Proceedings from The Seventh Conference The CareFusion Center for Safety and Clinical Excellence June 7-8, 2007, San Diego, CA Philip J. Schneider, MS, FASHP, Editor

Intensive Insulin for Tight Glycemic Control Research Therapy Monitoring Nursing

Conference Report Published by The CareFusion Center for Safety and Clinical Excellence www.cardinalhealth.com/clinicalcenter International Conference on Intensive Insulin Therapy for Tight Glycemic Control

The seventh invitational conference at the CareFusion Center for Safety and Clinical

Excellence in San Diego, held June 7-8, 2007, brought together a distinguished faculty from clinical practice, academia, and organizations. Judith Jacobi, PharmD, FCCM, FCCP,

BCPS, Critical Care Pharmacist, Methodist Hospital/Clarian Health, Indianapolis, IN and

Timothy S. Bailey, MD, FACE, CPI, Advanced Metabolic Care and Research, Escondido,

CA chaired the conference. Internationally recognized experts on research, current issues and opportunities in the use of intensive insulin therapy for tight glycemic control (TGC IIT) presented.

This conference report summarizes the information presented on TGC IIT with regard to research findings, safety concerns, emerging practices, monitoring, and nursing issues as researchers and clinicians seek to optimize insulin therapy to help maintain normoglycemia in critically ill patients. The proceedings were edited by Philip J.

Schneider, MS, FASHP, Clinical Professor and Director, Latiolais Leadership Program,

College of , The Ohio State University, Columbus, OH. 7th Invited Conference: Intensive Insulin Therapy for Tight Glycemic Control

Content Introduction p3 Tight Glycemic Control: Judith Jacobi, PharmD, FCCM, FCCP, BCPS An Overview Methodist Hospital/Clarian Health, Indianapolis, IN Timothy S. Bailey, MD, FACE, CPI Advanced Metabolic Care and Research Escondido, CA Research p4 A Brief History of Tight Glycemic Control: Tony Furnary, MD What We Know in 2007; and How We Got Here Starr-Wood Cardiac Group, Portland, OR p10 Intensive Insulin Therapy and the Simon Finfer, MBBS NICE-SUGAR Study Royal North Shore Hospital of Sydney, Australia p13 European Multi-center Trials with Tight Glucose Philippe Devos, MD Control by Intensive Insulin Therapy The George Institute of Liege, Belgium p17 Implementation of Tight Glycemic Control James Krinsley, MD at Stamford Hospital Stamford Hospital, Stamford, CT p21 Meta-analysis of Randomized Trials Anastassios G. Pittas, MD, MS of Tight Glycemic Control Tufts/New England Medical Center, Boston, MA p23 Perioperative Glucose Management Richard Prielipp, MD and IIT in The Operating Room University of Minnesota, Minneapolis, MN Douglas Coursin, MD Univertisty of Wisconsin, Madison,WI p26 Economic Advantages of Judi Jacobi, PharmD Tight Glycemic Control Methodist Hospital/Clarian Health, Indianapolis, IN Therapy p29 Intensive Insulin Therapy in the Robert Osburne, MD Intensive Care Unit Atlanta Medical Center, Atlanta, GA p33 Use of Computerized Algorithm in Patients Bruce Bode, MD Undergoing Cardiovascular : Atlanta Diabetes Associates, Atlanta, GA A Protocol for Tight Glycemic Control p36 Computerized Management of Pat Burgess, MD, PhD Tight Glycemic Control Carolinas Medical Center, Charlotte, NC p39 Analysis of Variation in Guy Soo Hoo, MD Insulin Protocols VA Hospital, Los Angeles, CA p43 Improving ICU Quality and Safety: Sean Berenholtz, MD Implications for Tight Glycemic Control Johns Hopkins, Baltimore, MD p47 Specialized Nutrition Support Kalman Holdy, MD and Glycemic Control Sharp Healthcare, San Diego, CA

Executive Summary Conference Report 1 7th Invited Conference: Intensive Insulin Therapy for Tight Glycemic Control

p50 Examining Medication Errors John Santell, MS Associated with Intravenous Insulin US Pharmacopeia, Washington, DC p53 The Portland-Vancouver Regional Chris Hogness, MD, MPH Inpatient Glycemic Control Collaborative Southwest Washington Medical Center Vancouver, WA p56 Building Transitions from the ICU to the Ward Greg Maynard, MD for the Hyperglycemic Patient: UCSD, San Diego, CA One Piece of the Puzzle Monitoring p59 Glucose Control and (Continuous) Christophe De Block, MD Glucose Monitoring in Critical Illness Antwerp University Hospital, Belgium p62 Glucose Sensor Technology: Timothy Bailey, MD Current State and Future Trends Advanced Metabolic Care and Research Escondido, CA p64 Assessing the Accuracy and Confounding Nam Tran, PhD (Candidate) Factors in Critical Care Glucose Monitoring UC Davis, Davis, CA p68 Glucose Sensor Augmented Insulin Delivery Jeffrey Joseph, DO in the Hospital: Open and Closed-Loop Methods T. Jefferson University, Philadelphia, PA Nursing p73 The Impact of Intensive Insulin Therapy Daleen Aragon, PhD, CCRN, FCCM on Nursing Orlando Regional, Orlando, FL p76 Nursing Education and Intensive Insulin Therapy Carol Manchester, MSN, APRN, BC-ADM, CDE University of Minnesota, Minneapolis, MN p79 Applying Glucometrics to Tight Jacqueline Thompson, MAS,RN,CDE Glycemic Control Sharp Healthcare, San Diego, CA Roundtable p81 Impact on Hospital Costs p81 Reasonable Target p82 Factors that Complicate Glycemic Control p82 Administrative Aspects of IIT p83 Risk of p84 Blood Glucose Measurement Appendix p85 Table. Major Published Randomized Controlled Anastassios G. Pittas, MD, MS Trials with Insulin Therapy in Critically Ill Patients Tufts/New England Medical Center, Boston, MA

2 Executive Summary Conference Report 7th Invited Conference: Tight Glycemic Control : An Overview

INTRODUCTION

Tight Glycemic Control: An Overview Judith Jacobi, PharmD, FCCM, FCCP, BCPS, Critical Care Pharmacist, Methodist Hospital/Clarian Health, Indianapolis, IN; Timothy S. Bailey, MD, FACE, CPI, Advanced Metabolic Care and Research, Escondido, CA

More than five years ago the publication There is significant workload associated will no longer ignore blood glucose values of a landmark trial of intensive insulin therapy with frequent glucose monitoring. Point-of- as a mere epiphenomenon of critical illness. (IIT) that demonstrated a reduction in surgi- care (POC) testing is a component of nurse- Glycemic control is essential, although the cal critical care mortality led clinicians to seek titrated protocols. Current POC methodolo- optimal target remains a topic of discussion. to evaluate and improve glycemic control in gies are less precise and more expensive The IIT process will need to be computer- their practice. Many protocols were devel- than standard central laboratory methods. ized to provide the most consistent ability oped and implemented in critical care units Potential errors arise from faulty operator to follow complex dosing algorithms, and with varying degrees of effectiveness. The technique, inadequate sample volume and glucose monitoring will need to be far more first protocols were paper-based and varied artifacts due to the altered physiology in ICU automated. Closed-loop insulin titration and greatly in complexity. Computer support is patients (e.g., hypoxia or low hematocrit). continuous monitoring would be most desir- now being developed to make intravenous Subcutaneous continuous glucose monitor- able. (IV) insulin (considered a high-risk drug) safer ing technology is only approved for use in The CareFusion Center for Safety and and easier to use. outpatients. Research with glucose sensors Clinical Excellence hosted an international that may be used in critically ill patients is The benefits of insulin and glucose con- conference that brought together some ongoing. trol were not surprising to endocrinologists of the world’s leaders in glycemic control or cardiovascular surgeons. Early reports Although single-center clinical trials have research, therapy and monitoring to discuss showed that the use of insulin infusions to suggested a benefit to lowering glucose to the latest findings in this area. Summaries of improve glucose control was associated with 80-110 mg/dL, a recent multi-center trial was their presentations and the spirited roundta- prevention of deep sternal wound infections stopped well before the target enrollment ble discussion that concluded the conference and lower mortality. Subsequent studies have because of safety concerns. A large, multi- are compiled in this monograph. added evidence to support the use of IIT center trial (NICE-SUGAR) is underway by the We hope that our readers will recognize to reduce morbidity and mortality in criti- Australia-New Zealand Critical Care Clinical the value of this information and experi- cally ill patients, including a subset of medi- Trials group with results expected in 2008. ence with IIT and that future systems can be cal patients who remain in the intensive care Without more large, prospective trials, ques- designed to achieve optimal safety and effi- unit (ICU) more than three days. Clinicians tions will remain about the optimal (both safe cacy. We wish to express our sincere thanks still struggle to provide IIT to achieve and effective) glucose endpoint. to CareFusion for their commitment to medi- near-euglycemia without causing hypogly- With regard to the future it is clear that cation safety and their sponsorship of this cemia. no matter what research will reveal, clinicians program.

Executive Summary Conference Report 3 7th Invited Conference: A Brief History of Tight Glycemic Control: What We Know in 2007; and How We Got Here

PROCEEDINGS A Brief History of Tight Glycemic Control: What We Know in 2007, and How We Got Here Tony Furnary, MD, Starr-Wood Cardiac Group, Portland, OR

patient) glycemic control strategies in this Key points patient population included subcutaneous • Elevated average three-day postoperative blood glucose (3BG) is a risk factor for morbidity (SQ) insulin therapy in 31% and only oral and mortality in hospitalized patients with diabetes. hyperglycemic agents in 52%, while 12% were • Continuous insulin infusions that control 3BG to near-euglycemic levels normalize the managed with diet control alone, and five diabetic patient outcomes to non-diabetic levels. percent were undiagnosed and not previously treated at the time of admission for cardiac • In patients target glucose level and duration of intravenous (IV) insulin surgery. therapy are both key elements of effective tight glycemic control (TGC). Simply stated, the “3” is just as important as the “BG.” One thousand of the 5,534 patients with diabetes were treated using our first protocol, • The effectiveness of TGC has been proven in: which was a very intensive SQ insulin proto- − Diabetes cardiac surgery patients in the intensive care unit (ICU), col, administering SQ doses of regular insulin − Non-ICU diabetes cardiac surgery patients through three postoperative days, every four hours. The other 4,500 patients were − Medical patients who reside in the ICU longer than three days. managed using some version of the “Portland • Glucose-insulin-potassium (GIK)therapy is not equivalent to continuous insulin infusions. Protocol.” In this protocol blood glucose was assessed for every patient every 30 minutes • Intensive glucose control in the ICU with continuous insulin infusions followed by to two hours throughout their hospital stay. subcutaneous control on the wards have not produced equivalent results to three days In the ICU glucose was measured using blood of IV insulin infusions. from an arterial or venous line, and in non-ICU general nursing units, with blood obtained This brief history reviews the major stud- Portland Diabetic Project using capillary fingerstick. If an ICU patient’s ies of tight glycemic control (TGC) from 1992 blood glucose was very high or very low, it Effects of on cardiac sur- to 2007. These five major studies combined was monitored every 30 minutes, and in the gery patients. The Portland Diabetic Project have evaluated the effects of TGC on more operating room, every 20 to 30 minutes. For was started in 1992 as a prospective, nonran- than 31,000 patients. The studies consid- purposes of data analysis the glycemic state domized interventional study to evaluate the ered here in chronological order include of each patient was described by a single effects of hyperglycemia and its reduction the Portland Diabetes Project1-6 on cardiac number, the average three-day postoperative with continuous intravenous (IV) insulin infu- surgery patients with diabetes, the Diabetes blood glucose (3BG) i.e., the average of all glu- sions on morbidity and mortality in cardiac Mellitus, Insulin Glucose Infusion in Acute cose measurements on the day of surgery and surgery patients. Between 1987 and end of Myocardial Infarction (DIGAMI) Study-17, the first and second postoperative days, derived 2005, 5,534 patients had been enrolled in this Leuven studies by Van den Berghe, et al. in from 24 to 72 glucose measurements made ongoing study. This number included approx- the surgical intensive care unit (SICU)8 and per protocol during that period of time. imately 4,500 coronary bypass, 470 valve, 570 medical ICU (MICU)9, the Stamford ICU studies valve coronary artery bypass graft (CABG) and The Portland Protocol blood glucose tar- by Krinsley10,11, the CREATE-ECLA12 multicenter 60 additional patients who had other cardiac get range became progressively lower over study and the multicenter DIGAMI-2 study13. surgical procedures. Pre-admission (out- time as our goal was to ultimately achieve

4 Executive Summary Conference Report 7th Invited Conference: A Brief History of Tight Glycemic Control: What We Know in 2007; and How We Got Here euglycemia in all patients. Between 1987 and tion rate approximately doubles with every insulin on deep sternal infection rates was 1991 treatment was started with SQ insulin 50 mg/dL increase in blood glucose above published in 19991. Not since the discovery with a target blood glucose of < 200 mg/dL. In 175 mg/dL3. These findings were presented and subsequent clinical introduction of peni- 1992 we were the first to implement intensive in 1995 and published in 1997. It took two cillin in the 1930s and early ‘40s, respectively, glycemic control using continuous insulin years to get these data published, because has there been a non-surgical intervention infusions. Our institutional review board (IRB) the findings were so different from the tra- that has so dramatically altered surgical-site required proof that intensive glycemic con- ditional concept of “benign hyperglyemia" infection rates. trol would not lead to hypoglycemia-related that reviewers were reluctant to believe the Based on these studies we concluded that fatalities, so we began the IV insulin phase results. diabetes itself was not a risk factor for infec- of the project with a target range of 150-200 Investigating further, we evaluated the tion. Rather it was the presence of hypergly- mg/dL and limited this therapy to the ICU. effect of the individual (daily) components cemia in the diabetes population that is the In 1995 after analyzing the initial data we of 3BG on infection rates. We found that with true risk factor for infection. Furthermore, realized that to truly cause change, we had three days of TGC, preoperative hemoglobin this risk can be reduced by 63% through the to implement intensive glycemic control not A1c has no bearing on infection. However, use of three postoperative days of TGC with only in the ICU but also in the operating room preoperative blood glucose > 180 mg/dL continuous insulin infusions. and in non-critical care patient care areas. The does have a significant effect. Blood glucose Acute mortality. In 1999 we presented 3BG concept began when we realized that on the day of surgery has no bearing on data at the American Heart Association that the second major factor in TGC is not where a infection, but blood glucose > 180 mg/dL on compared 3BG levels to mortality in the coro- patient is being treated in the hospital (e.g., in the first, second and third postoperative days nary bypass (CABG) population2. For CABG the ICU or in a general nursing unit) rather, it all have an independent effect on infection patients, when the 3BG was > 200 mg/dL, is the length of time (the duration of glycemic rates14. the mortality was 6% and when the 3BG control since a patient’s acute event) that truly We also found a highly significant differ- was < 200 mg/dL, the mortality was only matters. In 1995 we began using continuous ence between patients whose glucose levels 1.5%. Multivariate analysis showed that 3BG insulin infusions on the non-ICU telemetry were inadequately controlled or only partially is a highly significant independently predic- floor and maintained a target of 150–200 mg/ controlled with SQ insulin compared to those tive variable for mortality. The mortality rate dL. In 1999 we lowered the target range to who were managed using continuous insulin increases by two–fold for every 50 mg/dL 125-175 mg/dL; in 2001, to 100-150 mg/dL; in infusions. In the later group, the infection rate increase in 3BG. 2004, to 80-120 mg/dL; and in 2005, to 70-110 decreased to 0.7% as compared to a rate of mg/dL. In 2005 the 3BG, which includes the At a time when it was commonly thought 2% in the SQ group (p<0.001)1. Interestingly, initial phase of induction through transition that there was nothing wrong with hypergly- at that time, in the mid 1990s, the diabetic to a general nursing unit, for all our patients cemia in the postoperative patient, we estab- cardiac surgery patient population world- was 121 mg/dL. We had essentially eliminated lished hyperglycemia as an independent risk wide had an overall incidence of deep sternal hyperglycemia from our patient population. factor for mortality in CABG patients. Again, infection of 5.6%. So our infection rate of 2%, independent analysis of the various compo- Early results. We first reported on the rela- even in the subcutaneously treated popula- nents of 3BG showed that hemoglobin A1c is tionship between hyperglycemia and cardiac tion was very low compared to that reported not predictive of mortality, nor is preoperative surgery outcomes in a presentation to the in the world literature at the time. This dem- glucose, but elevated glucose is a significant annual meeting of the Society of Thoracic onstrated the significant effects of targeted independent risk for death. Blood glucose Surgeons in 19956. Whether the data were glycemic control even with SQ insulin therapy levels on the first and second postoperative analyzed with a single cutoff of > 175 mg/ when compared to the previous standard of days, and, for patients who remain in the ICU, dL or by 50 mg/dL-increments, our results care14. the third day are also significant predictors of showed that the postoperative level of blood The results obtained with continuous IV in-hospital death. glucose, as measured by 3BG, has an inde- insulin therapy were better still. Multivariate pendent effect on the incidence of deep Thus, the duration of hyperglycemia and, analysis showed that continuous insulin infu- sternal wound infection. Multivariate analysis conversely the duration of tight glucose man- sions independently reduced the risk of infec- showed that the deep sternal wound infec- agement is an important determinant of out- tion by 63%. The independent effect of IV comes related to hyperglycemia in cardiac

