear model used was obtained. Therefore, nonlinear control or multiple linear controllers should be applied to handle each glycemic range separately and mimic the natural pancreas secretions. The use of multiple linear controllers by gain scheduling approach is discussed here, while nonlinear control is addressed later in this

Blood glucose (mg/dL)

Fig. 2. Nonlinear steady-state insulin-glucose behavior in different models of diabetic

The idea behind using the GS strategy in artificial pancreas is to use multiple linear controllers to deal with the system nonlinear behavior and maintain the ability of handling each glycemic range separately according to its dynamics. Since most of the closed-loop control strategies use insulin only, the control algorithm should provide the different levels of insulin activity in different glycemic ranges by employing the GS technique. GS scheme requires the assignation of scheduling parameters that can be used to select the suitable linear controller for each range. The GS strategy overcomes the limitations of the linear control approach which is only valid in the neighborhood of a single operating point, and provides a performance similar to nonlinear

A simplified diagram of the GS control is shown in Figure 3. As it can be seen in the figure, the measured glucose level is used as a scheduling variable, and also delivered to the controllers box as feedback signal. The controllers receive the desired glucose level (glucose target) to calculate the required insulin based on the difference between target glucose and CGM measurements, and the glycemic range defined by the GS selection. A control approach combining linear MPC with GS was tested in (Abu-Rmileh & Garcia-Gabin (2010a;b)), and proved to enhance the performance of the closed-loop controller in avoiding hypoglycemia.

Regulation of blood glucose level after a meal is one of the main challenges for the fully developed artificial pancreas. Meals usually lead to a significant glucose flux into the blood stream. If feedback control is used to eliminate the meal effect, the controller reacts only after a rise in glucose has occurred and been detected by the CGM sensor. Elevated glucose level can lead to insulin overdosing, resulting in postprandial hypoglycemia (Steil et al. (2006)).

10 15 20 25

Basal insulin (mU/ min)

(b) Strong nonlinearity (*Hovorka model*)

14 15 16 17 18 19

Basal insulin (mU/ min)

(a) Smooth nonlinearity (*Meal model*)

chapter.

controllers with lower complexity.

**3.4 Meal announcement**

Blood

patients

glucose(mg/dL)

Fig. 3. Gain scheduling control scheme; the CGM output is delivered to the controllers box as a feedback signal, and to GS to select the controller to be used.

To avoid the limitation of purely reactive feedback control action and improve the controller response against meal effect, feedforward control (i.e. meal announcement) can be used. Feedforward is a well-known control technique used to eliminate the disturbance effect when the source of disturbance can be measured. In blood glucose control, the meal intake can be viewed as a known source of disturbance, and feedforward control can be used for meal announcement. In case information is given to the artificial pancreas system about the upcoming meal (size and time), a feedforward scheme may be implemented to deliver additional meal-time insulin bolus (Figure 1).

For the design of the feedforward controller, the effect of meal on blood glucose level should be modeled. The system model (insulin-glucose) in the feedforward element describes or predicts how each change in insulin will affect glucose, while the disturbance model (meal-glucose) is used to describe or predict how each change in meal will affect glucose. Let *Gs* and *Gd* be the system and disturbance models respectively, the feedforward control *u f f* is calculated as:

$$
\mu\_{ff} = -\frac{\mathbf{G}\_d}{\mathbf{G}\_s} \times \text{Mean} \tag{3}
$$

Feedforward controllers can range from simple scaling multipliers (static feedforward) to sophisticated differential equations (dynamic feedforward). Dynamic models give a better description of actual system and disturbance behaviors, often achieving improved disturbance rejection performance. However, the dynamic feedforward can be difficult to obtain and implement. In specific control algorithms such as MPC, the feedforward control signal can be calculated by the controller itself rather than using a separate feedforward controller. If the meal effect is included in the prediction model of the MPC, the controller predicts the future glucose levels as a function of insulin-glucose dynamics, CGM measurements, and meal information. Consequently, the meal effect on blood glucose will be considered in calculating the future insulin dose (i.e. predictive feedforward). In this controller configuration, the insulin dose has two parts: feedback insulin delivered in fasting conditions, and feedforward insulin bolus used at meal time to obtain better meal compensation.

hypoglycemic risk produced by erroneous insulin bolus or skipped meal, which may occur in

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 217

Three main types of meal detection algorithms currently exist. A voting scheme is used in (Dassau et al. (2008)) to detect meals based on a combination of four different methods for calculating glucose rates of change. Another algorithm is proposed in (Lee & Bequette (2009); Lee et al. (2009)), where the meal detection algorithm is developed by using a finite impulse response filter and a set of threshold values. The algorithm estimates the meal size at the time of detection. Since the main objective of the development of meal detection algorithms is the application to closed-loop artificial pancreas, Lee & Bequette (2009) tested the design algorithm in combination with a MPC closed-loop controller, and demonstrated that meal detection strategy is efficient and outperforms the stand-alone feedback control scheme. Cameron et al. (2009) presented a probabilistic and evolving algorithm to detect the meal and predict its shape, and to estimate the total appearance of glucose from the meal. The algorithm has proved to enhance the meal-compensation ability of the feedback controller.

It is well-known that the time delay in the subcutaneous route is a major challenge in the development of the artificial pancreas (Hovorka (2006)). Both physiological and technological delays exist in glucose sensing and insulin delivery. Such time delays can result in poorly controlled glucose since hypoglycemia can be induced and remains undetected for a significant time period. In an attempt to eliminate or minimize the effect of time delay, closed-loop control structures with time-delay compensation features can be used to improve the control outputs and reduce the hypoglycemic risk produced by physiological

Smith predictor structure is a control scheme that presents good properties in controlling systems with long time delay. The idea behind Smith predictor is to incorporate the system model within the closed-loop control structure (i.e. the system model becomes an explicit part of the controller). Thus, the design of Smith predictor scheme requires a model of the system dynamics and an estimate of the system time delay *t*0. In the Smith predictor scheme, there are two parallel paths for the control signal *u*(*t*) (see Figure 5); one passing through the real system (the patient), and one passing through the model of the system *Gs*. The function of the parallel path containing the model is to generate the difference *em*(*t*) between the actual system output *y*(*t*) and a model-based prediction of the control signal effect on the system output *ym*(*t*). The Smith predictor uses the model to predict the delay-free response of the

*<sup>m</sup>*(*t*). Then, it compares this prediction to the target glucose level *r*(*t*) to decide

what control actions are needed. To avoid drifting and reject external disturbances, the Smith predictor also compares the actual system output with a prediction that takes the time delay into account. The error *em*(*t*) contributes to the overall error signal *e*(*t*) delivered to the

The Smith predictor structure has been recently used in artificial pancreas studies (Abu-Rmileh et al. (2010a;b)). With an initial estimation of the time delay, the Smith predictor shows the ability to minimize the effect of time delays and the associated risk of hypoglycemia, and to enhance the controller performance. As mentioned before, the MPC strategy, which has been extensively studied in artificial pancreas applications, is another competitive control algorithm with inherited ability to deal with system time delays (Hovorka

the case of feedforward meal announcement.

**3.6 Time delay compensation**

and technological delays.

system *y*−

(2006)).

feedback controller.

The different configurations of feedforward (static, dynamic, and predictive) are being used in the artificial pancreas research, and their feasibility in improving the overall controller performance has been demonstrated in different clinical and simulation studies (Abu-Rmileh & Garcia-Gabin (2010a;b); Abu-Rmileh et al. (2010b); Lee & Bequette (2009); Marchetti et al. (2008); Weinzimer et al. (2008)). Since the feedforward action starts to deliver insulin before the meal effect appears in the CGM feedback loop, lower fluctuations in glucose levels are observed, with higher percentage of time within the acceptable glycemic range. An example of the improved performance achieved with feedforward control is shown in Figure 4. Finally, it should be mentioned that meal announcement must be done carefully, since an excess of insulin or badly-timed bolus may induce undesirable hypoglycemia episodes.

Fig. 4. Feedback (FB) vs. feedforward-feedback (FF-FB) control performance, (a) glucose level (b) insulin input

#### **3.5 Meal detection**

Beside feedback and feedforward control, meal detection techniques can be used to deal with meal challenge. Although feedforward-feedback control achieves better results than feedback alone, it is not uncommon that patients forget to announce upcoming meals. Therefore, a system for meal compensation that does not require information from the patient, would be preferable. The CGM measurements along with a set of thresholds on glucose levels and glucose rates of change (i.e. first and second derivative), can be used to build meal detection/compensation algorithms. When a meal is detected, the algorithm can be used to initiate extra meal-time insulin dose, or to activate an alarm for the patient. The meal-time dose can be delivered as insulin bolus or micro boluses, or a gain scheduling scheme can be used to adjust the controller output when a meal is detected. Meal detection and CGM-activated insulin dose remove the need for patient's interventions, and make the closed-loop artificial pancreas fully automatic. Meal detection algorithms also reduce the hypoglycemic risk produced by erroneous insulin bolus or skipped meal, which may occur in the case of feedforward meal announcement.

Three main types of meal detection algorithms currently exist. A voting scheme is used in (Dassau et al. (2008)) to detect meals based on a combination of four different methods for calculating glucose rates of change. Another algorithm is proposed in (Lee & Bequette (2009); Lee et al. (2009)), where the meal detection algorithm is developed by using a finite impulse response filter and a set of threshold values. The algorithm estimates the meal size at the time of detection. Since the main objective of the development of meal detection algorithms is the application to closed-loop artificial pancreas, Lee & Bequette (2009) tested the design algorithm in combination with a MPC closed-loop controller, and demonstrated that meal detection strategy is efficient and outperforms the stand-alone feedback control scheme. Cameron et al. (2009) presented a probabilistic and evolving algorithm to detect the meal and predict its shape, and to estimate the total appearance of glucose from the meal. The algorithm has proved to enhance the meal-compensation ability of the feedback controller.

#### **3.6 Time delay compensation**

10 Will-be-set-by-IN-TECH

The different configurations of feedforward (static, dynamic, and predictive) are being used in the artificial pancreas research, and their feasibility in improving the overall controller performance has been demonstrated in different clinical and simulation studies (Abu-Rmileh & Garcia-Gabin (2010a;b); Abu-Rmileh et al. (2010b); Lee & Bequette (2009); Marchetti et al. (2008); Weinzimer et al. (2008)). Since the feedforward action starts to deliver insulin before the meal effect appears in the CGM feedback loop, lower fluctuations in glucose levels are observed, with higher percentage of time within the acceptable glycemic range. An example of the improved performance achieved with feedforward control is shown in Figure 4. Finally, it should be mentioned that meal announcement must be done carefully, since an excess of insulin or badly-timed bolus may induce undesirable hypoglycemia episodes.

10 15 20 25 30 35

FB control FF−FB control Target glucose

FB control FF−FB control Meal−time bolus

10 15 20 25 30 35

Day t ime (hours)

Fig. 4. Feedback (FB) vs. feedforward-feedback (FF-FB) control performance, (a) glucose

Beside feedback and feedforward control, meal detection techniques can be used to deal with meal challenge. Although feedforward-feedback control achieves better results than feedback alone, it is not uncommon that patients forget to announce upcoming meals. Therefore, a system for meal compensation that does not require information from the patient, would be preferable. The CGM measurements along with a set of thresholds on glucose levels and glucose rates of change (i.e. first and second derivative), can be used to build meal detection/compensation algorithms. When a meal is detected, the algorithm can be used to initiate extra meal-time insulin dose, or to activate an alarm for the patient. The meal-time dose can be delivered as insulin bolus or micro boluses, or a gain scheduling scheme can be used to adjust the controller output when a meal is detected. Meal detection and CGM-activated insulin dose remove the need for patient's interventions, and make the closed-loop artificial pancreas fully automatic. Meal detection algorithms also reduce the

Day t ime (hours)

100

5

Insulin input (Units/h our)

level (b) insulin input

**3.5 Meal detection**

10

15

(b)

Glucose level(mg/d L)

150

200

(a)

It is well-known that the time delay in the subcutaneous route is a major challenge in the development of the artificial pancreas (Hovorka (2006)). Both physiological and technological delays exist in glucose sensing and insulin delivery. Such time delays can result in poorly controlled glucose since hypoglycemia can be induced and remains undetected for a significant time period. In an attempt to eliminate or minimize the effect of time delay, closed-loop control structures with time-delay compensation features can be used to improve the control outputs and reduce the hypoglycemic risk produced by physiological and technological delays.

Smith predictor structure is a control scheme that presents good properties in controlling systems with long time delay. The idea behind Smith predictor is to incorporate the system model within the closed-loop control structure (i.e. the system model becomes an explicit part of the controller). Thus, the design of Smith predictor scheme requires a model of the system dynamics and an estimate of the system time delay *t*0. In the Smith predictor scheme, there are two parallel paths for the control signal *u*(*t*) (see Figure 5); one passing through the real system (the patient), and one passing through the model of the system *Gs*. The function of the parallel path containing the model is to generate the difference *em*(*t*) between the actual system output *y*(*t*) and a model-based prediction of the control signal effect on the system output *ym*(*t*). The Smith predictor uses the model to predict the delay-free response of the system *y*− *<sup>m</sup>*(*t*). Then, it compares this prediction to the target glucose level *r*(*t*) to decide what control actions are needed. To avoid drifting and reject external disturbances, the Smith predictor also compares the actual system output with a prediction that takes the time delay into account. The error *em*(*t*) contributes to the overall error signal *e*(*t*) delivered to the feedback controller.