Executive Summary Conference Report 5 7th Invited Conference: A Brief History of Tight Glycemic Control: What We Know in 2007; and How We Got Here surgery patients. Thus, the second critical Non-diabetes patients. In 2007 we began lin treatment used in the ICU included IV factor in TGC management (the first being looking at the non-diabetes CABG patient insulin for more than 24 hours, then four SQ target blood glucose level) is not location population. Although we have applied our injections a day for the next three months. of the patient ( ICU or the operating room Portland Protocols to our non-diabetes Mortality was reduced by 20% during the or non-ICU floor); rather, it is the duration of patients with stress hyperglycemia since three to four years patients were evaluated. TGC therapy. It is the critical three-day peri- 1998, we have not seen any reduction in mor- In-hospital mortality was not reduced but od immediately following the seminal ICU tality in this population of patients. We are long-term survival improved. Lower glucose admission event during which hyperglycemia now hoping to randomize our non-diabetes upon admission was associated with lower significantly affects outcomes. For patients patient population with stress hyperglycemia in-hospital mortality. Although this associa- who remain in critical condition, it continues to TGC and non-TGC groups. However, based tion was not significant, there was a trend to affect outcomes for as long as the patient on our preliminary data, TGC may not make towards a lower mortality in the group with remains in the ICU. a difference in the non-diabetes cardiac sur- lower glucose. gery population. SQ vs. continuous insulin infusion. We The patients in the insulin-treated group have shown that continuous insulin infu- DIGAMI-1 also had better long-term survival. These sions reduce absolute unadjusted mortality patients were tightly controlled, versus the Diabetes and acute myocardial infarction rates by more than 50% in CABG patients control group that was not tightly controlled. (AMI). While we were publishing our findings who also have diabetes. Multivariate analy- For patients who were initially not on insulin from 1995 through 1998, others were work- sis shows that the risk-adjusted indepen- therapy at the time of admission for AMI and ing on this problem, including the effect of dent effect of continuous insulin infusions is who were then placed on insulin and very the management of diabetes on AMI mor- to reduce mortality by 65%. tightly controlled, the survival advantage was tality. Most studies had shown that in every even greater over the next few years. Those Our annualized mortality rates show that era of cardiac intervention – from the 1960s findings were published in 1997. after continuous insulin infusions were used through the present – patients with diabe- in the patients with diabetes, by 1995 the teshad a two-fold higher mortality for AMI DIGAMI-2 risk of death was normalized to that of compared to those without diabetes18. Glucose, insulin and potassium (GIK). The patients without diabetes. As the average When thrombolysis became part of car- randomized DIGAMI-2 study by Malmberg, protocol target and actual glucose levels diac care between the mid-1980s and 2000, Lars Ryden, et al. in 200513 included 48 were lowered, results continued to improve. overall mortality decreased, yet diabetes hospitals in six countries and 1,200 patients Between 2000 and 2006 the overall mortal- still had a higher mortality for AMI. Since who were assigned to three treatment arms. ity for patients with diabetes in our hospital 2000, patients with diabetes who have a AMI Group One received a solution of GIK for 24 was 0.9%, compared to the national repost- still have a two-times-higher incidence of hours followed by home insulin therapy. ed Society of Thoracic Surgeons mortality mortality than the total patient population. Group Two was given GIK infusion followed rate of 3.4% in CABG patients with diabe- Therefore, diabetes seems to be a risk factor by standard glucose control. Group Three tes15,16. for death following AMI. had routine metabolic management. No Complications. Publication of our results statistical differences were found between Even in patients without diabetes there showed that increasing 3BG is associated these three groups in terms of outcomes. is a relationship between severe hyperglyce- with an increasing number of complications, There were no differences among the major mia and mortality. Pooled meta-analysis data including death, transfusion, new-onset etiologic factor, glucose and the major pri- show that the pooled risk factor is about 2.8 atrial fibrillation and deep sternal wound mary outcome, mortality. Glucose control to 5.8 or about a four-fold increase in risk. infection. Low-cardiac-output syndrome in all three groups was exactly the same Hyperglycemia is also a risk factor for myocar- and length of stay also increase over time. over time and therefore did not produce a dial infarction. We did not see a relationship between 3BG separation of the outcomes curve. However, 7 and pneumonia, stroke, and other complica- The DIGAMI-1 study reported by Klas a multivariate analysis of mortality in the tions17. Malmberg in 1995 evaluated patients who DIGAMI-2 shows three very significant risk had an AMI and blood glucose concentra- factors for long-term death. Increased age, tions greater than 200 mg/dL. Intensive insu- serum creatinine, heart failure and higher

6 Executive Summary Conference Report 7th Invited Conference: A Brief History of Tight Glycemic Control: What We Know in 2007; and How We Got Here fasting blood glucose concentrations all neg- results similar to the early results from the Stamford / Krinsley atively affected survival. Portland Diabetic Project. In this prospective, Hyperglycemia in medical/surgical ICU 12 randomized study patients were assigned In the CREATE-ECLA trial cardiologists patients–retrospective data review. A corrob- to an IIT (target glucose < 110 mg/dL) or a at 470 centers around the world evaluated orating study by James Krinsley was based on non-IIT group (180-200 mg/dL). It is impor- patients with elevated-S-T myocardial infarc- a retrospective data review of 1,800 patients tant to note that 60% of enrollees were tion. The goal was to evaluate whether use at Stamford Hospital between 1999 and postoperative cardiac surgery patients. IIT of the GIK protocol reduced 30-day mortality 200310. The study showed a direct relation- was associated with a 34% reduction in and other measures in AMI patients. Results ship between increased mean ICU glucose mortality, a 46% reduction in infection, a showed no difference in the primary end- levels and increased mortality in a mixed, 41% reduction in dialysis, a 50% reduction point of mortality. The blood glucose levels in medical/surgical ICU that did not include in transfusions and a marked reduction in the control (non-GIK) group were lower than cardiac surgery patients. In Van den Berghe’s peripheral nerve polyneuropathy. the blood glucose levels in the GIK group, so study, 65% were cardiac surgery patients, and that any advantage that insulin might have The majority of the reduction in mortality the cardiac surgery population itself likely conferred was taken away by the disadvan- occurred in patients who were in the ICU and had a significant effect on the results seen. tage of increased glucose levels. The study kept on insulin infusion for five days or longer. In Krinsley’s non-cardiac surgery ICU popula- showed no reduction in mortality because In the study hospital, insulin infusions are not tion, patients observed with blood glucose the study design did not create a separation used in general nursing care areas. If patients levels of 150 mg/dL had a three-fold increase in the primary variable hypothesized to affect were transferred out of the ICU after the first in mortality compared to the group with the mortality, i.e., blood glucose levels. or second day, they only had one or two days lowest blood glucose values of 90 mg/dL. of TGC and then the glucose concentrations However, if these data are divided into Krinsley concluded that increased glucose increased. A reduction in mortality was not glucose terciles, mortality increased with levels adversely affect mortality rate even in seen in this subset of patients. Therefore, in increasing glucose levels. Patients in the low- non-cardiac surgery populations. Van den Berghe’s first surgical study, there is est third of glucose levels had the lowest Length of stay. In 2006 Krinsley report- confirmation of the duration component of mortality rate. In the middle third mortality ed19 that hyperglycemia was also related to continuous insulin infusions or ITT therapy was higher, and in the highest third there was increased length of stay. Insulin infusions, previously described by the Portland series. significantly higher mortality. Even though which reduce hyperglycemia, were shown to A follow-up study from Leuven showed that this study is considered a negative study, it decrease hospital costs. From the Portland the survival benefit achieved in the hospital shows a relationship between hyperglycemia data, IIT on the day of surgery and the first is maintained up to three or four years after and mortality. and second postoperative days has been surgery. From these studies on AMI one can con- shown to reduce length of stay. Overall, insu- Medical ICU. In 2006 in a population of clude that GIK is not effective in altering lin infusions reduce the length of stay by 1,200 MICU patients, Van den Berghe exam- outcomes. Over the past 40 years multiple about two days in cardiac surgery patients. ined in-hospital mortality between groups studies utilizing GIK have been carried out What Do We Know About TCG at the randomized to TGC (80 –110 mg/dL) or less to investigate its efficacy in reducing mortal- End of 2007? intensive glycemic control (< 180 mg/dL). ity. Not one of these studies has produced a They found a significant effect of ITT on mor- From the Portland study we know that: significant positive result. Outcome improve- tality for patients who remained in the ICU ment has only been associated with the use • Mortality is affected by glucose on the day three days or longer. For those who were in of insulin therapy to achieve glucose control. of surgery, the first day and the second the MICU for less than three days there was day post-surgery, but the effect becomes The Leuven studies / Van den Berghe no apparent effect of ITT on mortality. In insignificant on the third day. Intensive insulin therapy (IIT) in cardiac patients in the ICU for three days or longer, • Infection rates are affected preoperative- surgery patients–prospective, randomized the mortality risk reduction was about 18%, ly, are almost significant are the day of trial. In 2001 Van den Berghe8 published a which was highly significant. surgery (p = 0.7) and are significant landmark trial of 1,500 patients that showed on the first, second, and third day postop- eratively.

Executive Summary Conference Report 7 7th Invited Conference: A Brief History of Tight Glycemic Control: What We Know in 2007; and How We Got Here

• Length of stay is affected on the first, Table. TGC Proven Effectiveness and Hyperglycemia a Proven Risk Factor second and third day, and even preopera- as of December 2007 tively. Hospital Location: Admission OR/DOS ICU Ward-POD#3 > POD #3

• In cardiac surgery patients the “3” is just Patient Population:

as important as the “BG.” Both express DM-CABG I (A) ML (A) MIL (A) I (A) NS (A)

the important terms of this therapy–target Non-DM CABG NS NS (B) NS (B) NS (B) NS (B) level and duration. DM- non-CABG I (A) NS (B) I (A) I (A) NS (B) In general, this is what we know about TGC NonDM,non-CABG Unk Unk Unk Unk Unk in the cardiac surgery population is this Non-CTS Surgical Unk NS (A) MIL (A) Unk Unk (Table): Medical Unk NA M (A) Unk Unk • In CABG patients with diabetes who have AMI Cardiac NS (A) NS (A) NS (A) Unk S (A) hyperglycemia and insulin infusions, TGC Neurologic (CVA) Unk NA + (A) Unk Unk has been shown to be significant on admis- Significant for: M= Mortality, I = Infection, L = LOS; S = Long term Survival; NS = Non-significant; Unk = unknown; NA = Not Applicable; Levels of evidence: A = proven by multiple randomized trials; B = suggested by one or two observational studies; sion, in the operating room, on the day of C = based on consensus opinion, not proven in clinical trials surgery, in the ICU and in the ward. Beyond the third postoperative day, the relation- ship is not significant. For medical ICU patients, Van den Berghe – Medical ICU patients who are in the has shown that if a patient received TGC for ICU longer than three days. • In CABG patients without diabetes, no more than three days, it improves mortal- • Benefit is probable in surgical patients in significant association has been shown ity rates. There are no published data in the the SICU longer than three days. Benefit is between hyperglycemia and mortality, non-cardiac surgery populations about the considered as “probable” because non-car- infection or length of stay. impact of TGC in the operating room and in diac surgery patients were never separated general nursing units. • In the diabetes non-cardiac surgery , non- out from the cardiac surgery patients in the CABG cardiac surgery patients, i.e., isolated In the cardiac myocardial infarction pop- Van den Berghe study. valves, aorta, the only factor associated ulation, TGC has not shown a significant • Benefit is also possible in patients with dia- with glucose control is infection. This asso- impact on admission to or length of stay in betes who have a myocardial infarction and ciation occurs in the ICU and on the ward the ICU. We have not demonstrated changes have percutaneous coronary intervention. out to the third day. Beyond the third in outcomes despite an association between This possible benefit is only inferred from postoperative day there is no association. glucose concentrations and TGC. In surgical adverse data related to hyperglycemia, for Glucose control in the operating room on patients, beyond the third postoperative day the beneficial effects of insulin have never the day of surgery has no significant effect there is a significant association with survival been proven in this patient population. on infection rates. rates based in the results of the DIGAMI-1 • There have been no publications about Study. • Benefit has not been proven, or is unlikely in the impact of glucose control in non-CABG In patients with strokes in the ICU, insu- patients who do not have diabetes. lin infusions decrease the extent of the – Cardiac surgery patients without diabe- Although TGC is being widely advocated, stroke. tes, it has only been shown to be of significant Conclusions – Medical ICU patients in the ICU for less value in about 30% of patient populations. In than three days, non-cardiac surgery patients the supporting • The effectiveness of TGC is proven in: – AMI patients without diabetes. evidence is even less. TGC has been shown – Diabetes cardiac surgery patients in the to have significant impact on mortality, infec- ICU, We can conclude that: tion and length of stay for surgical patients – Non-ICU diabetes cardiac surgery • 3BG is a true risk factor for morbidity and while they are in the ICU. patients through three days, mortality in CABG patients with diabetes

8 Executive Summary Conference Report 7th Invited Conference: A Brief History of Tight Glycemic Control: What We Know in 2007; and How We Got Here

• Continuous insulin infusions that control 5. Furnary AP, Chaugle H, Zerr K, Grunkemeier G. 13. Malmberg K, Ryden L, Wedel H, et al. Intense meta- Postoperative hyperglycemia prolongs length bolic control by means of insulin in patients with 3BG can normalize the diabetic patient of stay in diabetic CABG patients. Circulation. diabetes mellitus and acute myocardial infarction outcomes to non-diabetic levels. 2000;102(18):II-556. (DIGAMI 2): effects on mortality and morbidity. Eur Heart J. Apr 2005;26(7):650-61. • GIK is not equivalent to continuous insulin 6. Zerr KJ, Furnary AP, Grunkemeier GL, Bookin S, Kanhere V, Starr A. Glucose control lowers the 14. Furnary AP, Wu Y. Eliminating the diabetic disadvan- infusions. risk of wound infection in diabetics after open tage: the Portland Diabetic Project. Semin Thorac heart operations. Annals of Thoracic Surgery. Cardiovasc Surg. Winter 2006;18(4):302-8. • TCG in the ICU with continuous infusions 1997;63(2):356-61. 15. Brown JR, Edwards FH, O'Connor GT, Ross CS, followed by SQ control in general nursing 7. Malmberg K. Prospective randomised study of Furnary AP. The diabetic disadvantage: historical units has not produced equivalent results intensive insulin treatment on long term survival outcomes measures in diabetic patients under- to three days of IV insulin infusions. after acute myocardial infarction in patients with going cardiac surgery--the pre-intravenous insu- diabetes mellitus. DIGAMI (Diabetes Mellitus, Insulin lin era. Semin Thorac Cardiovasc Surg. Winter References Glucose Infusion in Acute Myocardial Infarction) 2006;18(4):281-8. Study Group. Bmj. May 24 1997;314(7093):1512-5. 1. Furnary AP, Zerr KJ, Grunkemeier GL, Starr A. 16. Furnary AP. Diabetes, hyperglycemia, and the car- Continuous intravenous insulin infusion reduces 8. van den Berghe G, Wouters P, Weekers F, et al. diac surgery patient: introduction. Semin Thorac the incidence of deep sternal wound infection in Intensive insulin therapy in the critically ill patients. Cardiovasc Surg. Winter 2006;18(4):278-80. N Engl J Med. Nov 8 2001;345(19):1359-67. diabetic patients after cardiac surgical procedures. 17. Furnary AP, Cheek DB, Holmes SC, Howell WL, Kelly Ann Thorac Surg. Feb 1999;67(2):352-60; discussion 9. Van den Berghe G, Wilmer A, Hermans G, et al. SP. Achieving tight glycemic control in the operating 360-52. Intensive insulin therapy in the medical ICU. N Engl room: lessons learned from 12 years in the trenches 2. Furnary AP, Zerr KJ, Grunkemeier GL, Heller CA. J Med. Feb 2 2006;354(5):449-61. of a paradigm shift in anesthetic care. Semin Thorac Cardiovasc Surg. Winter 2006;18(4):339-45. Hyperglycemia: A predictor of mortality following 10. Krinsley JS. Association between hyperglycemia CABG in diabetics. Circulation. 1999;100(18):I-591. and increased hospital mortality in a heteroge- 18. Braunwald E. Shattuck lecture--cardiovascular 3. Furnary AP, Wu Y, Bookin SO. Effect of hyperglyce- neous population of critically ill patients. Mayo Clin at the turn of the millennium: triumphs, mia and continuous intravenous insulin infusions Proc. Dec 2003;78(12):1471-8. concerns, and opportunities. N Engl J Med. Nov 6 1997;337(19):1360-9. on outcomes of cardiac surgical procedures: the 11. Krinsley JS. Effect of an intensive glucose manage- Portland Diabetic Project. Endocr Pract. Mar-Apr ment protocol on the mortality of critically ill adult 19. Krinsley JS. Glycemic control, diabetic status, and 2004;10 Suppl 2:21-33. patients. Mayo Clin Proc. Aug 2004;79(8):992-1000. mortality in a heterogeneous population of criti- cally ill patients before and during the era of 4. Furnary AP, Gao G, Grunkemeier GL, et al. 12. Mehta SR, Yusuf S, Diaz R, et al. Effect of glucose- intensive glycemic management: six and one-half Continuous insulin infusion reduces mortality in insulin-potassium infusion on mortality in patients years experience at a university-affiliated commu- patients with diabetes undergoing coronary artery with acute ST-segment elevation myocardial infarc- nity hospital. Semin Thorac Cardiovasc Surg. Winter bypass grafting. J Thorac Cardiovasc Surg. May tion: the CREATE-ECLA randomized controlled trial. 2006;18(4):317-25. 2003;125(5):1007-21. Jama. Jan 26 2005;293(4):437-46.