The Smith predictor structure has been recently used in artificial pancreas studies (Abu-Rmileh et al. (2010a;b)). With an initial estimation of the time delay, the Smith predictor shows the ability to minimize the effect of time delays and the associated risk of hypoglycemia, and to enhance the controller performance. As mentioned before, the MPC strategy, which has been extensively studied in artificial pancreas applications, is another competitive control algorithm with inherited ability to deal with system time delays (Hovorka (2006)).

control is believed to be more appropriate for the closed-loop artificial pancreas, and will enhance hypoglycemia prevention features of closed-loop systems due to its ability to provide particular insulin profile for each glycemic region. However, the identification of nonlinear models is still a challenging task in the artificial pancreas research. In order to be used in closed-loop control, such nonlinear model should be sufficiently accurate to capture the main system behavior and nonlinearity, while being relatively simple to be identified from the

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 219

Nonlinear control strategies like nonlinear MPC (NMPC) and sliding mode control (SMC), have shown superior performance over classical linear controllers in the blood glucose control problem. Most of the available MPC strategies are based on a linear model of the system. For systems that are highly nonlinear, the performance of a linear MPC can be poor. This has motivated the design of the NMPC, where a more accurate nonlinear model of the system is used for prediction and optimization. NMPC has been used in a number of artificial pancreas studies (Hovorka et al. (2010; 2004); Schlotthauer et al. (2005); Trajanoski & Wach (1998)). SMC is a nonlinear robust procedure to synthesize controllers for linear and nonlinear systems. The design of SMC algorithm includes two main steps. 1) Choosing a switching (sliding) surface, along which the system can slide to its desired final value. The sliding surface is designed so that it describes the desired system dynamics. The sliding surface divides the phase plane into regions where the switching function has different signs. 2) By using appropriate control law: make the system reach the switching surface *(reaching phase)*, and keep it on the surface *(sliding phase)*. The structure of the controller is intentionally altered as its state crosses the surface in accordance with a prescribed control law. SMC exhibits good

SMC algorithms have been employed successfully in different *in-silico* studies of artificial pancreas (Abu-Rmileh et al. (2010a;b); Kaveh & Shtessel (2008)). The combination between SMC and Smith predictor used in (Abu-Rmileh et al. (2010a;b)) is simple in its formulation and implementation, yet has some good features such as accuracy and robustness, insensitivity to internal and external disturbances, time-delay compensation and finite time convergence. These features make the proposed control algorithm suitable for the blood glucose problem which incorporates many sources of uncertainty and disturbances, and imposes some specific time requirements to avoid hypoglycemia and extended hyperglycemia. Other nonlinear control and modeling techniques have been used in the artificial pancreas research. Brief descriptions of frequently used approaches are given here, while comprehensive reviews are provided in Bequette (2005); Chee & Fernando (2007); El-Youssef et al. (2009); Takahashi et al.

As mentioned before, the glucoregulatory system is nonlinear and difficult to model mathematically. Therefore, empirically-based and model-free control techniques such as fuzzy and neural network systems would be key components in artificial pancreas control systems. Fuzzy systems are based on the idea that input-output relationships are not crisp, but can change gradually from one state to the next, and partial membership rather than crisp membership can be used to adjust the control action. Fuzzy logic control takes the input variables and maps them into fuzzy levels by sets of membership functions. Each input variable has determined value's degree of membership in a fuzzy set. The process of converting crisp input values to fuzzy values is called *fuzzification*. The fuzzy controller makes decisions for what action to take based on a set of rules. The set of rules are built generally based on expert knowledge. The input signal is processed applying the corresponding rules and generating a result for each, then combining the results of these rules. Finally, the fuzzy

available data such as CGM measurements, and insulin and meal information.

robustness against parameter variations, modeling errors and disturbances.

(2008)).

Fig. 5. Smith predictor control structure for time-delay compensation

#### **3.7 Insulin on board and insulin feedback**

As discussed previously, the use of subcutaneous route faces a challenging problem represented by the delayed insulin action. The effect of subcutaneous insulin may remain active over an extended time period (3-5 hours) after administration. Insulin on board (IOB) is a term used to describe how much insulin is still active from previous doses. Modern insulin pumps include the IOB option that helps in calculating the next required insulin dose. Therefore, IOB curves (time-action profiles) can be used in the development of artificial pancreas control algorithms to consider the effect of previous insulin, and provide a type of safety measure to avoid the problem of overdosing and the associated hypoglycemia. Ellingsen et al. (2009) developed a MPC scheme with IOB constraints. The IOB was used as dynamic safety constraints with a set of curves, to account for the time profile of delayed insulin action. Lee et al. (2009) used the IOB safety constraints in an integrated control scheme for the artificial pancreas that includes MPC strategy, meal detection algorithm, IOB constraints, and pump suspension option to avoid hypoglycemia.

Another technique used to reduce insulin infusion is the *insulin feedback*, initially introduced by Steil et al. (2004). The algorithm aims at reproducing as close as possible the insulin secretion from the natural pancreas. The idea behind this technique is to consider that a part of previous insulin is still active, and can cause further reduction in glucose level. Based on a pharmacokinetic model (Steil et al. (2006)), the algorithm estimates the plasma insulin level, and reduces the output of a proportional-integration-derivative (PID) controller by using the insulin feedback term, that is proportional to the estimated plasma insulin. Different versions of the algorithm have been used in clinical studies (Steil et al. (2011; 2006); Weinzimer et al. (2008)). In a recent study (Steil et al. (2011)), the insulin feedback has been used to improve the PID controller response in avoiding hypoglycemia after breakfast, and has achieved the desired performance.

#### **3.8 Nonlinear modeling and control**

Since the effect of insulin is nonlinear across the different glycemic ranges, the use of nonlinear models able to describe this nonlinear behavior would facilitate the design of more robust nonlinear control strategies, to handle the difference between glycemic ranges and their insulin requirements. Nonlinear models are more flexible in capturing complex behavior than the linear models, and consequently, the nonlinear control strategies are considered to be more suitable for this type of systems than linear control strategies. Therefore, nonlinear 12 Will-be-set-by-IN-TECH

Gs

ym - (t)

.