Executive Summary Conference Report 9 7th Invited Conference: Intensive Insulin Therapy and the NICE-SUGAR Study

PROCEEDINGS Intensive Insulin Therapy and the NICE-SUGAR Study Simon Finfer, MB, BS, FRCP, FRCA, FJFICM, Director of Critical Care and Trauma, George Institute for International Health, Professor, Faculty of Medicine, University of Sydney, Sydney, Australia

Key points tion was observed in the control hospitals it illustrates the danger of attributing clinical • The landmark studies by Van den Berghe et al. have made control of blood glucose a changes to specific interventions. central part of the management of critically ill patients, but other randomized, controlled trials have been unable to replicate their results. A second study evaluated the mortality rates in patients presenting with severe sep- • As a result of the two conflicting Van den Berghe trials and concerns over case mix, higher- sis to emergency departments of Australian than-expected mortality in the control group, and routine use of high-dose intravenous and New Zealand hospitals. These data were glucose, intensive-care clinicians remain uncertain whether to use intensive insulin therapy obtained while planning a phase III trial to (IIT) in their patients. validate Rivers’ study of early goal-directed • The largest trial of IIT will be the NICE-SUGAR study, which compares the effects of the therapy5. There was a steady decrease in the a two blood glucose targets—4.5-6.0 mmol/L (80-110 mg/dL ) and 8.0–10.0 mmol/L case fatality rate with crude hospital mortal- (180-200 mg/dL)—on 90-day, all-cause mortality in intensive care patients who are ity decreasing from 37.7% in 1998 to 21.2% in predicted to be in the intensive care unit on more than two calendar days. 20056. There were no specific interventions to which this decrease could be attributed, illus- • If the NICE-SUGAR study demonstrates a favorable treatment effect, maintaining trating that dramatic reductions in mortality normoglycemia will most likely become a treatment standard worldwide. may occur over time both with and without changes in management. For common con- ditions, such as hyperglycemia in critical ill- Published observational data show a deaths (Figure 1) was observed. It would ness, decisions about appropriate choice of decrease in unwanted outcomes accompany- be natural to attribute this reduction to the therapy should be made on the basis of RCTs. ing improved glycemic control in cardiotho- introduction of the METs, but as the reduc- Observational data are useful for confirming racic surgery patients and in those admitted to a mixed population intensive care unit (ICU)1-3. While it is tempting to conclude that Figure 1. these data are evidence of “cause and effect,” 8 P=0.03 other reports suggest that it is hazardous to 7 Baseline period Study period make such inferences. Two examples illus- dmisions 6 trate this point: 5 4 P=0.004 3 The MERIT study was a cluster, random- e/1,000 A P=0.04 2 ized, controlled trial (RCT) that studied the 1 impact of introducing medical emergency Incidenc 0 teams (METs) on unanticipated ICU admis- rest regate sions, and on cardiac arrest and death in Agg patients without a DNR order in Australian Cardiac ar ICU admission Death (No NFR) hospitals4. In this study, a significant reduc- Incidence of aggregate outcome, cardiac arrest, death and unanticipated ICU admissions during baseline and study peri- ods in control hospitals in the MERIT study. (From data in [4]). tion in the incidence of cardiac arrests and

10 Executive Summary Conference Report 7th Invited Conference: Intensive Insulin Therapy and the NICE-SUGAR Study the findings of RCTsand for investigating rare were not able to predict accurately how long The two blood glucose targets are achieved but serious side effects, but are subject to each patient was likely to stay in the ICU. The with the aid of a web-based algorithm. The large errors due to bias7,8. publication of this second study has increased use of this algorithm promotes uniform blood clinicians’ uncertainty over the role of IIT in glucose management in all study sites and Intensive Insulin Therapy (IIT) critically ill patients. Van den Berghe called for enables the study management committee to Studies–Van den Berghe et al. additional large-scale RCTs of at least 5,000 monitor blood glucose management within In their first randomized trial, Van den participants to answer the important ques- the study. It is therefore known whether the Berghe et al. randomized 1,548 surgical inten- tion of whether IIT reduces mortality in ICU blood glucose targets are being met with suf- sive care patients to receive insulin to main- patients. Van den Berghe’s call was supported ficient separation between the two groups. tain blood glucose between 4.4-6.1 mmol/L by an accompanying editorial highlighting After more than 1.5 million hours of blood (80-110 mg/dLa) (intensive insulin group) the need for further study to answer this glucose management, the average blood glu- or between 10-11.1 mmol/L (180-200 mg/ important question12. cose derived from measurements entered 9 dL) (conventional insulin group) . The study into the treatment algorithm is 5.9mmol/L The NICE-SUGAR Study reported an absolute reduction in hospital (106.2 mg/dL) in the lower range group versus mortality of 3.7% (relative risk reduction, RRR As a result of the two conflicting Van den 8.4mmol/L (151.2 mg/dL) in the higher range 9 33%) with IIT . Other benefits reported in Berghe trials and concerns over case mix, group. This compares with 5.7 vs. 8.5mmol/L the intensive insulin group were a reduc- higher-than-expected mortality in the control (102.6 vs. 153.0 mg/dL) in the first Van den tion in hospital stay, blood stream infections, group and routine use of high-dose IV glucose, Berghe study and 6.2 vs. 8.5mmol/L (111.6 vs. acute renal failure requiring dialysis, incidence ICU clinicians are still uncertain about using IIT 153.0 mg/dL) in the second study. The aver- of critical-illness polyneuropathy and blood in their patients. To resolve this uncertainty, a age time on study treatment is 386 hours or transfusions. The external validity of the National Health and Medical Research Council 16.1 days. results has been questioned because study (NHMRC)-funded RCT of IIT commenced in The major safety concern with IIT is patients received high doses of intravenous Australia and New Zealand in 2005, and in hypoglycemia. The rate of hypoglycemia for (IV) glucose and the control group mortality 2006 the Normoglycaemia in Intensive Care patients randomized to the low-range arms was unexpectedly high. Many consider the Evaluation (NICE) Study Investigators joined of the two Van den Berghe trials was 5% and RRR to be implausible10. There was no differ- with the Survival Using Glucose Algorithm 18% respectively9,11. In 4,450 patients the rate ence in the number of deaths occurring dur- Regulation (SUGAR) trial investigators of the of hypoglycemia in the low-range group of ing the first five days in intensive care, and the Canadian Critical Care Trials Group to com- the NICE-SUGAR study is 10.2 events per reduction in mortality was limited to patients plete a single trial thereafter called the NICE- 100 patients, towards the lower end of the receiving more than five days’ treatment in SUGAR study13. rates reported for this treatment. All episodes the ICU. The incidence of hypoglycemia was The primary aim of the NICE-SUGAR study of hypoglycemia are classified as serious significantly increased in the intensive insulin is to compare the effects of the two blood adverse events (SAEs) and reported to partici- group (39 patients) compared to those in the glucose targets on 90-day, all-cause mortality pating centers’ ethics committees and to the conventional glucose group (6 patients). No in intensive care patients who are predicted study’s independent data and safety monitor- long-term sequelae from hypoglycemia were to be in the ICU on more than two calendar ing committee. All SAEs have been followed detected. days. The null hypothesis is that there is no dif- up by the study management committee and In February 2006, Van den Berghe et al. ference in the relative risk of death between to date there have been no harmful sequelae published a second RCT in 1200 critically ill patients assigned a blood glucose target of detected. medical patients expected to be treated in the 4.5-6.0 mmol/L (81-108 mg/dL) and those The data from the two Van den Berghe ICU for three or more days11. The study did not assigned a blood glucose target of less than studies suggest that in a combined medical find a significant reduction in mortality in the 10.0 mmol/L (<180 mg/dL) with insulin being and surgical population a RRR of 14% is a intention-to-treat population, although in an infused if blood glucose exceeds 10.0 mmol/L more appropriate target, and 6,100 patients a priori subgroup of 767 patients who were (180 mg/dL) and adjusted when needed to will be included in NICE-SUGAR to provide in the ICU on three or more calendar days, maintain blood glucose of 8.0–10.0 mmol/L 90% power to detect a 14% RRR from a base- 90-day mortality was reduced from 49.1% to (144-180 mg/dL). line mortality of 30% (α < 0.05). 42.2% (RRR 14.1%, p=0.06). The investigators

Executive Summary Conference Report 11 7th Invited Conference: Intensive Insulin Therapy and the NICE-SUGAR Study

NICE-SUGAR Research Planb • In the subgroup of patients admitted with b. See electronic supplement to Angus and Patient Selection diagnosis of traumatic brain injury, long- Abraham, 200513 term functional status will be determined The treatment effect in the first Van den References by Extended Glasgow Outcome Scores Berghe study was limited to patients who 1. Krinsley JS. Effect of an intensive glucose manage- (GOSE) at six months and two years. were in the ICU for five or more days. In the ment protocol on the mortality of critically ill adult patients. Mayo Clin Proc 2004;79:992-1000. second study only patients expected to be in Organization and Collaboration the ICU for three days were included. In the 2. Furnary AP, Zerr KJ, Grunkemeier GL, et al. The study is being conducted as a col- Continuous intravenous insulin infusion reduces NICE-SUGAR study patients expected to be laboration among the Australian and New the incidence of deep sternal wound infection in discharged alive or die before the end of the diabetic patients after cardiac surgical procedures. Zealand Intensive Care Society Clinical Trials day following admission are not being includ- Ann Thorac Surg 1999;67:352-60. Group (ANZICS CTG), the Canadian Critical ed. To exclude patients who will stay in the 3. Furnary AP, Gao G, Grunkemeier GL, et al. Care Trials Group (CCCTG) and The George Continuous insulin infusion reduces mortal- ICU for more than two calendar days but who Institute for International Health and over- ity in patients with diabetes undergoing coronary have a very low risk of death, patients who artery bypass grafting. J Thorac Cardiovasc Surg seen by the NICE-SUGAR study management are able to eat (or who are tube-fed due to 2003;125:1007-21. committee. Data analysis, data sharing, and pre-existing bulbar or laryngeal dysfunction) 4. Hillman K, Chen J, Cretikos M, et al. Introduction publication regulations will involve all inves- of the medical emergency team (MET) system: a and patients who do not have an arterial line tigators according to ANZICS CTG guide- cluster-randomized controlled trial. Lancet 2005; as part of their routine management are also 365:2091-7. lines and will be regulated by memoranda of being excluded. Patients who are moribund 5. Rivers E, Nguyen B, Havstad S, et al. Early goal- understanding. and at imminent risk of death (brain death or directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001;345:1368-77. cardiac standstill) are excluded on the basis The group assembled for this study that treatment allocation cannot alter their includes epidemiologists and intensive 6. The Australasian Resuscitation in Sepsis Evaluation (ARISE) Investigators and the Australian and New outcome. Randomization is achieved via a care who provide the expertise Zealand Intensive Care Society (ANZICS) Adult password-protected, encrypted, secure study and clinical and research skills to conduct Patient Database (APD) Management Committee. The outcome of patients with sepsis and sep- website with patients allocated to receive this study. The collaboration between the tic shock presenting to emergency departments one of two target ranges for glycemic control Australian and New Zealand and Canadian in Australia and New Zealand. Critical Care and Resuscitation 2007;9:8-18. in the ICU. A minimization program stratifies Critical Care Trials Groups and the Mayo Clinic treatment allocation by type of critical illness will provide reliable evidence about the com- 7. Collins R, MacMahon S. Reliable assessment of the effects of treatment on mortality and major mor- (medical vs. surgical) and by region: Australia parative effects of different targets for blood bidity, I: clinical trials. Lancet 2001;357:373-80. and New Zealand or North America. glucose concentration in patients treated in 8. MacMahon S, Collins R. Reliable assessment of the Australasian and North American inten- Study Outcomes the effects of treatment on mortality and major sive care setting. morbidity, II: observational studies. Lancet Primary outcome measure: 2001;357:455-62. Summary 9. Van den Berghe G, Wouters P, Weekers F, et al. • All-cause, 90-day mortality Intensive insulin therapy in critically ill patients. The two studies conducted by Van den N Engl J Med 2001;345:1359-67. Secondary outcomes: Berghe et al have made control of blood glu- 10. Bellomo R, Egi M. Glycemic control in the intensive • Death in the ICU, by day 28 and before cose an important issue in the management care unit: why we should wait for NICE-SUGAR. hospital discharge of critically ill patients. To date, other RCTs Mayo Clin Proc 2005;80:1546-8. have been unable to replicate the results of 11. Van den Berghe G, Wilmer A, Hermans G, et al. • Length of ICU stay these studies. The NICE-SUGAR Study will be N Engl J Med 2006;354:449-61. • Length of hospital stay the largest trial of IIT, and if it demonstrates a 12. Malhotra A. Intensive Insulin in Intensive Care. N Engl J Med 2006, 354: 516-8. favorable treatment effect, maintaining nor- • The need for organ support (inotropes, moglycemia will most likely become a treat- 13. Angus DC, Abraham E. Intensive Insulin Therapy renal replacement therapy and positive in Critical Illness. Am J Respir Crit Care Med 2005; ment standard worldwide. pressure ventilation) 172:1358-9. a. mmol/L = (md/dL x 10) divided by atom- • Incidence of bloodstream infections ic weight of glucose (MW = 180), i.e., • Incidence and severity of hypoglycemia mmol/L X 18 = mg/dL of glucose

12 Executive Summary Conference Report 7th Invited Conference: European Multi-center Trials with Tight Glucose Control by Intensive Insulin Therapy

PROCEEDINGS European Multi-center Trials with Tight Glucose Control by Intensive Insulin Therapy

Philippe Devos, MD, Jean-Charles Preiser, MD, PhD, Department of General Intensive Care, University Hospital of Liege–Sart-Tilman, Liege, Belgium

Key points ferent teams tried to confirm the results of the Leuven team in prospective, random- • In contrast to the decreases in mortality and low severity of adverse effects reported when ized trials, including the German VISEP insulin infusion rate was titrated to keep blood glucose levels between 80-110 mg/dL, these trial, the Australian NICE-SUGAR trial and benefits were not confirmed in multi-center prospective studies. the European Gluconcontrol study, while • Retrospective data analysis found an association between a mean blood glucose level others analyzed retrospectively collected below 140-150 mg/dL and improved outcome. data9,10,11,12.

• Currently unresolved issues in intensive insulin therapy (IIT) include the optimal blood Multi-center Trials of TGC IIT glucose target, the effects of high variability in blood glucose levels, the risks and hazards of VISEP. The German Competence Network hypoglycemia and the potential influence of an underlying disorder on the effects of tight Sepsis (SepNet), a publicly funded, indepen- glucose control (TGC). dent, collaborative study group, designed • Although recommendations regarding the practical aspects of TGC IIT cannot presently be the randomized VISEP trial13 to address two made, an intermediate target level for blood glucose seems to be associated with the lowest questions in a group of septic patients (col- risk-to-benefit ratio. loids versus crystalloids and TGC IIT). This trial was stopped for safety reasons after 488

Tight glucose control (TGC) by intensive in the structure of macromolecules3,4. In the patients in 17 centers were enrolled between insulin therapy (IIT) is defined as the main- presence of high glucose concentrations, April 2003 and March 2005. Of these, 247 tenance of “normoglycemia” (blood glucose several steps in the glycolytic pathways can received intensive insulin therapy (IIT [goal: levels 80-110 mg/dL). Achieving TGC by induce the release of toxic derivates. These 80-110 mg/dL]) and 241 received conven- the titration of intravenous (IV) insulin has effects, sometimes collectively referred to tional insulin therapy (CIT, [goal: 180-200 mg/ become a major topic of interest. The com- as “the Brownlee theory,” are reversible with dL]). Interim data analysis showed that 30 mon condition of “stress hyperglycemia” as the pharmacological inhibition of Poly-ADP- patients (12.1%) treated with IIT developed a physiological response to a critical illness ribosyl-polymerase5, suggesting the involve- hypoglycemia, compared to 5 patients (2.1%) was challenged by the report of Greet Van ment of the activation of this enzymatic com- treated with CIT (p < 0.001). No adverse event den Berghe and co-workers in Leuven. They plex of nuclear repair enzymes. This is prob- was classified as leading to patient death. No reported a 4% decrease in absolute mortality ably related to the involvement of reactive differences were found in the 28-day (21.9% associated with TGC IIT in a surgical intensive oxygen intermediates in the toxic effects of vs 21.6%; p = 1.0) and 90-day mortality rates care unit (ICU) population1. The beneficial hyperglycemia6,7. (32.8% vs 29.5%; p = 0.43) for IIT and CIT, effects of TGC IIT were partially confirmed by respectively. Since the observed rate of hypo- These clinical and biochemical findings the same team in a medical ICU population2. glycemia was considered unacceptably high support the concept of hyperglycemia as a and since there was no treatment efficacy New insights into the mechanisms of mediator for rather than a marker of criti- (no significant difference in 28- or 90-day glucose toxicity have also been described. cal illness. Proof that hyperglycemia is an mortality), the Independent Data Monitoring Severe hyperglycemia was found to induce independent risk factor for ICU mortality in Committee (IDMC) strongly recommended 8 acute changes in cellular metabolism and critically ill patients is lacking . Several dif- that the insulin arm of the trial be stopped.

Executive Summary Conference Report 13 7th Invited Conference: European Multi-center Trials with Tight Glucose Control by Intensive Insulin Therapy

Glucontrol. This prospective, randomized, for critically ill patients in ICUs worldwide can Detrimental Effects of controlled, multi-center study14 compared the be recommended. High-glucose Variability effects of TGC IIT to a control group with less Optimal Target Egi and colleagues18 performed a multi- abnormal blood glucose concentrations than for Blood Glucose variate logistic regression analysis of retro- in patients in the Leuven studies (140-180 spectively collected data from 7,049 critically mg/dL). The primary outcome measure was The answer to the question of optimal ill patients. The coefficient of variability cal- ICU mortality. Twenty-one ICUs participated blood glucose target level can probably be culated from the standard deviation of blood on a voluntary basis (i.e., no financial incen- inferred from clinical data rather than from glucose values recorded for each patient tive or defrayment of study-related costs). experimental findings. Indeed, in the various appeared to be closely related with survival. This study was stopped for safety reasons studies the detrimental effects of hyperglyce- In patients with diabetes, blood glucose vari- by the Data Safety Monitoring Board (DSMB) mia were observed in the presence of blood ability was a stronger predictor of ICU mortal- after the first interim analysis because of a glucose levels higher than those observed ity than was the absolute blood glucose value. high rate of unintended protocol violations. in patients and therefore could be irrelevant Outside the ICU, recent data recorded in dia- A total of 1,011 patients (550 in the IIT arm, for the determination of the optimal glyce- betic patients and compared to volunteers 551 in the CIT arm) were enrolled. Patient mia. Based on the data from the two Leuven 1,2 also indicate that blood glucose fluctuations characteristics (median age 65 years, medical studies , blood glucose > 200 mg/dL can increase the oxidative stress19. These clinical patients 41%, males 62.7%, APACHE II score at probably no longer be considered an accept- data may reflect “cellular” data that showed admission 16.5 ± 7.0) did not differ between able target for insulin therapy in critically cell damage to be most prominent when groups. From the time of admission the mean ill patients. However, the issue of the safest blood glucose changed rapidly from a normal blood glucose levels calculated from individu- range below this level is still unresolved and to an elevated level (reviewed in Brownlee3). al blood glucose values were higher in the CIT has not been specifically addressed in pro- than in the IIT group with a median value of spective clinical trials to date. This potentially important issue of glucose variability was not analyzed in the large trials 119 (IQR 110-131) mg/dL in the IIT group and Three large retrospective trials9,10,11 found performed in critically ill patients published 147 (IQR 128-165) mg/dL in the CIT group, that blood glucose levels < 140 mg/dL were to date. In the Glucontrol study14, the blood p < 0.0001. The adherence to the experi- associated with an improved outcome com- glucose standard deviation was identical in mental protocol was confirmed by the pro- pared with higher levels. portion of time spent in the assigned range the two treatment arms. (40.8% and 38.2% for the IIT and CIT groups, Ideally, the optimal target for blood glu- One implication of the discovery of the respectively). The ICU mortality was slightly cose levels should be defined by large pro- 15 importance of keeping blood glucose as sta- higher in the IIT compared to the CIT group spective trials comparing two ranges . The ble as possible could be to favour the use (16.7% versus 15.2%, NS). Multivariate analy- Normoglycaemia in Intensive Care Evaluation of strict algorithms to maintain blood glu- sis showed a significant association between and Survival Using Glucose Algorithm cose within a narrow range. Although several APACHE II and SOFA scores on admission and Regulation (NICE-SUGAR) and the Glucontrol different validated algorithms are available, higher mortality. The rate of hypoglycemia study were designed and launched to com- indices of blood glucose variability usually was higher in the IIT (9.8 %) than in the CIT pare the effects of insulin therapy titrated to were not assessed and not used to compare group (2.7%, p < 0.0001). Assignment to target blood glucose levels of 80-110 mg/ different protocols. the IIT group, death in the ICU, and APACHE dL versus 140-180 mg/dL. The results of the II scores were significantly associated with Glucontrol study to date suggest that a blood Risks and Hazards of Hypoglycemia glucose target of 140-180 mg/dL is safer hypoglycemia. Hypoglycemia is the major fear when start- than 80-110 mg/dL. Even though further ing IIT and justified the interruption of the Multi-center Trial Results. The currently confirmation of these findings is desirable, two European multi-centre prospective trials available results of both multi-center trials most clinicians presently use this intermedi- mentioned above. Even if the incidence of do not seem to confirm the Leuven data and ate range of 140-180 mg/dL as a target for hypoglycemia was substantial in both Leuven actually raise additional clinically important 16, 17 IIT . 1,2 concerns, questions and difficulties that must studies , the condition of the patients expe- be resolved before widespread use of TGC IIT riencing hypoglycemia was not worsened. Of note, blood glucose monitoring was very

14 Executive Summary Conference Report 7th Invited Conference: European Multi-center Trials with Tight Glucose Control by Intensive Insulin Therapy tight, which implies that the duration of the susceptible to the deleterious effects of hyper- 6. Szabo C, Biser A, Benko R, et al. Poly(ADP-Ribose) polymerase inhibitors ameliorate nephropathy of hypoglycemic episodes was definitely short. glycemia (reviewed in Devos et al. 22). type 2 diabetic Leprdb/db Mice. Diabetes 2006; Therefore, the possibility that long-lasting 55:3004-12. Interventional studies performed in spe- hypoglycemia may be deleterious or even cific subgroups or, at least, subgroup analyses 7. Ceriello A. Oxidative stress and diabetes-associated life-threatening cannot be ruled out. Using IIT complications. Endocr Pract 2006; 12(Suppl 1):60-2. of the large multi-center trials are needed titrated to maintain normoglycemia requires 8. Corstjens AM, van der Horst IC, Zijlstra JG, et al. to define the categories of patients that will careful blood glucose monitoring, since the Hyperglycaemia in critically ill patients: marker or selectively benefit from IIT. Meanwhile, the mediator of mortality? Crit Care 2006; 10:216. classical neurological symptoms can be offset use of an intermediate blood glucose target 9. Krinsley JS. Effect of an intensive glucose manage- by sedation or by an underlying impairment is probably more prudent and is presently ment protocol on the mortality of critically ill adult of the mental status. patients. Mayo Clin Proc 2004; 79:992-1000. recommended23,24,25,26. Some categories of patients with signifi- 10. Finney SJ, Zekveld C, Elia A, et al. Glucose control Conclusions and mortality in critically ill patients. JAMA 2003; cant dysfunctions of neoglucogenic organs 290:2041-7. (liver and kidney), with adrenal failure lead- In spite of the findings that mortality can 11. Gabbanelli V, Pantanetti S, Donati A, et al. Correlation ing to an impaired responsiveness of coun- be decreased in critically ill patients by TGC between hyperglycemia and mortality in a medical ter-regulatory hormones, or with a delayed and restoring normal blood glucose values and surgical intensive care unit. Minerva Anestesiol 2005; 71:717-25. elimination of insulin could experience longer using IIT, several important questions are still 12. Ouattara A, Lecomte P, Le Manach Y, et al. Poor episodes of hypoglycemia. The effects of TGC unanswered. These include the issues of the intraoperative blood glucose control is associated IIT in these subgroups need to be carefully best target range, the importance of mini- with a worsened hospital outcome after cardiac assessed. mizing blood glucose variability, the avoid- surgery in diabetic patients. 2005; 103:687-94. ance of hypoglycemia and the delineation Potential Influence of the Underlying of the categories of patients in whom the 13. Brunkhorst FM, Kuhnt E, Engel C et al. Intensive Disorder on the Effects of TGC IIT insulin therapy in patients with severe sepsis and restoration of “normal” blood glucose is most septic shock is associated with an increased rate of In the second Leuven study2, the improve- beneficial. With the notable exception of the hypoglycaemia–results from a randomized multi- center study. Abstr. Infection 2005; 33(Suppl 1):19. ment in mortality was seen only in patients VISEP trial, the titration of insulin in order (http://webanae.med.uni-jena.de/WebObjects/ who were in the ICU three days or longer. to maintain blood glucose < 180 mg/dL is DSGPortal.woa/WebServerResources/sepnet/visep. Mortality tended to increase in patients with supported by the currently available clinical html) a shorter length of stay who were random- data, and an improvement in outcome was 14. Devos P, Preiser J, Mélot C, on behalf of the Glucontrol steering committee. Impact of tight ized to the IIT group. Some secondary out- consistently associated with blood glucose glucose control by intensive insulin therapy on 20 come variables (see Devos and Preiser for < 140-150 mg/dL. ICU mortality and the rate of hypoglycaemia: Final discussion) such as requirement for dialysis, results of the glucontrol study. Crit Care Med 2007– References Abstract: oral presentation # 0735 (supplement for incidence of bacteremia, requirement for pro- the annual congress): In press. 1. Van den Berghe G, Wouters P, Weekers F, et al. longed antibiotic therapy, incidence of hyper- Intensive insulin therapy in the critically ill patients. 15. Angus DC, Abraham E. Intensive insulin therapy bilirubinemia and “hyperinflammation” were N Engl J Med 2001; 345:1359-67. in critical illness. Am J Respir Crit Care Med 2005; 172:1358-9. not improved in the IIT group. These differ- 2. Van den Berghe G, Wilmer A, Hermans G, et al. ences could point out subgroups of patients Intensive insulin therapy in the medical ICU. N Engl 16. McMullin J, Brozek J, Jaeschke R, et al. Glycemic J Med 2006; 354:449-61. control in the ICU: a multicenter survey. Intensive that do not benefit from IIT. Another category Care Med 2004; 30:798-803. of patients that may not benefit from IIT is the 3. Brownlee M. Biochemistry and molecular cell biolo- gy of diabetic complications. Nature 2001; 414:813- 17. Devos Ph, Ledoux D, Preiser JC, on behalf of the subset of patients with pre-existing diabetes, 20. GLUCONTROL Steering Committee. Current prac- tice of glycaemia control in European intensive care as shown by the aggregation of the results of 4. Hirsch IB, Brownlee M. Should minimal blood glu- units (ICUs). Abstr. Intensive Care Med 2005; 31:130. both Leuven studies21. cose variability become the gold standard of glyce- mic control?J Diabetes Complications 2005; 19:178- 18. Egi M, Bellomo R, Stachowski E, et al. Variability of At the present stage, there is no definite 81. blood glucose concentration and short-term mor- tality in critically ill patients. Anesthesiology 2006; answer to the question of which subgroups 5. Garcia Soriano F, Virag L, Jagtap P, et al. 105:244-52. are likely to benefit more from IIT. Patients Diabetic endothelial dysfunction: the role of poly(ADP-ribose) polymerase activation. Nat Med 19. Monnier L, Mas E, Ginet C, et al. Activation of oxida- with myocardial ischemia and after cardiac 2001; 7:108-13. tive stress by acute glucose fluctuations compared surgery may represent a subset of patients with sustained chronic hyperglycemia in patients with type 2 diabetes. JAMA 2006; 295:1681-7.

Executive Summary Conference Report 15 7th Invited Conference–European Multi-center Trials with Tight Glucose Control by Intensive Insulin Therapy

20. Devos P, Preiser JC. Is it time for implementation of tight glycaemia control by intensive insulin therapy in every ICU ? Crit Care 2006; 10(2):130.

21. Van den Berghe G, Wilmer A, Milants I, et al. Intensive insulin therapy in mixed medical/surgical intensive care units: benefit versus harm. Diabetes 2006; 55:3151-9.

22. Devos P, Chiolero R, Van den Berghe G, et al. Glucose, insulin and myocardial ischaemia. Curr Opin Clin Nutr Metab Care 2006; 9:131-9.

23. Devos P, Preiser JC. Tight blood glucose control: a recommendation applicable to any critically ill patient? Crit Care 2004; 8:427-9.

24. Preiser JC, Devos P, Van den Berghe G. Tight control of glycaemia in critically ill patients. Curr Opin Clin Nutr Metab Care 2002; 5:533-7.

25. McMahon MM, Miles JM. Glycemic control and nutrition in the intensive care unit. Curr Opin Clin Nutr Metab Care 2006; 9:120-3.

26. Dellinger RP, Carlet JM, Masur H, et al. Surviving Sepsis Campaign guidelines for management of severe sepsis and septic shock. Crit Care Med 2004; 32:858-73.

16 Executive Summary Conference Report 7th Invited Conference: Implementation of Tight Glycemic Control at Stamford Hospital

PROCEEDINGS Implementation of Tight Glycemic Control at Stamford Hospital James Krinsley, MD, FCCM, FCCP, Director of Critical Care, Stamford Hospital, Professor of Clinical Medicine Columbia University College of Physicians and Surgeons, Stamford, CT

Key points was agreed that 140 mg/dL would be an acceptable initial treatment threshold. As a • Elements of a successful implementation of a tight glycemic control (TGC) protocol include result, a 80-140 mg/dL target was used in the multidisciplinary collaboration; prior staff experience with complex protocols; an agreed- 1,600 patient before-and-after interventional upon, achievable blood glucose target; robust data-analysis capabilities; and clinician study3. access to relevant intensive care unit (ICU) outcome data. This experience highlights some of the • Avoidance of deleterious severe hypoglycemia requires recognition of the risk factors such important principles about successful imple- as severity of illness, the presence of renal failure (delayed clearance of insulin), sepsis mentation of tight glycemic control (TGC). and liver failure (impaired gluconeogenesis), discontinuation of source of calories with Developing and implementing an ITT proto- continuation of insulin therapy, and diabetes. col is most successful where there is multi- • TGC has emerged as a standard of care in the ICU because of its biologic plausibility and disciplinary collaboration. Before addressing the available interventional studies. Further corroboration may come from the ongoing TGC, nearly all routine aspects of care in the NICE-SUGAR trial. unit had been “protocolized.” The nurses and

• The appropriate treatment threshold–110, 120, 140 or even 150 mg/dL– is not known with physicians had experience crafting evidence- certainty and may vary based on diagnostic category. Available evidence, however, does based protocols and standardized care was not support maintenance of higher levels of sustained hyperglycemia. accepted. This was important, because TGC presented a level of complexity not present in most other protocols. Medical and nursing Before the project discussed in this arti- gate the relationship between glycemic lev- leadership needed to share the same vision cle, glycemic management in the Stamford els and mortality. The results were striking. A for the unit for this work to be done success- Hospital intensive care unit (ICU)—a 14-bed nearly linear relationship was found between fully. unit where medical, surgical and cardiac these two variables even within the range of Another key point is the decision to patients were treated—was similar to that euglycemia. Further elaboration of these data choose an achievable goal. There was no in most ICUs. Moderate degrees of hypergly- was published in 20032. buy-in from the nursing staff regarding the cemia were tolerated without consideration This evaluation identified the importance idea of TGC until they were comfortable with for treatment, because this condition was of glycemic control and prompted discus- the glycemic target. A final component for common and believed to be adaptive. After sions with the ICU nurses about the type of successful execution of TGC was the avail- the publication of Van den Berghe’s random- treatment protocol that should be instituted ability of robust data-analysis capabilities. ized study of intensive insulin therapy (IIT), a in the unit. It became clear that the nurses Clinicians who practice in the ICU need to program for tight glycemic control (TGC) was were unwilling to implement a protocol with have ready and timely access to glycemic implemented at our hospital. a blood glucose target of 80-110 mg/dL. The values. The treating clinicians need to know The process of making this change began nurses feared the work burden imposed by whether hyperglycemia is being adequately with a review of a database that includes strict monitoring and treatment and the risk controlled and whether the insulin therapy detailed clinical and financial data for 8,600 of hypoglycemia, particularly if such a strict is associated with an unacceptable rate of consecutively admitted patients to investi- target was established. After discussion, it hypoglycemia.