As discussed previously, the use of subcutaneous route faces a challenging problem represented by the delayed insulin action. The effect of subcutaneous insulin may remain active over an extended time period (3-5 hours) after administration. Insulin on board (IOB) is a term used to describe how much insulin is still active from previous doses. Modern insulin pumps include the IOB option that helps in calculating the next required insulin dose. Therefore, IOB curves (time-action profiles) can be used in the development of artificial pancreas control algorithms to consider the effect of previous insulin, and provide a type of safety measure to avoid the problem of overdosing and the associated hypoglycemia. Ellingsen et al. (2009) developed a MPC scheme with IOB constraints. The IOB was used as dynamic safety constraints with a set of curves, to account for the time profile of delayed insulin action. Lee et al. (2009) used the IOB safety constraints in an integrated control scheme for the artificial pancreas that includes MPC strategy, meal detection algorithm, IOB

Another technique used to reduce insulin infusion is the *insulin feedback*, initially introduced by Steil et al. (2004). The algorithm aims at reproducing as close as possible the insulin secretion from the natural pancreas. The idea behind this technique is to consider that a part of previous insulin is still active, and can cause further reduction in glucose level. Based on a pharmacokinetic model (Steil et al. (2006)), the algorithm estimates the plasma insulin level, and reduces the output of a proportional-integration-derivative (PID) controller by using the insulin feedback term, that is proportional to the estimated plasma insulin. Different versions of the algorithm have been used in clinical studies (Steil et al. (2011; 2006); Weinzimer et al. (2008)). In a recent study (Steil et al. (2011)), the insulin feedback has been used to improve the PID controller response in avoiding hypoglycemia after breakfast, and has achieved the

Since the effect of insulin is nonlinear across the different glycemic ranges, the use of nonlinear models able to describe this nonlinear behavior would facilitate the design of more robust nonlinear control strategies, to handle the difference between glycemic ranges and their insulin requirements. Nonlinear models are more flexible in capturing complex behavior than the linear models, and consequently, the nonlinear control strategies are considered to be more suitable for this type of systems than linear control strategies. Therefore, nonlinear

P atient and C GM

+

+

T ime Delay (t0) ym(t)

em(t)


y(t)

Feedback controller

Fig. 5. Smith predictor control structure for time-delay compensation

constraints, and pump suspension option to avoid hypoglycemia.

r(t) u(t)

+


e(t)

**3.7 Insulin on board and insulin feedback**

desired performance.

**3.8 Nonlinear modeling and control**

control is believed to be more appropriate for the closed-loop artificial pancreas, and will enhance hypoglycemia prevention features of closed-loop systems due to its ability to provide particular insulin profile for each glycemic region. However, the identification of nonlinear models is still a challenging task in the artificial pancreas research. In order to be used in closed-loop control, such nonlinear model should be sufficiently accurate to capture the main system behavior and nonlinearity, while being relatively simple to be identified from the available data such as CGM measurements, and insulin and meal information.

Nonlinear control strategies like nonlinear MPC (NMPC) and sliding mode control (SMC), have shown superior performance over classical linear controllers in the blood glucose control problem. Most of the available MPC strategies are based on a linear model of the system. For systems that are highly nonlinear, the performance of a linear MPC can be poor. This has motivated the design of the NMPC, where a more accurate nonlinear model of the system is used for prediction and optimization. NMPC has been used in a number of artificial pancreas studies (Hovorka et al. (2010; 2004); Schlotthauer et al. (2005); Trajanoski & Wach (1998)).

SMC is a nonlinear robust procedure to synthesize controllers for linear and nonlinear systems. The design of SMC algorithm includes two main steps. 1) Choosing a switching (sliding) surface, along which the system can slide to its desired final value. The sliding surface is designed so that it describes the desired system dynamics. The sliding surface divides the phase plane into regions where the switching function has different signs. 2) By using appropriate control law: make the system reach the switching surface *(reaching phase)*, and keep it on the surface *(sliding phase)*. The structure of the controller is intentionally altered as its state crosses the surface in accordance with a prescribed control law. SMC exhibits good robustness against parameter variations, modeling errors and disturbances.

SMC algorithms have been employed successfully in different *in-silico* studies of artificial pancreas (Abu-Rmileh et al. (2010a;b); Kaveh & Shtessel (2008)). The combination between SMC and Smith predictor used in (Abu-Rmileh et al. (2010a;b)) is simple in its formulation and implementation, yet has some good features such as accuracy and robustness, insensitivity to internal and external disturbances, time-delay compensation and finite time convergence. These features make the proposed control algorithm suitable for the blood glucose problem which incorporates many sources of uncertainty and disturbances, and imposes some specific time requirements to avoid hypoglycemia and extended hyperglycemia. Other nonlinear control and modeling techniques have been used in the artificial pancreas research. Brief descriptions of frequently used approaches are given here, while comprehensive reviews are provided in Bequette (2005); Chee & Fernando (2007); El-Youssef et al. (2009); Takahashi et al. (2008)).

As mentioned before, the glucoregulatory system is nonlinear and difficult to model mathematically. Therefore, empirically-based and model-free control techniques such as fuzzy and neural network systems would be key components in artificial pancreas control systems. Fuzzy systems are based on the idea that input-output relationships are not crisp, but can change gradually from one state to the next, and partial membership rather than crisp membership can be used to adjust the control action. Fuzzy logic control takes the input variables and maps them into fuzzy levels by sets of membership functions. Each input variable has determined value's degree of membership in a fuzzy set. The process of converting crisp input values to fuzzy values is called *fuzzification*. The fuzzy controller makes decisions for what action to take based on a set of rules. The set of rules are built generally based on expert knowledge. The input signal is processed applying the corresponding rules and generating a result for each, then combining the results of these rules. Finally, the fuzzy

has been tested and proved to improve the performance of the closed-loop control and reduce

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 221

While partial results obtained in different artificial pancreas studies are promising, several aspects regarding the fully developed artificial pancreas are still open, and further improvements are needed. Obtaining models from patient's input-output data using advanced modeling techniques is recommended for blood glucose control. Nonlinear identification of insulin-glucose models for control is desirable. Development of advanced control techniques is needed due to the nonlinear behavior, unmodeled disturbances, delay and inaccuracy in measurements, together with modeling errors and patient variability. Another required improvement is the modeling of different meal contents, since most of the available models are restricted to carbohydrates effect. Using multiple variable control (i.e. considering insulin, glucagon, exercise, stress, etc.), and incorporating the effect of insulin sensitivity change during the day in the control algorithm design, would increase the reliability of models in representing the real conditions of the diabetic patient, and

consequently, improve the overall performance of the designed artificial pancreas.

Although the nonlinearity in the insulin-glucose system is quite obvious, the available hypoglycemia detection and prediction algorithms do not consider the nonlinear nature of the system through the different glycemic ranges (Chan et al. (2010)). Taking into account the nonlinearity of the system would be a possible way to enhance the performance of the algorithms and increase their effectiveness in preventing hypoglycemia (Chan et al. (2010)). The inclusion of IOB effect in predicting future hypoglycemic episodes could be another technique to improve the feasibility of these algorithms (Buckingham et al. (2009)). Finally, improving the accuracy and reliability of CGM systems is an essential task, since both control algorithms and hypoglycemia alarms depend widely on CGM measurements. Poorly functioning sensor increases the risk of system-induced and undetected hypoglycemia, while

The first author acknowledges the support of the University of Girona through the (BR-UdG)

Abu-Rmileh, A. & Garcia-Gabin, W. (2010a). Feedforward-feedback multiple predictive

Abu-Rmileh, A. & Garcia-Gabin, W. (2010b). A Gain Scheduling Model Predictive Controller

Abu-Rmileh, A., Garcia-Gabin, W. & Zambrano, D. (2010a). Internal model sliding mode

Abu-Rmileh, A., Garcia-Gabin, W. & Zambrano, D. (2010b). A robust sliding mode

*Biological Engineering and Computing* 48(12): 1191 – 1201.

controllers for glucose regulation in type 1 diabetes, *Computer Methods and Programs*

for Blood Glucose Control in Type 1 Diabetes, *IEEE Transaction on Biomedical*

control approach for glucose regulation in type 1 diabetes, *Biomedical Signal Processing*

controller with internal model for closed-loop artificial pancreas, *Medical and*

the hypoglycemia episodes. Other techniques are still under study.

accurate sensor improves the control quality and reduces the risk.