Executive Summary Conference Report 17 7th Invited Conference: Implementation of Tight Glycemic Control at Stamford Hospital

ICU clinicians should also have access to in the intensively treated group of the second control study matched 102 index cases with relevant ICU outcome data. It is not enough to Leuven trial explains why this study was only SH to 306 controls, using APACHE II score, know that the mean glucose level in the unit “equivocally positive,” instead of unequivo- age, diabetic status, diagnostic category and has decreased. It is also important to know cally positive as with their first trial. era of treatment (historic vs. TGC) as matching if TGC had an effect on clinical outcomes, parameters. Mortality was significantly higher Two other recent studies were discontin- including infection rates, severity-adjusted in the 102 patients with SH than in their 306 ued before completion of patient enrollment length of stay and severity-adjusted mortal- matched controls (Table 1). A multi-variate and therefore can be considered failed trials, ity. An analysis of outcomes data from our regression analysis controlling for these same not necessarily refutations of the beneficial database showed a decline in mortality within factors yielded the same conclusion (Table effect of TGC. The VISEP trial was a multi- several months of initiation of the protocol. 1). However, a sensitivity analysis indicated center European trial of TGC and fluid resus- This positive feedback motivated the staff that TGC would still be beneficial even with citation strategies, with a 2X2 design (creat- to continue doing the extra work that TGC a quadrupling of the baseline rate of SH rate ing four separate treatment groups) among required. The nurses knew that the protocol and an associated doubling of the mortality patients with septic shock5. The final results was having a positive effect. percentage attributable to SH. are not available; the study was discontinued Hypoglycemia and the Risk: prematurely because of an excessive rate of Recognition of the relevant risk factors Benefit Ratio of TGC SH in the TGC group (12.1% vs. 2.1%). The is crucial to avoid deleterious SH. The most GLUCONTROL trial was another multi-center important include severity of illness, the pres- In the Van den Berghe surgical ICU study, European study performed in medical and ence of renal failure (delayed clearance of severe hypoglycemia (SH, defined as glucose surgical ICUs6. The results of this study have insulin), sepsis and liver failure (impaired glu- < 40 mg/dL) was seen in 0.8% of the con- also not been published, and it was also termi- coneogenesis), discontinuation of source of ventionally treated group and 5.1% of the nated prematurely because of unacceptable calories with continuation of insulin therapy intensively treated group1. In the Stamford rate of protocol violations and SH. and diabetes. interventional trial, SH represented 0.35% of the values in the historic era and 0.34% of A detailed review of the risk factors for The rate of SH in our ICU was 0.28% among the values in the treatment era2. Both studies the development of SH and the clinical con- patients in the pre-TGC historic era and 0.35% stated that the hypoglycemia that occurred sequences of SH has just been completed among patients treated with the 80-140 mg/ in the treated patients had no adverse clini- in a cohort of 5,365 patients admitted con- dL target. In January 2005, the nurses initi- cal consequences, likely due to its transient secutively to our ICU7. Two separate analyses ated a reduction in the glycemic treatment nature. Later work has refuted this view4,7. confirmed that a single occurrence of SH threshold to 125 mg/dL, because they were conferred increased risk of mortality. A case comfortable with the protocol and wanted The second Leuven trial, published in 2006, was performed in 1,200 medical ICU patients4. The glycemic goal was 80-110 mg/ Table1. A Single Episode of Severe Hypoglycemia (Glucose < 40 mg/dL) dL but the monitoring was performed differ- Increases Risk of Mortality ently than in Van den Berghe’s earlier work. In • Multiple logistic regression analysis the first study all patients had arterial lines, – Model included APACHE II score with age component deleted, age, from which blood was sampled. In the second diabetic status, septic shock, mechanical ventilation, renal insufficiency, study only a minority of patients had arte- treatment era rial lines, and glycemic monitoring was done OR 2.28 (1.41 - 2.70) using a variety of different sources, including p = 0.0008 fingerstick capillary blood. The rate of SH in • Case: control study (1:3) the medical ICU study was 3.1% among the control group and 18.3% among the patients – Matched for APACHE II score with age component deleted, age, in the TGC group, and the investigators noted diagnostic category, treatment era that multivariate analysis indicated that SH Mortality rate: 55.9% vs. 39.5% mitigated the beneficial effect of intensive p = 0.0057 treatment. I believe that the high rate of SH

18 Executive Summary Conference Report 7th Invited Conference: Implementation of Tight Glycemic Control at Stamford Hospital to decrease the threshold. This led to a sig- Future Directions: Advances in Summary nificant increase in the rate of SH (0.73% of Monitoring and Treatment TGC has emerged as a standard of care glucose values) and prompted a review the Current standards of monitoring involve in the ICU because of its biologic plausibil- ICU database leading to the study described the use of a variety of different sources of ity and the available interventional studies. above. A detailed review of the individual blood–arterial, venous and capillary–and dif- Further corroboration may come from the occurrences of SH led to the recognition that ferent measurement systems with different ongoing NICE-SUGAR trial, a multi-center trial when a continuous infusion was not being levels of accuracy. Considerable effort is now being conducted in Australia and Canada used, the use of regular insulin for subcutane- being expended to find a reliable continu- among more than 4,000 medical and surgical ous “correctional” dosing was used. This was ous glucose monitor11. Such a device would ICU patients. The success of the NICE-SUGAR implicated in many of the instances of SH eliminate two important barriers to success- trial will depend in part on whether the cen- because of the potential for “dose stacking” to ful implementation of TGC. The work burden ters can achieve the desired glycemic goals, follow multiple administrations. associated with protocol implementation with clear separation of the interventional In January 2007, regular insulin for sub- would be greatly reduced. The risk of hypo- and control groups and without an unaccept- cutaneous administration was eliminated glycemia would be dramatically reduced, able level of severe hypoglycemia. There may from the protocol, replaced by fast-acting since the nurse would have ready warning never be a true “historic” or “control” groups aspart analogue insulin. With monitoring per- about impending hypoglycemia and be able in any interventional study. It is not known formed every three hours and subcutane- to take corrective action. Another innovation whether the treatment threshold should be ous treatment given every three hours (for is the automated insulin dosing tool. This 110, 125, 140, or even 150 mg/dL. The depth those patients not on continuous infusions device integrates clinical information, recent and breadth of the currently randomized and of insulin), the rate of SH has fallen to 0.13% glucose results and recent history of insulin observational data would not allow an ethical of values, less than half that seen during the dosing to determine a suggested insulin dose Institutional Review Board to allow a control pre-TGC historic era. Careful review of data for each successive glucose value recorded. group to maintain glucose values >200 mg/ allowed staff to recognize an important prob- Our nurses are empowered to make treat- dL without treatment, as was seen in the first lem and track its successful resolution. ment decisions based on a written treat- Van den Berghe trial. Differences Between Patients With ment guideline; however, insulin dosing tools and Without Diabetes might be especially useful in ICUs in which nurses do not have experience with nurse- Another recent publication from our insti- driven protocols. tution has highlighted a different impact of hyperglycemia on mortality of diabetic com- 8 Table 2. Mortality Associated with Mean Glucose Level During ICU Stay pared to non-diabetic ICU patients . Similar findings have been noted in populations of 5,365 patients admitted to the Stamford Hospital ICU between 10/99 and 6/06 patients with acute myocardial infarction9,10. Mortality percentage Odds ratio Hyperglycemia among non-diabetic patients Mean glucose Non- Diabetics Non- Diabetics in our ICU was a much more significant risk during ICU diabetics diabetics factor for mortality than it was among the 70-99 9.2 13.0 0.45 0.45 patients with diabetes (Table 2). There are no data that report non-diabetic hyperglycemic 100-119 11.3 18.9 0.67 NS ICU patients longitudinally to assess what per- 120-139 16.3 17.1 NS NS centage were “latent” diabetics or subsequent- ly became overtly diabetic. Nevertheless, our 140-179 24.2 22.0 1.82 NS experience highlights the importance of tar- 180+ 39.4 25.5 2.35 1.65 geting all patients in the ICU for intensive gly- cemic monitoring and treatment, not just the Odds ratio adjusted for age, severity of illness (APACHE II Score) small minority who are previously identified Krinsley JS. Sem in Thor and Cardiovasc Surg. 2006; 18: 317-25. diabetics.

Executive Summary Conference Report 19 7th Invited Conference: Implementation of Tight Glycemic Control at Stamford Hospital

References 5. Brunkhorst FM, Kuhnt E, Engel C, et al. Intensive 9. Sala J, Masia R, Gonzalez de Molina, et al. Short- insulin therapy in patients with severe sepsis and term mortality of myocardial infarction patients 1. Van den Berghe G, Wouters P, Weekers F, et al. septic shock is associated with an increased rate with diabetes or hyperglycaemia during admission. Intensive insulin therapy in critically ill patients. N of hypoglycemia–results from a randomized mul- J Epidemiol Comm Health 2002; 56:707-12. Engl J Med 2001; 345:1359-67. ticenter study (VISEP). Infection 2005; 33(S1): 19. 10. Capes SE, Hunt D, Malmberg K, Gerstein HC. Stress 2. Krinsley JS. Association between hyperglycemia 6. Glucontrol study. Available online at: hyperglycaemia and increased risk of death after and increased hospital mortality in a heteroge- http://www.glucontrol.org/ myocardial infarction in patients with and without neous population of critically ill patients. Mayo diabetes: a systematic overview. Lancet 2000; 7. Krinsley JS, Grover A. Severe hypoglycemia in Clinic Proc 2003; 78:1471-8. 355:773-8. critically ill patients: Risk factors and outcomes. Crit 3. Krinsley JS. The effect of an intensive glucose man- Care Med. 2007;35:2262-7 . 11. Krinsley JS, Zheng, P, Hall D et al. ICU validation of agement protocol on the mortality of critically ill the OptiScanner, a continuous glucose monitoring 8. Krinsley JS. Glycemic control, diabetic status and adult patients. Mayo Clinic Proc 2004; 79:992-1000. device. Crit Care Med 2006; 34:A67. mortality in a heterogeneous population of criti- 4. Van den Berghe G, Wilmer A, Hermans G, et al. cally ill patients before and during the era of tight Intensive insulin therapy in the medical ICU. N Engl glycemic control. Seminars in Thor and Cardiovasc J Med 2006; 354:449-61. Surg 2006;18:317-25.

20 Executive Summary Conference Report 7th Invited Conference: Meta-analysis of Randomized Trials of Tight Glycemic Control

PROCEEDINGS Meta-analysis of Randomized Trials of Tight Glycemic Control Anastassios G. Pittas, MD, MS, Associate Professor of Medicine, Tufts-New England Medical Center, Boston, MA

Key points The DIGAMI study was the first to provide concrete evidence that controlling glycemia • The results of observational studies show that in-hospital hyperglycemia is an independent in hospitalized patients may improve out- risk factor for adverse outcomes, including mortality. comes. However, it was not until 1991, when • A single-center randomized trial found that tight glycemic control with insulin improves the results of a study by Van den Berghe et mortality and morbidity in the surgical intensive care unit; benefit in other hospital settings al. were published, that hyperglycemia and has not been established by randomized trials. its therapy with insulin in the critically ill 2. Van • Clinicians should aim for a pre-prandial glucose concentration < 150 mg/dL in most became of great interest to clinicians den Berghe et al. conducted a randomized hospitalized patients, with stricter glucose goals in the critically ill patient, especially among controlled trial where patients admitted patients with cardiac disease. to the surgical intensive care unit (ICU), of which 63% had cardiac surgery, received either ‘routine’ therapy or intensive insulin Hyperglycemia is a common occurrence when an infusion of glucose, insulin and therapy (IIT) aiming at tight glycemic con- in the critically ill patient. Historically, in- potassium (GIK) was developed as a potential trol (TGC). The results were remarkable: IIT hospital hyperglycemia, especially of new therapy to improve cardiac-related outcomes resulted in a blood glucose (BG) level of 103 onset, has been seen as an adaptive response following acute myocardial infarction (AMI). mg/dL vs. 153 mg/dL in the control group to heightened medical stress and considered Over the next 30 years multiple small inter- and significantly reduced in-hospital mortal- a marker of illness severity rather than a vention studies with GIK were completed ity and morbidity (see Table in Appendix). distinct medical entity that requires manage- with conflicting results that could be due, at ment. Recent evidence, however, challenges least in part, to the differences in the various Since then, three large studies with insu- that notion. Several observational studies regiments used. lin therapy in hospitalized patients were published with neutral results. have established that in-hospital hyperglyce- Large Intervention Studies mia is an independent risk factor for adverse • The CREATE-ECLA study was an interna- outcomes, including mortality, particularly The first large intervention study with a tional trial of patients with AMI, who were in critically ill patients with known diabetes glucose-insulin infusion aiming at euglyce- randomized upon hospital admission to but also in those with new-onset hyperglyce- mia was the DIGAMI study, which random- receive either GIK infusion for 24 hours mia. Recently, several randomized trials have ized patients with diabetes and AMI to either (without a glucose target) or usual care3. tested the hypothesis that lowering glycemia ‘routine’ therapy or intensive therapy with No difference on mortality or morbidity in hospitalized patients would improve out- glucose-insulin infusion for 48 hours with a was seen with the GIK infusion. comes, but with conflicting results. Here I glucose goal less than 10 mmol/L, followed 1 summarize the evidence from these trials. by subcutaneous insulin therapy . Although • The DIGAMI-2 study tried to distinguish there was no short-term statistically signifi- the short-term from the long-term ben- Early Trials cant benefit with intensive insulin-glucose efits of an insulin-based glucose manage- Experimentally, the concept of altering therapy, at one year the latter group had ment protocol in patients hospitalized for glycemia in hospitalized patients to affect a relative mortality reduction of 26% com- AMI by randomizing patients into three outcomes was first introduced in the 1960s pared to the control group. groups (Table in Appendix)4. There were

Executive Summary Conference Report 21 7th Invited Conference: Meta-analysis of Randomized Trials of Tight Glycemic Control

no differences in mortality or morbidity After combining data from studies that used References

among the three groups. This study was a non-GIK insulin regimen, the relative risk 1. Malmberg K, Ryden L, Efendic S, et al. Randomized underpowered by not meeting enroll- for mortality remained essentially unchanged trial of insulin-glucose infusion followed by sub- cutaneous insulin treatment in diabetic patients ment numbers and did not achieve its (Relative Risk 0.92 [95% CI, 0.74-1.15]). It is with acute myocardial infarction (DIGAMI study): treatment goals. interesting to note that since the positive effects on mortality at 1 year. J Am Coll CardioL Jul results in the surgical ICU reported by Van de 1995;26(1):57-65. The results of the CREATE-ECLA and Berghe et al. in 2001, 14 trials with a variety of 2. Van den Berghe G, Wouters P, Weekers F, et al. DIGAMI-2 trials suggest that insulin ther- Intensive insulin therapy in critically ill patients. N insulin regimens, including GIK, in hospital- apy without targeting euglycemia prob- Engl J Med 2001;345(19):1359-67. ized patients have been published, and none ably has no effect on outcomes. 3. Mehta SR, Yusuf S, Diaz R, et al. Effect of glucose- of them have shown a statistically significant insulin-potassium infusion on mortality in patients • The third study was conducted by Van benefit for insulin therapy. Two randomized with acute ST-segment elevation myocardial infarc- tion: the CREATE-ECLA randomized controlled trial. den Berghe et al. in patients in the medi- trials (VISEP and Glucontrol) whose results J Amer Med Assn Jan 26 2005;293(4):437-46. cal ICU and followed a protocol identical are pending were stopped early because of 4. Malmberg K, Ryden L, Wedel H, et al. Intense meta- to the surgical intensive care study by the frequent hypoglycemia (these are discussed bolic control by means of insulin in patients with same investigators5. Overall, there was no elsewhere in these Proceedings). diabetes mellitus and acute myocardial infarction benefit of IIT reduced BG levels but did (DIGAMI 2): effects on mortality and morbidity. Eur Conclusion Heart J Apr 2005;26(7):650-61. not significantly reduce in-hospital mor- 5. Van den Berghe G, Wilmer A, Hermans G, et al. tality (40% vs. 37% in the conventional There is general agreement that improved Intensive insulin therapy in the medical ICU. N Engl vs intensive group, respectively). Among glycemic control should be an important J Med Feb 2 2006;354(5):449-61.

patients who stayed in the ICU for less component of care in the hospitalized patient. 6. Pittas AG, Siegel RD, Lau J. Insulin therapy for than three days, mortality was greater Although the evidence supports TGC in cardi- critically ill hospitalized patients: a meta-analysis of randomized controlled trials.Arch Intern Med Oct 11 among those receiving intensive therapy. ac patients in the surgical ICU, currently there 2004;164(18):2005-11. In contrast, among patients who stayed in is not enough evidence from randomized the ICU for three or more days, in-hospital trials to recommend the same degree of strict mortality was reduced from 53% to 43% glycemic control in all hospitalized patients. with IIT. Until further evidence becomes available, it is prudent for clinicians to aim for a pre-pran- Updated Review and Meta-analysis of dial blood glucose concentration less than Randomized Trials 150 mg/dL in all hospitalized patients, with My colleagues and I recently conducted stricter blood glucose goals in the critically a systematic review and meta-analysis of ill patient, especially among patients with randomized trials to determine the effect on cardiac disease. mortality of insulin therapy initiated during hospitalization in patients with critical illness defined as AMI, stroke, cardiac surgery or an illness requiring a stay in the ICU6. I updated the search and analyses for this conference.

The search revealed 41 published ran- domized trials (n=32,573 patients) that have employed an insulin regimen, including GIK, and reported data on mortality. Combining results from all 41 trials, there was a trend that insulin therapy decreased short-term mortal- ity (Relative Risk 09.94 [95% CI, 0.85-1.03]).