*in Biomedicine* 99(2): 113–123.

*Engineering* 57(10): 2478–2484.

*and Control* 5(2): 94 – 102.

**5. Acknowledgement**

research grant.

**6. References**

controller output is obtained via *defuzzification* combining result back into a specific crisp control output value. Different fuzzy control schemes have been implemented in artificial pancreas studies (see for example Atlas et al. (2010); Campos-Delgado et al. (2006); Ibbini (2006); Ibbini & Massadeh (2005)). In Atlas et al. (2010), a personalized fuzzy logic controller has been validated clinically, and proved to minimize hyperglycemic peaks while preventing hypoglycemia.

Neural networks are modeling techniques that result in a nonlinear model based on experimental data. It is a black-box model organized in sequential layers containing neurons. The network output is obtained as a weighted sum of inputs through the hidden layers. The weights are found during a training process by minimizing the error between desired and network output. Neural networks show excellent adaptation and learning ability. Neural networks deal with the blood glucose problem without explicit description of the exact model of the insulin-glucose system. Such approach is very useful in irregular situations (e.g. patients have a disease or abnormal conditions) that limit the usability of normal models (Takahashi et al. (2008)). Neural networks have been used to obtain insulin-glucose models for the design of nonlinear closed-loop controllers (El-Jabali (2005); Schlotthauer et al. (2005); Takahashi et al. (2008); Trajanoski & Wach (1998)). A combination between fuzzy logic and neural network (neuro-fuzzy) control strategy was applied by Dazzi et al. (2001) in clinics, and proved to provide superior glycemic control compared to conventional algorithms, with hypoglycemic events reduced to half.

Adaptive control is another approach used for glucose regulation. The complexity of glucose control mechanism highlights the need for an adaptive control algorithm to compensate for variations in patient dynamics (e.g. time-varying insulin sensitivity, stress and physical exercise) or disturbances by adapting the controller and model parameters to the changing patient conditions (Eren-Oruklu et al. (2009a); Hovorka (2005)). Adaptive control includes several configurations that allow not only outputs of the controller to be changed over time, but also the method by which those outputs are generated; the controller continuously monitors its own adaptation through a defined metric, and is capable of altering its own control scheme to better meet the adaptation criterion. For blood glucose control, different adaptation schemes have been employed (Chee & Fernando (2007)), in systems that use the sensor measurements to track the changes in glucose dynamics and update the controller structure to assign the required insulin regime. In model-based adaptive control, patient model is used to predict future glucose levels based on current and past insulin infusions. The model parameters are continuously updated and used in the control algorithm to calculate the required insulin. Adaptive control strategies have the ability to individualize the control scheme and/or patient model to represent the inter- and intra-patient variability. Adaptive schemes have achieved safe control while avoiding hypoglycemia in spite of all the challenges facing the closed-loop artificial pancreas (Eren-Oruklu et al. (2009a); Shaller et al. (2006)).

#### **4. Conclusions**

Closed-loop insulin delivery by the artificial pancreas gives hope to achieve tight glycemic control in T1DM by reducing the risk of hypoglycemia while solving the problem of hyperglycemia. The prevention of life-threatening hypoglycemia is considered as a possible goal for the first generation of the artificial pancreas before reaching the fully developed device that mimics the function of natural pancreas in night, fasting and prandial conditions. The closed-loop system can be subjected to different modifications to implement control techniques that reduce the risk of hypoglycemia. The feasibility of some of these techniques has been tested and proved to improve the performance of the closed-loop control and reduce the hypoglycemia episodes. Other techniques are still under study.

While partial results obtained in different artificial pancreas studies are promising, several aspects regarding the fully developed artificial pancreas are still open, and further improvements are needed. Obtaining models from patient's input-output data using advanced modeling techniques is recommended for blood glucose control. Nonlinear identification of insulin-glucose models for control is desirable. Development of advanced control techniques is needed due to the nonlinear behavior, unmodeled disturbances, delay and inaccuracy in measurements, together with modeling errors and patient variability.

Another required improvement is the modeling of different meal contents, since most of the available models are restricted to carbohydrates effect. Using multiple variable control (i.e. considering insulin, glucagon, exercise, stress, etc.), and incorporating the effect of insulin sensitivity change during the day in the control algorithm design, would increase the reliability of models in representing the real conditions of the diabetic patient, and consequently, improve the overall performance of the designed artificial pancreas.

Although the nonlinearity in the insulin-glucose system is quite obvious, the available hypoglycemia detection and prediction algorithms do not consider the nonlinear nature of the system through the different glycemic ranges (Chan et al. (2010)). Taking into account the nonlinearity of the system would be a possible way to enhance the performance of the algorithms and increase their effectiveness in preventing hypoglycemia (Chan et al. (2010)). The inclusion of IOB effect in predicting future hypoglycemic episodes could be another technique to improve the feasibility of these algorithms (Buckingham et al. (2009)). Finally, improving the accuracy and reliability of CGM systems is an essential task, since both control algorithms and hypoglycemia alarms depend widely on CGM measurements. Poorly functioning sensor increases the risk of system-induced and undetected hypoglycemia, while accurate sensor improves the control quality and reduces the risk.

#### **5. Acknowledgement**

The first author acknowledges the support of the University of Girona through the (BR-UdG) research grant.

#### **6. References**

14 Will-be-set-by-IN-TECH

controller output is obtained via *defuzzification* combining result back into a specific crisp control output value. Different fuzzy control schemes have been implemented in artificial pancreas studies (see for example Atlas et al. (2010); Campos-Delgado et al. (2006); Ibbini (2006); Ibbini & Massadeh (2005)). In Atlas et al. (2010), a personalized fuzzy logic controller has been validated clinically, and proved to minimize hyperglycemic peaks while preventing

Neural networks are modeling techniques that result in a nonlinear model based on experimental data. It is a black-box model organized in sequential layers containing neurons. The network output is obtained as a weighted sum of inputs through the hidden layers. The weights are found during a training process by minimizing the error between desired and network output. Neural networks show excellent adaptation and learning ability. Neural networks deal with the blood glucose problem without explicit description of the exact model of the insulin-glucose system. Such approach is very useful in irregular situations (e.g. patients have a disease or abnormal conditions) that limit the usability of normal models (Takahashi et al. (2008)). Neural networks have been used to obtain insulin-glucose models for the design of nonlinear closed-loop controllers (El-Jabali (2005); Schlotthauer et al. (2005); Takahashi et al. (2008); Trajanoski & Wach (1998)). A combination between fuzzy logic and neural network (neuro-fuzzy) control strategy was applied by Dazzi et al. (2001) in clinics, and proved to provide superior glycemic control compared to conventional algorithms, with