22 Executive Summary Conference Report 7th Invited Conference: Perioperative Glucose Management in the Operating Room

PROCEEDINGS Perioperative Glucose Management and IIT in the Operating Room Richard C. Prielipp, MD, MBA, FCCM, Professor and Chair of Anesthesiology, University of Minnesota, Minneapolis, MN; Douglas B. Coursin, MD, Professor of Anesthesiology and Medicine, University of Wisconsin School of Medicine and , Madison, WI

operative periods. This variation is captured Key points in a recent web-based poll conducted by • Anesthesia professionals vary in their adoption and use of intensive insulin therapy (IIT) the Anesthesia Patient Safety Foundation and debate the appropriate application of inensive care unit (ICU ) insulin protocols and (apsf.org), when the following question was treatment goals in patients during the pre-, intra-, and postoperative periods. asked: • There are very few data to answer questions such as: “During general anesthesia in the OR, what is − Does tight glycemic control (TGC) during the period of surgery affect long-term patient your current upper limit of glucose that triggers outcomes? (intravenous bolus or infusion) insulin therapy?” − What is the optimal glucose concentration for patients in the OR during periods Results are illustrated in the Figure. The most of stress? common response was that insulin therapy is − What is the danger of hypoglycemia for patients under general anesthesia? initiated during surgery only when patients’ • The original data from Belgium suggest substantial benefit from maintenance of blood glucose is ≥ 200 mg/dL (11.1 mmol/L). This glucose ≤ 110mg/dL in cardiac surgery patients, but the results of subsequent studies are may surprise or even distress some ICU prac- insufficient to mandate this level of TGC for patients in the OR. titioners. Three factors may account for cur- rent practice for glucose management in the Leaders in anesthesiology, intensive care, including cardiac catheterization laborato- OR: , surgery, hospitalist medicine, ries, neuroradiology suites, and endoscopy 1. Lack of definitive data that IIT and TGC , nutrition, nursing, pharma- centers. Anesthesia professionals vary in during surgery (a period typically lasting cy, biostatistics, and biotechnology recognize their adoption and use of these concepts only two to three hours) improves peri- that intensive insulin therapy (IIT) improves and debate the appropriate application of operative outcomes, especially in subsets outcomes in select critically ill intensive care ICU insulin protocols and treatment goals of patients such as ambulatory surgery unit (ICU) patients. There are core questions in patients during the pre-, intra-, and post- patients. that must still be resolved. For example, do ICU patients benefit most from ‘tight’ glyce- Figure. Threshold that Triggers Insulin Therapy mic control (TGC) (usually defined as plasma 35 [blood glucose] in the range 80–110 mg/ 30 dL) or 'slightly less intense’ control (typi- 25 cally translated as blood glucose in the range 20 110–150 mg/dL)? Other questions include: 15

• What glucose concentration is the appro- 10

priate threshold to initiate treatment? 5

• Which IIT treatment algorithm is best (and 0 110 140 180 200 >240 Only if safest)? (Glucose)–mg/dL acidotic

These questions are the same for patients Responses to a June, 2007 APSF poll (www.apsf.org) of anesthesia professionals asked: “During general anesthesia in the in the operating room (OR) and other areas OR, what is your current upper limit of glucose that triggers (intravenous bolus or infusion) insulin therapy?”

Executive Summary Conference Report 23 7th Invited Conference: Perioperative Glucose Management in the Operating Room

2. The uncertainty of the ideal blood glucose Although this was a modest-sized study, mor- not clear that all patients derive these ben- target/goal in the OR (similar to the chal- tality rates were less than half that reported efits. An improvement has been documented lenge and dilemma faced by ICU practitio- in the commonly discussed IIT ICU study by in complex cardiac surgery patients for whom ners). Van den Berghe2. Van den Berghe wrote an an aggressive IIT protocol was used for an accompanying editorial suggesting that TGC extended time after surgery. These results 3. The ill-defined danger of iatrogenic during the “brief” duration of OR care is insuf- have not been reconfirmed, however. hypoglycemia associated with IIT during ficient to affect patient outcomes. Data from anesthesia, when the classic autonomic Factor # 3: What is the danger of hypoglyce- other studies such as the Portland Diabetic and neurological signs and symptoms of mia for patients under general anesthesia? Project4 draw different conclusions. This non- low blood glucose are masked or absent. randomized, prospective, observational study Anesthesia professionals prefer to avoid Implicit with this concern are the medico- of 5,510 cardiac surgery patients found that risk and are sensitive to the possibility of iat- legal consequences of unintended hypo- hyperglycemia in the first three post-opera- rogenic hypoglycemia when using IIT during glycemia during anesthesia. tive days was an independent and robust pre- anesthesia. The usual autonomic and neu- Factor # 1: Does TGC during the period of dictor of mortality, sternal wound infections, rological signs and symptoms of low blood surgery impact long-term patient outcomes? and increased LOS for patients with diabetes. sugar are masked or absent during anes- thesia. Little is known about the frequency, There are very few data to guide glucose So, how do anesthesiology clinicians rec- severity and consequences of intraoperative management for patients who are in the oncile this conflicting information? Until con- hypoglycemia. Hypoglycemia in awake out- OR. Many anesthesia practitioners therefore clusive data are published, some anesthesia patients is defined as < 50 mg/dL in males question whether it is appropriate to apply practitioners suggest that routine glucose and < 40 mg/dL in females. The clinically rel- data from pre-operative, post-operative, and management is sufficient for the OR, and it evant symptoms associated with this degree ICU periods to the relatively brief period of is reasonable to delay TGC until the patient of hypoglycemia are summarized in Table 1. surgery. Others think that data derived from arrives in the ICU. They note the ICU environ- studies of specific ICU populations such as ment is ideally suited to engage multi-disci- Clinicians are naturally cautious when con- those patients suffering MI, cerebral ischemia, plinary teams to implement IIT protocols. ditions exist which blunt the usual responses sepsis, or undergoing cardiac surgery are not to hypoglycemia. Drugs such as anesthet- Until then the lack of definitive evidence- applicable to patients under general anes- ics and beta-blockers, various medical con- based outcome data about intraoperative thesia having routine surgery lasting two to ditions, or autonomic sympathetic hypo-re- glycemic control will result in variable prac- three hours. A recent prospective, random- sponsiveness (termed “hypoglycemia-associ- tices for insulin therapy in the OR. The core ized, open-label, controlled study at the Mayo ated autonomic failure,” HAAF) can blunt the question remains as to whether two to three Clinic of 400 cardiac surgery patients1 was physiologic response to low blood glucose, hours of hyperglycemia in the OR constitutes unable to demonstrate a benefit of IIT during and serious neuroglycopenia can occur in the a critical risk which increases adverse patient surgery. absence of changes in vital signs or observ- outcomes. As this discussion continues, it able neurological symptoms. The Mayo Clinic patients were random- is likely that intra-operative glucose man- ized to tight intraoperative control utilizing agement and IIT will be a component of The concern surrounding iatrogenic hypo- an insulin infusion to maintain blood glucose overall perioperative care, and maybe even a glycemia is one factor limiting routine aggres- between 80–100 mg/dL while in the OR, or benchmark for anesthesiology P4P (pay-for- sive glucose management in the OR. It is likely conventional treatment where they received performance). that the sequelae of hypoglycemia are dura- insulin only when blood glucose exceeded tion (time)–and “dose” (severity)–related. Low Factor # 2: What is the optimal glucose con- 200 mg/dL. All patients received the “standard glucose concentrations are associated with centration for patients in the OR during peri- ICU insulin protocol” and achieved TGC within predictable neurological dysfunction (Table ods of stress? four to six hours of arrival in the cardiac ICU. 2). There is minimal permanent risk from a sin- Outcomes including hospital and ICU length This question parallels the challenge facing gle episode of hypoglycemia (blood glucose of stay (LOS) were identical for both groups ICU practitioners. While it is widely recognized ≤ 40 mg/dL), providing it is diagnosed and of patients. The IIT group had more deaths (4 that patient outcomes improve when glucose managed in a timely fashion. A recent study5 vs. 0; p = 0.06), and more strokes (8 vs. 1; p = is tightly controlled during ICU treatment for of a long-term, outpatient, intensive diabetic- 0.02) than the conventional treatment group. certain subsets of hospitalized patients, it is care algorithm showed that patients experi-

24 Executive Summary Conference Report 7th Invited Conference: Perioperative Glucose Management in the Operating Room

tions for patients who are receiving IIT under • Clinical experience suggests that a single Table 1: Signs and Symptoms of general anesthesia. This is important because episode of hypoglycemia (blood glucose Hypoglycemia in Awake Patients additional variables such as blood sampling ≤ 40 mg/dL) carries minimal risk if diag- Behavior/mood alterations site (venous vs. capillary vs. arterial), anemia, nosed and managed in a timely fashion. – Emotional lability and confounding substrates may lead to spu- The issue of hypoglycemia is still a con- – Irritability rious glucose determinations7. cern for the many anesthesia profession- als. Patients with sepsis exhibit a much Physical Symptoms A Summary of Perioperative Glucose higher risk of hypoglycemia and control of – Diaphoresis Management – Tremor their blood sugar is more difficult. Conclusions about perioperative glucose – Paresthesia • Practitioners in both the OR and ICU management include: – Tachycardia struggle with the question of what is the • The prevalence of type 2 diabetes mellitus Neuroglycopenic Signs and Symptoms necessary, appropriate and “ideal” glucose is increasing rapidly. – Hypothermia target. The original cardiac surgery data – Weakness • TGC requires an interdisciplinary team from Leuven, Belgium suggest substan- – Fatigue approach, a culture of safety and a focus tial benefit from maintenance of blood – Slurred speech on professional education. Benchmarks to glucose ≤ 110mg/dL. Current data appear – Loss of consciousness (LOC) evaluate effectiveness are needed. insufficient to mandate this level of TGC for – Hemiparesis patients in the OR. A recent randomized • It is important to note that perioperative – Seizures study in cardiac surgery patients found no hyperglycemia occurring in “non-diabet- – Brain damage difference in ICU or hospital LOS despite ics” may actually indicate undiagnosed TGC throughout the operative period. type 2 diabetes that may result in increased enced frequent hypoglycemic episodes. The morbidity and mortality. Providers should References extensive battery of neuropsychiatric tests, consider hemoglobin A1c determinations 1. Gandhi GY, Nuttall GA, Abel MA, et al. Intensive however, found no difference in neurologi- intraoperative insulin therapy versus conventional in these patients to direct optimal meta- cal deficits between intensive and routine glucose management during cardiac surgery: A bolic management and potentially alter randomized trial. Ann Intern Med 2007;146:233-43. diabetic care for study patients. In nearly two the timing of procedural intervention, par- decades, no adverse neurological sequelae 2. Van den Berghe G, et al. Intensive insulin therapy ticularly for elective such as joint in critically ill patients. N Engl J Med 2001; 345: could be correlated with hypoglycemic epi- 1359-67. replacement, spine surgery or bariatric sodes. procedures. 3. Van den Berghe G. Does tight glycemic control dur- ing cardiac surgery improve patient outcome? Ann More data are needed to define the • It is now being recognized that insulin is Intern Med 2007;146:307-8. Editorial. minimal duration and severity of hypogly- appropriate therapy for all acute stress 4. Furnary AP, Wu Y. Clinical effects of hyperglycemia cemia that constitutes “neurological risk” for in the cardiac surgery population: the Portland and perioperative hyperglycemia. The patients in the OR (where brain temperature Diabetic Project. Endocr Pract. 2006; 12 Suppl treatment of patients who do not have 3:22-6. may be an important co-variate, for example). diabetes but become hyperglycemic may This information will define the optimal time 5. Jacobson AM, Musen, G, Ryan CM, et al. Long-term achieve the greatest benefit with appro- effects of diabetes and its treatment on cognitive increment between blood sugar determina- function. New Engl J Med 2007;356:1842-52. priate treatment. 6. Kroll HR, Maher TR. Significant hypoglycemia sec- ondary to icodextrin peritoneal dialysate in a dia- Table 2: Neurological Consequences of Hypoglycemia betic patient. Anesth Analg 2007;104:1473-4.

Blood sugar below 45 mg/dL = neuroglycopenia 7. Marik PE, Raghavan M. Stress-hyperglycemia, insu- – Altered mentation, eventually leading to seizures, lin and immunomodulation in sepsis. Intensive Care Med 2004; 30:748-56. unconsciousness and coma

Below 36 mg/dL = EEG changes – EEG changes persist even after restoration of plasma sugar

Below 18 mg/dL = neuronal necrosis likely

Executive Summary Conference Report 25 7th Invited Conference: Economic Advantages of Tight Glycemic Control

PROCEEDINGS Economic Advantages of Tight Glycemic Control Judith Jacobi, PharmD, FCCM, FCCP, BCPS, Critical Care Pharmacist, Methodist Hospital/Clarian Health, Indianapolis, IN

and the need for prolonged ventilation is an Key points important contributor to more rapid ICU dis- • Measures of importance in evaluating the economic impact of tight glycemic control (TGC) charge. The need for MV has a large impact on include clinical outcomes, mortality and morbidity, reduction in length of stay, equipment ICU costs, as demonstrated by Dasta and col- utilization and personnel time. leagues5. Mean intensive care costs and LOS • Studies have shown that achieving TGC reduces intensive care unit-related complications were higher than for those not requiring MV. and length of stay compared with conventional glycemic management that produces Daily costs were greatest on day 1 ($10,794 higher mean glucose values of 150-170 mg/dL. versus $6,667 for MV vs. no MV, respectively) and decreased daily, becoming stable after • The degree of cost saving may vary with other patient populations and with other glycemic day 3 ($3,968 vs. $3,184, respectively). The management programs, depending on the baseline complication rate and the effectiveness mean incremental cost of MV was $1,522 per and safety of the TGC protocol; however, outcome and economic benefits of TGC appear to day, although costs appear to be higher in be significant and positive. surgical and trauma patients than in medical ICU patients.

Evaluation of the economic impact of any mortality reduction of at least 3% to 4% in all Infectious complications are another intervention in the critical care unit is a com- patients and above 7% in surgical intensive important contributor to hospital costs. The plex proposition. Many overhead expenses care (SICU) patients and medical ICU (MICU) importance of glucose control has been are fixed in a technology-rich environment. patients who remained in the ICU for more shown by the Portland Diabetic Project. However, there have been several reports than 3 days3. A combined dataset from these Reductions in the three-day post-operative examining the potential impact of tight gly- two trials demonstrated consistent benefits average blood glucose (3BG) have been asso- cemic control (TGC) in critical care. in all subsets with the exception of patients ciated with progressive reductions in deep with diabetes who were previously treat- sternal wound infection (DSWI) and mortal- The economic measures of importance 6 ed with insulin, where a tendency toward ity . The risk of post-cardiac bypass mortality include clinical outcomes, mortality and mor- increased risk of death was suggested. While is doubled for each 50mg/dL increase in 3BG. bidity, length of stay, equipment utilization the benefit on mortality may be progressive The reported 3.2% reduction in DSWI trans- and personnel time. A reduction in mortal- as the glucose is lowered to normal ranges lates into a number needed to treat (NNT) of ity is the most important clinical outcome; of 80-110 mg/dL, achieving normoglycemia 31 diabetic patients to prevent 1 DSWI. The however, it can lead to additional expenses 7 appears essential to prevent renal failure and cost of a DSWI was reported to be $81,000 . if associated with a longer intensive care critical-illness polyneuropathy4. The average cost per day in open heart sur- unit (ICU) length of stay (LOS) or additional gery patients has been reported to be as high therapeutic interventions. The large trials of Specific morbidity measures have great 8 as $1,150 . In summary, actual cost saving per tight TGC have also demonstrated significant impact on cost analysis. TGC reduced ICU patient is $2,613 and savings from reduced reductions in morbidity1,2. and hospital LOS and, importantly, reduced length of stay is $3,150 per day, producing a the percentage of patients requiring more The two large clinical trials from Leuven, net saving of $5,580 ($138-$5,718) per patient 1 than two weeks of ICU care . Reduction of 6 Belgium demonstrated significant absolute treated with the Portland Protocol . the duration of mechanical ventilation (MV)

26 Executive Summary Conference Report 7th Invited Conference: Economic Advantages of Tight Glycemic Control

Cost savings are also associated with pre- economic impact of a therapeutic interven- ferences in total cost per patient was 7,931€ vention of other complications such as blood tion is essential. The Leuven SICU study pro- (6,746-9,031€) for the TGC group vs. 10,569€ stream infections (BSI) and renal failure. The duced significant reductions in mortality and (9,214-11,441€) for the conventional treat- Leuven SICU study with TGC demonstrated morbidity related to reductions in septicemia, ment group. This translates into a savings of significant reductions in ICU septicemia and renal failure, red-cell transfusions, need for 2638€ (183-4,695€) per patient or approxi- the number of patients who are treated with drug therapy (pressors, antibiotics, inotropes) mately $3,429. more than 10 days of antibiotics1. The attribut- and development of critical illness (polyneu- A similar analysis was performed in a com- able costs of an ICU-acquired BSI are substan- ropathy). The comprehensive cost analysis munity-based medical-surgical ICU based on tial and vary from approximately $9,400 to included bed, therapy and monitoring costs the results of a before-and-after comparison $18,000 (Table 1). The impact from avoidance (Table 2)16. The costs for insulin therapy and of cohorts17. The historical cohort did not have a focused glucose management pro- Table 1. Attributable Costs Reported Per Acquired Blood Stream Infection gram, while the after cohort was managed to Country Cost Approx. US Cost achieve blood glucose of 80-140 mg/dL. The glucose management protocol resulted in a Canada9 $7,885 per case significant reduction in outcome measures, $16,099 per survivor including a 13.9% reduction in ICU days and Italy10 €16,356 per case $17,694 per survivor a 34.3% relative reduction in the duration of €13,611 per survivor MV. These positive results produced a net cost savings per patient of $2,311 or an adjusted Belgium11 € 13,585 per case $17,661 per case saving of $1,580 when correcting for differ- ences in ventilation at baseline. These savings USA12 $11,971 per case considered hospital costs, imaging, pharmacy, Missouri Severity adjusted laboratory, and the higher costs for intensive USA13 $40,179 per case insulin therapy, but do not account for every Pittsburgh $9,419-$170,565 ICU cost. An annualized, adjusted total cost Ratio € to US of 1:1.3 applied, if not reported in the paper savings was predicted to be $1,339,500 with the application of TGC in a 14-bed unit. of renal failure and the need for renal replace- monitoring were 72€ (approx. US $93.60) ment therapy is also potentially important. higher in the TGC group. However, the dif- The Leuven group reported that maintaining TGC (80-110 mg/dL) was essential to prevent Table 2. Expenses Associated With Tight Glycemic Control16 renal impairment and produced an overall 42% risk reduction (p=0.0009) versus conven- Variable Cost, € Cost Approx. US$ tional glucose control4. The cost of continuous veno-venous hemofiltration (CVVH) has been ICU length of stay (per day) 1030.00 1339 reported to be approximately $390 per 24 Mechanical ventilation (per day) 40.80 53.04 9 hours ($292 to $488) . Hemodialysis (per day) 386.00 501.80 A potential drawback to the use of TGC is IV tubing (changed daily) 4.77 6.20 the impact on nursing workload. This is an IV pump (per day) 4.75 6.18 important component of analysis, consider- ing the fixed number of critical care nurses. A 0.9% NaCl for in injection 1.20 1.56 detailed discussion of the impact on workload Regular Human Insulin (per unit) 0.03 0.04 15 is available from Aragon in this summary . Whole blood glucose measure 0.90 1.17

While avoiding individual complications US = 1.3 x €, if not reported in the paper is important, an overall assessment of the

Executive Summary Conference Report 27 7th Invited Conference: Economic Advantages of Tight Glycemic Control

Conclusion References 9. Laupland KB, Lee H, Gregson DB, Manns BJ. Cost of intensive care unit-acquired bloodstream infec- Achieving TGC has been shown to reduce 1. Van den Berghe G, Wouters P, Weekers F, et al. tions. J Hosp Infect 2006;63:124-32. Intensive insulin therapy in critically ill patients. N ICU-related complications and LOS compared Engl J Med 2001;345:1359-67. 10. Orsi GB, DiStefano L, Noah N. Hospital-acquired, with conventional glycemic management laboratory-confirmed bloodstream infection: 2. Van den Berghe G, Wilmer A, Hermans G, et al. increased hospital stay and direct costs. Inf Control that produces higher mean glucose values Intensive insulin therapy in medical ICU. N Engl J Hosp Epidemiol 2002;23:190-7. Med 2006;354:449-61. of 150-170 mg/dL. The degree of cost sav- 11. Blot SI, Depuydt P, Annemans L, et al. Clinical and ing may vary with other patient populations 3. Vanhorebeek I, Langouche L, Van den Berghe G. economic outcomes in critically ill patients with Tight blood glucose control with insulin in the ICU. and other glycemic management programs, nosocomial catheter-related bloodstream infec- Facts and controversies. Chest 2007;132:268-78. tions. Clin Infect Dis 2005;41:1591-8. depending on the baseline complication rate 4. Van den Berghe G, Wilmer A, Milants I, et al. 12. Warren DK, Quadir WW, Hollenbeak CS, et and the effectiveness and safety of the TGC Intensive insulin therapy in mixed medical/surgical al. Attributable cost of catheter-associated protocol, but outcome and economic ben- intensive care units. Benefit versus harm. Diabetes bloodstream infections among intensive care 2006;55:3151-9. efits of TGC appear to be significant and posi- patients in a nonteaching hospital. Crit Care Med 2006;34:2084-9. tive. 5. Dasta JF, McLaughlin TP, Mody SH, Piech KT. Daily cost of an intensive care unit day: the contri- 13. Pitts Shannon RP, Patel B, Cummins D, et al. bution of mechanical ventilation. Crit Care Med Economics of central-line associated bloodstream 2005;33:1266-71. infections. Am J Med Qual 2006;21(Suppl):7s-16s.

6. Furnary AP, Wu X, Bookin SO. Effect of hypergly- 14. Klarenbach SW, Pannu N, Tonelli MA, Manns BJ. cemia and continuous insulin infusions on out- Cost-effectiveness of hemofiltration to prevent comes of cardiac surgical procedures: The Portland contrast nephropathy in patients with chronic kid- Diabetic Project. Endocr Pract 2004;10(Suppl ney disease. Crit Care Med 2006;34:1044-51. 2):21-3. 15. Aragon D, Evaluation of nursing work effort and 7. Cimochowski GE, Harostock MD, Brown R, et al. perceptions about blood glucose testing in tight Intranasal mupirocin reduces sternal wound infec- Glycemic control. Am J Crit Care 2006;15:370-7. tion after open heart surgery in diabetics and nondiabetics. Ann Thorac Surg 2001;71:1572-9. 16. Van den Berghe G, Wouters PJ, Kesteloot K, et al. Analysis of healthcare resource utilization with 8. Furnary AP, Gao G, Grunkemeier GL, et al. Continuous intensive insulin therapy in critically ill patients. Crit infusion insulin reduces mortality in patients with Care Med 2006;34:612-6. diabetes undergoing coronary artery bypass graft- ing. J Thorac Cardiovasc Surg 2003;125:1007-21. 17. Krinsley JS, Jones RL. Cost analysis of intensive Glycemic control in critically ill adult patients. Chest 2006;129:644-50.

28 Executive Summary Conference Report 7th Invited Conference: Intensive Insulin Therapy in the Intensive Care Unit

PROCEEDINGS Intensive Insulin Therapy in the Intensive Care Unit Robert Osburne, MD, Endocrinology, Atlanta Medical Center, Atlanta, GA

Key points Georgia Hospital Association, saw tight glyce- mic control (TGC) as an issue involving quality • New approaches, algorithms and tools are necessary to achieve tighter glycemic targets of care and patient safety that confronted a without substantially increasing the risk of harmful hypoglycemic events. large number of Georgia Hospitals. In 2003, • Use of a protocol at Atlanta Medical Center (AMC) led to better overall blood glucose the Diabetes Special Interest Group* (DSIG) control in the intensive care unit, hypoglycemia episodes less than 1% of blood glucose was organized as a part of the PHA, parallel- measurements and infrequent episodes of sustained hypoglycemia. ing similar efforts in other parts of the coun- 11. Nurses at the Atlanta Medical Center • The average blood glucose level declined for all hyperglycemic patients whether or not the try (AMC) ICU, a 50-bed, open ICU with medical, ordered constant insulin infusion per the protocol. surgical, coronary, trauma and open-heart • AMC experience showed that: surgery patients, were concerned about the • Tight glycemic control (TGC) initiatives must be nurse-driven with the support of a variability in average daily blood glucose physician champion and hospital administration. levels and frustrated by the wide variety of approaches to treatment. A multidisciplinary • Successful implementation of TGC requires much more than just a good insulin-infusion Glucose Control Team (GCT) was formed and protocol. concluded that a standardized approach was • Minimizing prolonged hypoglycemia is imperative. needed to reduce confusion among the nurs- es who administered glucose control orders. • Measuring the impact of new approaches on average blood glucose measurements and GCT nurses participated in the DSIG and outcomes is important. decided to pilot test the group’s work prod- • Computerized systems to automate calculation and documentation are needed to uct as an evidence-based approach to TGC. reduce nurse workload and facilitate compliance. Developing and Implementing the DISG • A successful program requires a change in thinking at many levels within the Protocol for TGC organization and takes longer than expected. For more than two years, the DSIG held monthly meetings that involved well over a 9-10 Although there is still disagreement about in the medical ICU . Regardless of the target hundred different people representing at least the appropriate target for glycemic control, range finally agreed upon, new approaches, 50 hospitals and other organizations inter- current evidence suggests that for patients algorithms and tools are necessary to achieve ested in improving inpatient glycemic control in the intensive care unit (ICU) peak blood tighter glycemic targets without substantially in Georgia hospitals. The group reviewed glucose should be lower than traditional- increasing the risk of harmful hypoglycemic 12 protocols, which showed wide variability ly allowed. Studies from Van den Berghe1-3, events. in insulin-dose recommendations12. Several Furnary4-5, and Krinsley6-8 suggest the target Diabetes Special Interest Group years’ work by the DSIG produced the colum- may be as low as 80-110 mg/dL. Others are studying whether blood glucose levels as The Partnership for Health and * The history and work of the DSIG can be viewed on Accountability (PHA), supported by the high as 180 mg/dL are acceptable, especially the PHA/DSIG webpage at http://diabetes.gha.org.

Executive Summary Conference Report 29 7th Invited Conference: Intensive Insulin Therapy in the Intensive Care Unit nar insulin dosing chart shown in Figure 1. This method requires hourly fingerstick The nurse is responsible for implementing The decision rules were first published by blood glucose tests until the patient stabi- the treatment plan. Every attempt should be Steed, Bode and Davidson13-14 describing an lizes. When this occurs the interval between made to make a treatment algorithm as user- early computerized version that became the blood glucose checks can be lengthened to friendly as possible. Nurses should lead this Glucommander®. two hours. The chart includes recommenda- effort or at least be integrally involved. tions for correction of “below target” blood The chart may appear more complicated Several steps were important to the suc- glucose levels using 50% dextrose. The lower than a calculation method, but most nurs- cessful IIT implementation at AMC (Table). target is set at 80 mg/dL, well above the es can use it after an hour of instruction. To reduce the confusion with the many AMC “hypoglycemic threshold” of 60 mg/dL. The numbered columns in the chart rep- approaches, the interdisciplinary team chose Sustained hypoglycemia is rare, because of resent a series of parallel correction scales a single approach from those available. After the frequent blood glucose checks, drip-rate with different conversion factors. The most the protocol was determined to be suitable, adjustments as needed and below-target cor- highly insulin-sensitive patients will be con- nurses were trained and a pilot test was con- rections with 50% dextrose. trolled using the information in column one. ducted. This allowed staff to detect unantici- Recommendations for the management of While an effective algorithm, chart or com- pated barriers before proceeding with general patients with progressive degrees of insu- puter program is necessary, more is needed implementation. Following a successful pilot, lin resistance are represented in the higher- to safely achieve the TGC target currently the protocol was implemented as standard numbered columns to the right. The DSIG believed necessary to reduce hyperglyce- policy through the normal medical staff poli- web-page has a second chart for more resis- mia-related complications. Few treatment cy-making system. The protocol then was well tant patients (typically ICU patients who are schemes are as complex as an insulin-dosing publicized within the institution. Major factors have an infection, are obese and/or are being regimen for TGC with minimum hypogly- for success are education, education and edu- given glucocorticoids). The left column and cemia. This complex regimen must be inte- cation. the color codes provide decision rules that grated into the workload of a modern, very Results change with the blood glucose level. busy ICU in which multiple treatments are given and multiple parameters monitored. After a successful AMC pilot test, full imple- mentation of the protocol led to better overall blood glucose control in the ICU, although Figure 1. PHA/DSIG Columnar Insulin Dosing Chart* for a variety of reasons the target range is not always reached during the first 24 hours. The number of hypoglycemia episodes as a percent of blood glucose measurements of patients receiving insulin infusions is signifi- cantly less than 1% and sustained hypoglyce- mia is rare.

Our experience has been that when staff focused on blood glucose control, the average glucose level declined for all hyperglycemic patients, regardless of whether or not the phy- sician ordered constant insulin infusion using the “Columnar Dosing Chart” (Figure 3). This is a good demonstration of the “Hawthorne Effect” observed by social scientists in the 1930s (performance improves when workers know it is being measured)16. The mean blood glucose in the constant insulin infusion group

Copies of this chart can be downloaded at no charge at from www.diabetes.gha.org. is higher than the blood glucose targets for several reasons: 1) a delay in instituting the

30 Executive Summary Conference Report 7th Invited Conference: Intensive Insulin Therapy in the Intensive Care Unit

Any attempt to achieve TGC requires an Figure 2. Blood Glucose Measurements Post-implementation of Standardized investment of time and resources. While pub- Protocol for TGC lished studies demonstrate a return on invest- 77 Patients Columnar Drip Chart Sugars – Post Implementation ment based on shorter stays, reduced infec- 300 tions, and reduced mortality, the debate still surrounding TGC creates doubt in the mind of administrators asked to fund the cost. TGC 250 requires more frequent blood glucose checks and hence more bedside glucose monitors 200 and increased use of related consumable sup- plies. Blood glucose checks, algorithm-related

150 nurse decisions and nurse documentation will add 30 to 45 minutes per shift of nurse work Glucose (mg/100mL) time. Our hospital is like many others in its lack 100 of sophisticated outcomes-tracking systems. This makes it difficult to observe and docu-

50 ment the benefits of TGC. The increased costs 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 are very noticeable and will result in questions Hour from Drip Start and requests for justification from the finance (Osburne, unpublished data) staff in the hospital. infusion, 2) premature termination of the infu- • Despite considerable, consistent effort on The requirement for hourly fingerstick sion when glycemic targets were reached the part of the GCT, acceptance comes blood glucose determinations is one of the (while the patient still had fluctuating insulin slowly. major hurdles to nurse acceptance of safe TGC sensitivity because of other active problems), efforts. Frequent blood glucose determina- • Constant reeducation is needed. and 3) poor transition to subcutaneous insulin tions minimize the risk of sustained hypo- once a decision was made to discontinue con- • A dedicated, nurse-expert GCT is impor- glycemia. Transient blood glucose decreases stant insulin infusion. tant to sustaining the effort required to below 60, 50, or even 40 mg/dL rarely cause change. long-term harm, although they are likely to Lessons Learned be uncomfortable for the patient. Sustained Implementation of TGC at AMC was suc- cessful and educational. The following lessons Figure 3. Raw Average Blood Glucose for Entire ICU Stay were learned.

• Nurses must be the principal drivers since 200 the work required falls most heavily on their shoulders.

• A physician champion is important for sup- 150 port and representation to the physician staff. Key physicians in the ICU must be recruited early and educated to reduce the 100

effects of fear of hypoglycemia, increase Glucose (mg/100mL) understanding of the benefits of TGC and

minimize the effects of clinical inertia. Hawthorne Eect 50 All ICU Patients ICU Patients

(Osburne, unpublished data)

Executive Summary Conference Report 31 7th Invited Conference: Intensive Insulin Therapy in the Intensive Care Unit hypoglycemia is to be avoided because of Conclusions 5. Krinsley JS. Association Between Hyperglycemia and Increased Hospital Mortality in a Heterogeneous the risk of central nervous system damage Based on the experience at AMC, the fol- Population of Critically Ill Patients. Mayo Clinic or even death. A low incidence of sustained Proceedings 2003;78:1471-8. lowing key factors affect implementation of hypoglycemia with the use of TGC treatment TGC in the ICU: 6. Krinsley JS. Effect of an Intensive Glucose strategies builds confidence and hence physi- Management Protocol on the Mortality of Critically Ill Adult Patients. Mayo Clinic Proceedings cians’ and nurses’ support. Intermittent feed- • TGC initiatives must be nurse-driven with 2004;79:992-1000. ing such as meals or bolus tube feeding com- the support of a physician champion and 7. Krinsley JS. Glycemic control, diabetic status, and promises the impact of currently published hospital administration. mortality in a heterogeneous population of criti- algorithms, and these feedings should be cally ill patients before and during the era of • Successful implementation of TGC requires intensive glycemic management: six and one-half avoided for patients requiring insulin infusion much more than just published articles years experience at a university-affiliated commu- for TGC. and a good insulin-infusion protocol. nity hospital. Semin Thorac Cardiovasc Surg. 2006 Winter;18(4):317-25. Transition to subcutaneous (SC) insulin. A • Minimizing prolonged hypoglycemia is 8. Finney SJ. Glucose control in the critically ill-still so successful inpatient TGC plan requires a good imperative. many questions. J Crit Care. 2007 Jun;22(2):118-9. system for the transition to an effective basal- Epub 2007 Jan 31. bolus SC insulin therapy regimen. Transition • Measuring the impact of new approaches 9. Devos P, Preiser JC. Current controversies around from the insulin infusion algorithm is as least on average blood glucose measurements tight glucose control in critically ill patients. Curr and outcomes is important. Opin Clin Nutr Metab Care. 2007 Mar;10(2):206-9. as difficult as execution of the insulin infusion Review. algorithm, and there is not much published • Computerized systems that automate cal- 10. Cook CB, Stockton L, Baird M, Osburne RC, Davidson information about this. Transitioning patients culation and documentation are needed PC, Steed RD, Bode BW, Reid J, McGowan KA. is one reason why the raw mean blood glu- Working to improve care of hospital hyperglyce- to reduce nurse workload and facilitate mia through statewide collaboration: the Georgia cose level in the AMC ICU is not as close to compliance. Hospital Association Diabetes Special Interest target as desirable. Group. Endocr Pract. 2007 Jan-Feb;13(1):45-50. • A successful program requires a change in 11. Wilson M, Weinreb J, Hoo GW. Intensive insulin While there are multiple published insu- thinking at many levels within the organi- therapy in critical care: a review of 12 protocols. lin infusion protocols and at least two FDA- zation and takes longer than expected. Diabetes Care. 2007 Apr;30(4):1005-11. Epub 2007 approved computerized systems, there is no Jan 9. Review. Davidson PC, Steed RD, Bode BW, References Sivitz WI. Computer-controlled intravenous insulin published study of head-to-head compari- infusion using intermittent bedside glucose moni- sons of various protocols in a clinical setting. 1. Van den Berghe G, Wouters P, Weekers F, et al. toring: one year’s experience [abstract]. Diabetes. Intensive insulin therapy in critically ill patients. N 1986;35(Suppl 1):126. There is little published experience using Engl J Med 2001;19:1359-67. protocols for transition from IV insulin infu- 12. Steed RD, Davidson PC, Bode BW, et al. Computer- 2. Van den Berghe G, Wouters PJ, Bouillon R, et al. controlled intravenous insulin infusion using inter- sion to basal-bolus SC regimens. Basal-bolus Outcome benefit of intensive insulin therapy in the mittent bedside glucose monitoring: one year’s regimens should be used in patients able to critically ill: insulin dose versus glycemic control. experience (Abstract). Diabetes 35 (Suppl. 1):32A, Crit Care Med 2003 August;31(2): 359-66. 1986. take oral feeding. Finally, debate continues on 2. Van den Berghe G, Wilmer A, Hermans G, et al. the appropriate target ranges for TGC should 13. Davidson PC, Steed RD, and Bode BW. Intensive insulin therapy in the medical ICU. N Engl Glucommander: A computer-directed intrave- be outside of the cardiac surgery ICU setting. J Med 2006 Feb 2;354(5):449-61. nous insulin system shown to be safe, simple, and effective in 120,618 h of operation. Diabetes Care 3. Furnary, A P et al. Continuous intravenous insulin 2005;28:2418-23. infusion reduces the incidence of deep sternal wound infection in diabetic patients after cardiac 14. Osburne RC, Cook CB, Stockton L, et al. Improving surgical procedures Ann Thorac Surg 1999;67:352- hyperglycemia management in the intensive care 60. unit: preliminary report of a nurse-driven qual- ity improvement project using a redesigned insu- 4. Furnary, A P et al. (2003) Continuous insulin infu- lin infusion algorithm. Diabetes Educ. 2006 May- sion reduces mortality in patients with diabe- Jun;32(3):394-403. tes undergoing coronary artery bypass grafting J Thorac Cardiovasc Surg 2003;125:1007-21. 15. Parsons HM. What Happened at Hawthorne? Science 1974 8 March;183(4128);922-32.