Adaptive control is another approach used for glucose regulation. The complexity of glucose control mechanism highlights the need for an adaptive control algorithm to compensate for variations in patient dynamics (e.g. time-varying insulin sensitivity, stress and physical exercise) or disturbances by adapting the controller and model parameters to the changing patient conditions (Eren-Oruklu et al. (2009a); Hovorka (2005)). Adaptive control includes several configurations that allow not only outputs of the controller to be changed over time, but also the method by which those outputs are generated; the controller continuously monitors its own adaptation through a defined metric, and is capable of altering its own control scheme to better meet the adaptation criterion. For blood glucose control, different adaptation schemes have been employed (Chee & Fernando (2007)), in systems that use the sensor measurements to track the changes in glucose dynamics and update the controller structure to assign the required insulin regime. In model-based adaptive control, patient model is used to predict future glucose levels based on current and past insulin infusions. The model parameters are continuously updated and used in the control algorithm to calculate the required insulin. Adaptive control strategies have the ability to individualize the control scheme and/or patient model to represent the inter- and intra-patient variability. Adaptive schemes have achieved safe control while avoiding hypoglycemia in spite of all the challenges facing the closed-loop artificial pancreas (Eren-Oruklu et al. (2009a); Shaller et al. (2006)).

Closed-loop insulin delivery by the artificial pancreas gives hope to achieve tight glycemic control in T1DM by reducing the risk of hypoglycemia while solving the problem of hyperglycemia. The prevention of life-threatening hypoglycemia is considered as a possible goal for the first generation of the artificial pancreas before reaching the fully developed device that mimics the function of natural pancreas in night, fasting and prandial conditions. The closed-loop system can be subjected to different modifications to implement control techniques that reduce the risk of hypoglycemia. The feasibility of some of these techniques

hypoglycemia.

**4. Conclusions**

hypoglycemic events reduced to half.


Dalla Man, C., Camilleri, M. & Cobelli, C. (2006). A system model of oral glucose

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 223

Dalla Man, C., Rizza, R. & Cobelli, C. (2007). Meal simulation model of the glucose-insulin system, *IEEE Transactions on Biomedical Engineering* 54(10): 1740–1749. Dassau, E., Bequette, B., Buckingham, B. & Doyle III, F. (2008). Detection of a meal using

Dazzi, D., Taddei, F., Gavarini, A., Uggeri, E., Negro, R. & Pezzarossa, A. (2001). The control of

DCCT Research Group (1993). The effect of intensive treatment of diabetes on the

Doyle III, F., Jovanovic, L. & Seborg, D. (2007). Glucose control strategies for treating type 1

Dua, P., Doyle III, F. & Pistikopoulos, E. (2009). Multi-objective blood glucose control for type 1 diabetes, *Medical and Biological Engineering and Computing* 47(3): 343–52. El-Jabali, A. (2005). Neural network modeling and control of type 1 diabetes mellitus,

El-Khatib, F. H., Jiang, J. & Damiano, E. R. (2009). A feasibility study of bihormonal

El-Khatib, F. H., Russell, S. J., Nathan, D. M., Sutherlin, R. G. & Damiano, E. R. (2010). A

El-Youssef, J., Castle, J. & Ward, W. (2009). A review of closed-loop algorithms for glycemic

Eren-Oruklu, M., Cinar, A., Quinn, L. & Smith, D. (2009a). Adaptive control strategy for

Eren-Oruklu, M., Cinar, A., Quinn, L. & Smith, D. (2009b). Estimation of future glucose

Finan, D., Palerm, C., Doyle, F., Seborg, D., Zisser, H., Bevier, W. & Jovanovic, L. (2009). Effect

Garcia-Gabin, W., Zambrano, D., et al. (2009). A sliding mode predictive control approach to

*Control in Biomedical Systems*, Aalborg, Denmark, pp. 85–90.

control in the treatment of type 1 diabetes, *Algorithms* 2(1): 518–532. Ellingsen, C., Dassau, E., Zisser, H., Grosman, B., Percival, M., Jovanovic, L. & Doyle 3rd.,

closed-loop blood glucose control using dual subcutaneous infusion of insulin and glucagon in ambulatory diabetic swine., *Journal of Diabetes Science and Technology*

bihormonal closed-loop artificial pancreas for type 1 diabetes, *Science Translational*

F. (2009). Safety constraints in an artificial pancreatic *β* cell: an implementation of model predictive control with insulin on board., *Journal of Diabetes Science and*

regulation of blood glucose levels in patients with type 1 diabetes, *Journal of process*

concentrations with subject-specific recursive linear models, *Diabetes Technology and*

of input excitation on the quality of empirical dynamic models for type 1 diabetes,

closed-loop glucose control for type 1 diabetes, *7th IFAC Symposium on Modelling and*

diabetes mellitus, *New England Journal of Medicine* 329(14): 977–986.

diabetes mellitus, *Journal of Process Control* 17(7): 572–576.

*Bioprocess and Biosystems Engineering* 27(2): 75–79.

*Engineering* 53(12): 2472–2478.

*and its Complications* 15(2): 80–87.

31(2): 295–300.

3(4): 789–803.

*Medicine* 2(27).

*Technology* 3(3): 536–544.

*control* 19(8): 1333–1346.

*Therapeutics* 11(4): 243–253.

*AICHE Journal* 55(5): 1135–1146.

absorption: validation on gold standard data, *IEEE Transactions on Biomedical*

continuous glucose monitoring: Implications for an artificial *β*-cell, *Diabetes care*

blood glucose in the critical diabetic patient: a neuro-fuzzy method, *Journal of Diabetes*

development and progression of long term complications in insulin-dependent


16 Will-be-set-by-IN-TECH

Atlas, E., Nimri, R., Miller, S., Gurmberg, E. & Phillip, M. (2010). MD-logic artificial

Bequette, B. (2005). A critical assessment of algorithms and challenges in the development of a closed-loop artificial pancreas, *Diabetes Technology and Therapeutics* 7(1): 28–46. Bergman, R., Ider, Y., Bowden, C. & Cobelli, C. (1979). Quantitative estimation of insulin

Briscoe, V. & Davis, S. (2006). Hypoglycemia in type 1 and type 2 diabetes: Physiology,

Buckingham, B., Cobry, E., Clinton, P., Gage, V., Caswell, K., Kunselman, E., Cameron, F. &

insulin pump suspension, *Diabetes Technology and Therapeutics* 11(2): 93–97. Cameron, F., Niemeyer, G. & Buckingham, B. A. (2009). Probabilistic evolving meal detection

Cameron, F., Niemeyer, G., Gundy-Burlet, K. & Buckingham, B. (2008). Statistical hypoglycemia prediction, *Journal of Diabetes Science and Technology* 2(4): 612–621. Campos-Delgado, D., Hernandez-Ordoñez, M., Femat, R. & Gordillo-Moscoso, A. (2006).

Chan, A., Heinemann, L., Anderson, S., Breton, M. & Kovatchev, B. (2010). Nonlinear

Clarke, W. L., Anderson, S., Breton, M., Patek, S., Kashmer, L. & Kovatchev, B. (2009).

Cobelli, C., Dalla Man, C., Sparacino, G., Magni, L., De Nicolao, G. & Kovatchev, B. (2009).

Cryer, P. (2008). Hypoglycemia: Still the limiting factor in the glycemic management of

diabetes mellitus., *Journal of diabetes science and technology* **4**(4): 873–881. Chee, F. & Fernando, T. (2007). *Closed-Loop Control of Blood Glucose*, Springer-Verlag, London. Choleau, C., Dokladal, P., Klein, J. ., Kenneth Ward, W., Wilson, G. S. & Reach, G.

pathophysiology, and management, *Clinical Diabetes* 24(3): 115–121. Bruttomesso, D., Farret, A., Costa, S., Marescotti, M. C., Vettore, M., Avogaro, A., Tiengo,

sensitivity, *American Journal of Physiology* 236: E667.

33(5): 1072–1076.

*and Technology* 3(5): 1014–1021.

*technology* 3(5): 1022–1030.

–96.

diabetes, *Diabetes Care* 33(6): 1282 – 1287.

monitoring system, *Diabetes* 51(11): 3263–3273.

*of Diabetes Science and Technology* 3(5): 1031–1038.

diabetes, *Endocrine Practice* 14(6): 750–756.

pancreas system: A pilot study in adults with type 1 diabetes mellitus, *Diabetes Care*

A., Dalla Man, C., Place, J., Facchinetti, A., Guerra, S., Magni, L., De Nicolao, G., Cobelli, C., Renard, E. & Maran, A. (2009). Closed-loop artificial pancreas using subcutaneous glucose sensing and insulin delivery and a model predictive control algorithm: preliminary studies in padova and montpellier., *Journal of Diabetes Science*

Chase, H. P. (2009). Preventing hypoglycemia using predictive alarm algorithms and

and estimation of meal total glucose appearance., *Journal of diabetes science and*

Fuzzy-based controller for glucose regulation in type 1 diabetic patients by subcutaneous route, *IEEE Transations on Biomedical Engineering* 53(11): 2201–2210. Castle, J., Engle, J., El-Youssef, J., Massoud, R., Yuen, K., Kagan, R. & Ward, W. (2010). Novel

use of glucagon in a closed-loop system for prevention of hypoglycemia in type 1

metabolic effect of insulin across the blood glucose range in patients with type 1

(2002). Prevention of hypoglycemia using risk assessment with a continuous glucose

Closed-loop artificial pancreas using subcutaneous glucose sensing and insulin delivery and a model predictive control algorithm: the virginia experience., *Journal*

Diabetes: Models, signals, and control, *IEEE Reviews in Biomedical Engineering* 2: 54


Maciejowski, J. (2002). *Predictive Control with Constraints*, Prentice Hall.

*Science and Technology* 1(6): 804–812.

*of Process Control* 18(2): 149–162.

pp. 3822–3825.

23(1): 90–93.

*Physics* 28(3): 240 – 250.

*Technology* 4(1): 34–40.

*Metabolism*, DOI:10.1210/jc.2010-2578.

*Therapeutics* 5(6): 953–964.

Engineering, MIT.

–937.

Magni, L., Raimondo, D., Bossi, L., Dalla Man, C., De Nicolao, G., Kovatchev, B. & Cobelli, C.

Hypoglycemia Prevention in Closed-Loop Artificial Pancreas for Patients with Type 1 Diabetes 225

Marchetti, G., Barolo, M., Jovanovic, L., Zisser, H. & Seborg, D. (2008). A

Nguyen, H., Ghevondian, N. & Jones, T. (2009). Real-time detection of nocturnal

Palerm, C. C., Willis, J. P., Desemone, J. & Bequette, B. W. (2005). Hypoglycemia prediction and detection using optimal estimation, *Diabetes Technology and Therapeutics* 7(1): 3–14. Parker, R., Doyle lll, F. & Peppas, N. (1999). A model-based algorithm for blood glucose control

Schlotthauer, G., Gamero, L., Torres, M. & Nicolini, G. (2005). Modeling, identification and

Shaller, H. C., Schaupp, L., Bodenlenz, M., Wilinska, E., Chassin, L. J.and Wach, P.,

Skladnev, V., Tarnavskii, S., McGregor, T., Ghevondian, N., Gourlay, S. & Jones, T. (2010).

Sorensen, J. (1985). *A physiologic Model of Glucose Metabolism in Man and its Use to Design and*

Sparacino, G., Zanderigo, F., Corazza, S., Maran, A., Facchinetti, A. & Cobelli, C. (2007).

Steil, G., Palerm, C., Kurtz, N., Voskanyan, G., Roy, A., Paz, S. & Kandeel, F. (2011). The effect

Steil, G., Pantaleon, A. & Rebrin, K. (2004). Closed-loop insulin delivery - the path to physiological glucose control, *Advanced Drug Delivery Reviews* 56(2): 125–144. Steil, G., Rebrin, K., Darwin, C., Hariri, F. & Saad, M. (2006). Feasibility of automating insulin delivery for the treatment of type 1 diabetes, *Diabetes* 55(12): 3344–3350. Steil, G., Rebrin, K., Janowski, R., Darwin, C. & Saad, M. (2003). Modeling *β*-cell insulin

Takahashi, D., Xiao, Y. & Hu, F. (2008). A survey of insulin dependent diabetes part II: Control

methods, *International Journal of Telemedicine and Applications* .

(2007). Model predictive control of type 1 diabetes: an in silico trial, *Journal of Diabetes*

feedforward-feedback glucose control strategy for type 1 diabetes mellitus, *Journal*

hypoglycemic episodes using a novel non-invasive hypoglycemia monitor, *Proceedings of the 31st Annual International Conference of the IEEE Engineering in Medicine and Biology Society: Engineering the Future of Biomedicine, EMBC 2009*,

in type 1 diabetic patients, *IEEE Transactions on Biomedical Engineering* 46(2): 148–157.

nonlinear model predictive control of type i diabetic patient, *Medical Engineering and*

Vering, T., Hovorka, R. & Pieber, T. R. (2006). On-line adaptive algorithm with glucose prediction capacity for subcutaneous closed loop control of glucose: evaluation under fasting conditions in patients with type 1 diabetes, *Diabetic Medicine*

Hypoglycemia alarm enhancement using data fusion, *Journal of Diabetes Science and*

*Assess Improved Insulin Therapies for Diabetes*, PhD thesis, Department of Chemical

Glucose concentration can be predicted ahead in time from continuous glucose monitoring sensor time-series, *IEEE Transactions on Biomedical Engineering* 54(5): 931

of insulin feedback on closed loop glucose control, *Journal of Clinical Endocrinology &*

secretion - implications for closed-loop glucose homeostasis, *Diabetes Technology and*


Maciejowski, J. (2002). *Predictive Control with Constraints*, Prentice Hall.

18 Will-be-set-by-IN-TECH

Grosman, B., Dassau, E., Zisser, H., Jovanovic, L. & Doyle III, F. (2010). Zone model predictive

Hovorka, R. (2005). Management of diabetes using adaptive control, *International Journal of*

Hovorka, R. (2006). Continuous glucose monitoring and closed-loop systems, *Diabetic*

Hovorka, R., Allen, J. M., Elleri, D., Chassin, L. J., Harris, J., Xing, D., Kollman, C., Hovorka, T.,

Hovorka, R., Canonico, V., Chassin, L., Haueter, U., Massi-Benedetti, M., Orsini Federici, M.,

Hovorka, R., Shojaee-Moradie, F., Carroll, P., Chassin, L., Gowrie, I., Jackson, N., Tudor,

Hovorka, R., Wilinska, M., Chassin, L. & Dunger, D. (2006). Roadmap to the artificial pancreas,

Hughes, C., Patek, S., Breton, M. & BP, K. (2010). Hypoglycemia prevention via

Ibbini, M. (2006). A PI-fuzzy logic controller for the regulation of blood glucose level in diabetic patients, *Journal of Medical Engineering and Technology* 30(2): 83 – 92. Ibbini, M. & Massadeh, M. (2005). A fuzzy logic based closed-loop control system for the blood

Kaveh, P. & Shtessel, Y. (2008). Blood glucose regulation using higher order sliding mode control, *International Journal of Robust and Nonlinear Control* 18: 557 – 569. Kirchsteiger, H. & Del Re, L. (2009). Reduced hypoglycemia risk in insulin bolus therapy using

Klonoff, D. (2005). Continuous glucose monitoring, roadmap for 21st century diabetes

Kovatchev, B., Breton, M., Dalla Man, C. & Cobelli, C. (2009). In silico preclinical trials: a

Kumareswaran, K., Evans, M. & Hovorka, R. (2009). Artificial pancreas: an emerging approach to treat type 1 diabetes, *Expert Reviews of Medical Devices* 6(4): 401–410. Lee, H. & Bequette, B. (2009). A closed-loop artificial pancreas based on model predictive

Lee, H., Buckingham, B., Wilson, D. & Bequette, B. (2009). A closed-loop artificial pancreas

proof of concept in closed-loop control of type 1 diabetes, *Journal of Diabetes Science*

control: Human friendly identification and automatic meal disturbance rejection,

using model predictive control and a sliding meal size estimator., *Journal of Diabetes*

asymmetric cost functions, *Asian Control Conference*, pp. 751–756.

*Science and Technology* 4(4): 961–975.

*Physiological Measurements* 25(4): 905–920.

therapy, *Diabetes Care* 28(5): 1231–1239.

*Science and Technology* 3(5): 1082–1090.

*Biomedical Signal Processing and Control* 4(4): 347–354.

*and Technology* 3(1): 44–55.

*Endocrinology and Metabolism* 282(5): 992–1007.

*Diabetes Research and Clinical Practice* 74(2): S178–S182.

*Medicine* 23: 1–12.

4(5): 1146–1155.

29(2): 64 – 69.

*Adaptive Control and Signal Processing* 19(5): 309–325.

control: A strategy to minimize hyper- and hypoglycemic events, *Journal of Diabetes*

Larsen, A. M. F., Nodale, M., De Palma, A., Wilinska, M. E., Acerini, C. L. & Dunger, D. B. (2010). Manual closed-loop insulin delivery in children and adolescents with type 1 diabetes: a phase 2 randomised crossover trial, *The Lancet* 375(9716): 743–751.

Pieber, T., Schaller, H., Schaupp, L., Vering, T. & Wilinska, M. (2004). Nonlinear model predictive control of glucose concentration in subjects with type 1 diabetes,

R., Umpleby, A. & Jones, R. (2002). Partitioning glucose distribution/transport, disposal, and endogenous production during ivgtt, *American Journal of Physiology -*

pump attenuation and red-yellow-green "traffic" lights using continuous glucose monitoring and insulin pump data, *Journal of diabetes science and technology*

glucose level regulation in diabetes, *Journal of Medical Engineering and Technology*


**12** 

*Poland* 

Leszek Szablewski

*Medical University of Warsaw* 

**Glucose Homeostasis – Mechanism and Defects** 

Glucose is an essential metabolic substrate of all mammalian cells. D-glucose is the major carbohydrate presented to the cell for energy production and many other anabolic requirements. Glucose and other monosaccharides are transported across the intestinal wall to the hepatic portal vein and then to liver cells and other tissues. There they are converted to fatty acids, amino acids, and glycogen, or are oxidized by the various catabolic pathways

Most tissues and organs, such as the brain, need glucose constantly, as an important source of energy. The low blood concentrations of glucose can causes seizures, loss of consciousness, and death. On the other hand, long lasting elevation of blood glucose concentrations, can result in blindness, renal failure, vascular disease, and neuropathy. Therefore, blood glucose concentrations need to be maintained within narrow limits. The process of maintaining blood glucose at a steady-state level is called glucose homeostasis. This is accomplished by the finely hormone regulation of peripheral glucose uptake, heaptic glucose production and glucose uptake during carbohydrate ingestion. This maintenance is achieved through a balance of several factors, including the rate of consumption and intestinal absorption of dietary carbohydrate, the rate of utilization of glucose by peripheral tissues and the loss of glucose through the kidney tubule, and the rate of removal or release of glucose by the liver and kidney. To avoid postprandial hyperglycemia (uncontrolled increases in blood glucose levels following meals) and fasting hypoglycemia (decreased in blood glucose levels during periods of fasting), the body can adjust levels by a variety of cellular mechanisms. Important mechanisms are conveyed by hormones, cytokines, and fuel

Diabetes mellitus is one of the clinical manifestations of long-term metabolic abnormalities involving multiple organs and hormonal pathways that impair the body's ability to maintain glucose homeostasis. As a result of impaired glucose homeostasis is a hyperglycemia. Prolonged elevation of blood glucose concentrations causes a number of complications like blindness, renal failure, cardiac and peripheral vascular disease, neuropathy, foot ulcers, and limb amputation. Vascular complications represent the leading

Hypoglycemia is abnormally low levels of sugar (glucose) in the blood. Low levels of sugar in the blood interfere with the function of much organ system. A person with hypoglycemia may feel weak, drowsy, confused, hungry, and dizzy. The other signs of low blood sugar are: paleness, headache, irritability, trembling, sweating, rapid heart beat, and a cold. The

substrates and are sensed through of cellular mechanisms.

cause of mortality and morbidity in diabetic patients.

**1. Introduction** 

of cells.