32 Executive Summary Conference Report 7th Invited Conference: Use of a Computerized Algorithm in Patients Undergoing Cardiovascular Surgery: A Protocol for Tight Glycemic Control

PROCEEDINGS Use of a Computerized Algorithm in Patients Undergoing Cardiovascular Surgery: A Protocol for Tight Glycemic Control Bruce W. Bode, MD, FACE, Atlanta Diabetes Associates, Atlanta, GA; Member of the Diabetes Special Interest Group of the Georgia Hospital Association

Key points with diabetes who had pre-existing glycemia based on an A1c > 6%. Discharge planning • Hospital protocols to control glucose must be simple and user-friendly. should be done for all patients with diabe- • These protocols should allow patients with hyperglycemia to be treated the same, allowing tes who have not received insulin before, modification and refinement of the protocol only when necessary. including those who are newly diagnosed.

• There are two effective methods for normalizing glucose for hospitalized patients: Case-specific recommendations are made for continuous, variable-rate intravenous (IV) insulin infusions and basal bolus subcutaneous at home. (SC) insulin therapy Methodology

• Computerized algorithms for IV insulin infusion and transition to SC insulin therapy enable Based on the above goals, Piedmont standardization that can minimize medical errors made in insulin dosing. Hospital in Atlanta decided to implement a TGC protocol in CV patients undergoing • The use of a computerized protocol for cardiovascular surgery patients resulted in bypass and valve surgery. This decision was near-normal glycemia (mean blood glucose 107 mg/dL) with no blood glucose < 40 mg/dL. made after analysis of data from the preced- ing six months that revealed that only 50% to Introduction Diabetes Association (ADA). Both organi- 60% of CV patients had met the Institute for zations recommend a blood glucose level Studies have shown the benefits and risks Healthcare Improvemement’s goal of having < 110 mg/dL in the ICU. AACE recommends of tight glycemic control (TGC) in hospitalized all blood glucose levels be < 200 mg/dL in the < 110 mg/dL premeal with peak postprandial 7 patients, including the intensive care unit 24 hours post-surgery . To achieve this goal, < 180 mg/dL, while the ADA recommends (ICU)1-4. TGC methodologies can be difficult a team of endocrinologists, nurses, pharma- 90-130 mg/dL premeal6. to implement because they often require cists and educators met with a CV surgeon complex, insulin-dosing formulas that are To accomplish these goals while minimiz- every week for three months to develop the restricted to the ICU, and highly skilled nurses ing the risk of hypoglycemia, protocols can protocol. All parties agreed to the following: to work with only one or two patients at a be developed and implemented to screen 1. The protocol would be used for all patients, time. Severe hypoglycemia also can be a patients at high risk for hyperglycemia and so that all patients would be treated the problem with many intravenous (IV) insulin to modify and initiate insulin therapy when same. protocols and has been shown to be an needed. If a patient is acutely ill or unable 2. Since CV patients are at high risk for hyper- independent risk factor for mortality5. In this to eat, a continuous, variable-rate IV insulin glycemia, upon arrival at the hospital all article, the methodology, results and benefits infusion based on hourly or more frequent patients would be screened for both A1c of a computerized protocol in cardiovascular glucose measurements should be used. If a and glucose. (CV) patients undergoing bypass or valve patient is stable and able to eat, subcutane- surgery are discussed. ous (SQ) basal bolus therapy can be used, 3. If a patient were admitted > 24 hours with blood glucose measured premeal, at before surgery, a weight-based basal bolus The recommended goals for glyce- bedtime and at 0300. Transition from IV to SC SQ insulin would be initiated if the A1c mic control in the hospital vary slightly basal bolus insulin therapy should be done were > 6% or the premeal blood glucose between the American Association of Clinical with known or newly diagnosed patients was > 140 mg/dL. The weight-based for- Endocrinologist (AACE) and the American

Executive Summary Conference Report 33 7th Invited Conference: Use of a Computerized Algorithm in Patients Undergoing Cardiovascular Surgery: A Protocol for Tight Glycemic Control

mula used in the protocol is weight in patient is stable, with RAI given premeal Results: Kg times 0.5 equals total daily dose (Kg x the following morning and stopping of the The above protocol was initiated in January 0.5 = TDD), with 50% of the basal being IV insulin and IVFs post-breakfast. The TDD 2006 with a one-month pilot followed by use given as glargine insulin at bedtime and was calculated by using the multiplier on for all patients in February 2006. As of October the remaining 50% of insulin divided by the Glucommander® at 11 p.m. the night 2006, more than 1,800 patients have been 3 and given as a pre-meal dose of rapid- of transition, with the glargine dose being treated using this protocol. Data from the first acting insulin (RAI). Bedside capillary blood 500 times the multiplier, the RAI dose 470 patients were analyzed, producing the glucose would be measured premeal, at being the same dose divided by 3 and following results. bedtime and 0300, with a correction dose given in proportion to the food consumed, of RAI given for a blood glucose > 140 mg/ and the correction dose given as listed 1. 28% of patients had pre-existing diabetes dL. The correction formula would be blood in the weight-based formula. Insulin SQ with 10% of all patients being on insulin glucose minus 100, divided by correction doses would be adjusted by 20% if blood pre-admission; 48% of patients had an A1c factor as determined by the 1700 rule, glucose readings were outside the 70-140 > 6% with 52% having one or more of these which is 1700 divided by the TDD ([blood mg/dL range. Diet therapy would be ini- three criteria. glucose – 100] / [1700/TDD]). tially 1800 Kcal ADA diet in all patients with 2. 96% of patients needed IV insulin peri- 4. Within the the 24-hour period pre-sur- adjustment based on nutrition consulta- and post-operatively with the time to gery through the peri-operative period, tion. < 120 mg/dL being 3 hours. IV insulin was computerized IV insulin delivery using 7. Discharge planning and diabetes educa- continued for a mean of 37 hours. Mean Glucommander® would be initiated for tion would be provided for all patients blood glucose was 107 mg/dL with 2% of anyone with one blood glucose > 140 who have not self-administered SQ insulin patients having a transient blood glucose mg/dL or two blood glucose > 110 mg/ and patients with a change in their insulin < 50 mg/dL and no patient having a blood dL8. Targets for glucose control would be regimen. All patients with A1c > 7% would glucose < 40 mg/dL (Figure 1). 90-120 mg/dL with blood glucose mea- be instructed to remain on their basal 3. 98% of patients have been controlled with surement in response to a Glucommander® bolus insulin therapy for at least 4 to 6 no blood glucose > 200 mg/dL within the alert (average, hourly; range, every 20 to weeks or longer, until advised otherwise first 48 hours postoperatively. 120 minutes based on the stability of the by their primary care practitioner. 4. 55% of patients were transitioned to basal- glucose). The lower alert target of 90 mg/ 8. An endocrinologist would be consulted bolus therapy whereby > 90% of the blood dL was selected to minimize the risk of for any patient with a blood glucose < 70 glucose values remained in the target range hypoglycemia. Maintenance intravenous or > 140 mg/dL or a patient who had not (70-140 mg/dL) with no blood glucose < 40 fluids (IVFs) are administered using D10W received insulin before. mg/dL (Figure2). at 50mL/hr to prevent catabolism and allow a 1-to-1 conversion of IV units to SC Figure 1. Average Blood Glucose of All Glycemic Protocol Runs units. With Standard Deviation (N=470) 5. Hypoglycemia (< 60 mg/dL) would be Average Blood Glucose (with SD) of All Glucommander® Runs treated with 50% Dextrose IV, based on the 200 following formula: (100 – blood glucose) x 180 0.4 = mL of D50 IV push. 160 6. IV insulin therapy would be continued until the morning of postoperative day (POD) #2 140 and longer if a patient were unable to eat 120

or still critically ill. Transition to SC basal 100 bolus therapy would be done only if the Blood Glucose (mg/dL) 80 patient had known diabetes or an A1c > 6%. The transition is initiated by giving 60 0 5 10 15 20 25 glargine insulin at 11p.m. POD #1 if the Hours on Glucommander®

34 Executive Summary Conference Report

Basal: Multiplier *500; CIR: 0.5/Multiplier; Correction Factor: 1.7/Multiplier 220 200 180 n=209 160 140 120 100 80 Blood Glucose (mg/dL) 60 40 0 8 16 24 32 40 48 56 64 72

e st e st e e h er GM kfa kfa t ch ch c n n n Las Lu Dinner Bedtim 3:00 AM Brea Lu Dinner Bedtim 3:00 AM Brea Lu Dinn Bedtim 3:00 AM Bedtim Hours after IV insulin 7th Invited Conference: Use of a Computerized Algorithm in Patients Undergoing Cardiovascular Surgery: A Protocol for Tight Glycemic Control

5. 23% of patients were discharged with were involved in achieving this success. The team members that such a glycemic protocol home insulin therapy: > 90% of those were protocol was designed and implemented as could also be used in normalizing glucose in discharged on basal-bolus therapy. a group effort by diabetes specialists, nurses, other patients in the hospital system. pharmacists and educators under the guid- 6. Nursing feedback after the first 60 days Glucommander® is a product of Glucotec, ance of a CV surgeon who championed the was uniformly positive: the only major ini- Inc., based in Greenville, SC. program to his fellow surgeons and hospital tial complaint was the frequency of glu- administration. Empowerment of the nurses References cose monitoring while a patient is on IV in the design, ownership and implementation 1. Van den Berghe G, Wouters P, Weekers F, et al. insulin. Complaints subsided after nurses of the protocol was crucial. Use of a proven, Intensive insulin therapy in the critically ill patients. realized that patients had fewer infections N Engl J Med 2001; 345:1359-67. 2001. computerized, IV-insulin-dosing algorithm and shorter length of stay. Agency and 2. Furnary AP, Gao G, Grunkemeier GL, et al. Continuous (Glucommander®) with appropriate setting of float nurses were able to learn the protocol insulin infusion reduces mortality in patients with the target range of 90-120 mg/dL allowed all diabetes undergoing coronary artery bypass graft- within their first shift on the CV ICU post- patients to obtain a mean glucose of 107 mg/ ing. J Thorac Cardiovasc Surg. 2003 May;125(5):1007- operative unit or CV floor. 21. dL with no glucose < 40 mg/dL. 7. Physician acceptance of the protocol has 3. Krinsley JS. Effect of intensive glucose management Average Blood Glucose (with SD) of All Glucommander® Runs Transition from IV to SQ insulin in all on the mortality of critically ill adult patients. Mayo been uniformly positive with no criticisms Clin Proc. 2004;79:992-1000. 200 patients with known diabetes or an A1c > 6% or complaints. 180 allowed normalization of blood glucose after 4. Gandhi GY, Nutta GA, Abel MD, et al. Intensive 8. Post-operative length of stay was reduced intraoperative insulin therapy versus conventional 160 eating during the post-operative period until glucose management during cardiac surgery, a ran- by 0.8 days for an estimated annual savings discharge. To accomplish this, all patients were domized trial. Ann Intern Med 2007;146:233-43. 140 of over > $800,000. screened for diabetes and had an A1c drawn 5. Krinsley JS, Grover A. Severe hypoglycemia in criti- 120 Conclusions to identify patients to transition from IV to SQ cally ill patients: risk factors and outcomes. Crit Care 100 Med. 2007 Oct;35(10):2262-7.

Blood Glucose (mg/dL) insulin and who would benefit from insulin This glycemic protocol using a computer- 6. American College of Endocrinology and American 80 after discharge. Any patients admitted with ized system for IV insulin infusion and transi- Diabetes Association Consensus Statement on 60 poorly controlled diabetes were normalized in In-Patient Diabetes and Glycemic Control. Endocr tion to SC basal bolus insulin therapy was 0 5 10 15 20 25 the hospital with insulin and discharged with Pract. 2006 Jul-Aug 12(14):458-68 and Diabetes Care Hours on Glucommander®highly effective in normalizing blood glu- 2006 Aug 29; (8):1955-62. an insulin regimen to keep their blood glu- cose without significant hypoglycemia in all 7. Institute for Healthcare Improvement web- cose levels normal as long as they complied patients undergoing CV surgery at Piedmont site: www.IHI.org/IHI/Topics/PatientSafety/ with their case-specific regimen. It is felt by all SurgicalSiteInfections/Changes/ Hospital in Atlanta. Several important factors SSI+Maintain+Glucose+Control.htm

8. Davidson PC, Steed RD, Bode BW. Glucommander, A Computer-Directed Intravenous Insulin System, Shown to be Safe, Simple, and Effective in 120, Figure 2 . Transition from Glucommander® to Basal-Bolus Insulin 618 h of Operation. Diabetes Care 2005;28(10): Glargine and Aspart 2418-23. Basal: Multiplier *500; CIR: 0.5/Multiplier; Correction Factor: 1.7/Multiplier 220 200 180 n=209 160 140 120 100 80 Blood Glucose (mg/dL) 60 40 0 8 16 24 32 40 48 56 64 72

e st e st e e h er GM kfa kfa t ch ch c n n n Las Lu Dinner Bedtim 3:00 AM Brea Lu Dinner Bedtim 3:00 AM Brea Lu Dinn Bedtim 3:00 AM Bedtim Hours after IV insulin

Executive Summary Conference Report 35 7th Invited Conference: Computerized Management of Tight Glycemic Control “The challenge to imitate a healthy pancreas”

PROCEEDINGS Computerized Management of Tight Glycemic Control: “The challenge to imitate a healthy pancreas” W. Patrick Burgess, MD, PhD, Carolinas Medical Center, Charlotte, NC

Key points dosing relationship. There are thousands of variations in how hospitals and physicians • A primary unresolved issue in critical care is a methodology to reach and maintain blood 20% expect caregivers to perform these calcula- glucose control safely. 18% tions.n > 200,000 Figure dose 1 illustrates calculations a type of risk-benefit in over 5,500 patients • Most paper-based glucose-control protocols rely on bedside16% calculations using linear analysis that compares the performance of a 14% mathematics, whereby lowering the mean blood glucose increases the incidence of hypo- number of published bedside protocols by 12% glycemia. plotting the mean blood glucose (benefit) 10% to the incidence of hypoglycemia (risk). The • Since higher glucose levels are associated with increased8% insulin clearance, an insulin dose is dashed lines in this figure connect the control not linearly related to blood glucose level. 6% and study cohorts from these reported retro- 4% • A computerized system using a nonlinear, physiologic insulin dosing function more closely 1.1% spective or randomized studies. The protocols 2% 0.5% aligns glucose control with the behavior of the human pancreas0.2% and optimizes patient out- 0% 0.1% usually use linear mathematics to determine comes with minimal hypoglycemia. Blood Glucose Range, mg/dLinsulin dosing and have the same characteris- tic: lowering the mean blood glucose increas- • The database generated by a computerized system also provides5 previously10 unavailable 15 20 25 information to help clinicians assess and treat critically ill patients with hyperglycemia. es the incidence of hypoglycemia, although most studies claim no ill effects from the observed hypoglycemia. The notable excep- Glycemic control has value in clinical Protocols medicine because it can reduce the cost of tion is the large 2006 Van den Berghe study, The vast majority of paper protocols care, shorten length of stay, improve heal- in which hypoglycemia was reported as an rely on bedside calculation to make dosing ing, reduce nosocomial infections and reduce independent predictor of death. adjustments based on a linear IV insulin- acute renal failure. A primary unresolved issue is a practical method to reach and maintain glycemic control safely. Current standards Figure 1. Mean Blood Glucose of Study (mg/dL) of care to achieve blood glucose control in (mean of all blood glucose readings on IV insulin) critically ill patients include use of point-of- 20 19 care devices to measure the blood glucose 15 *Hypoglycemia is de ned as BG 40 mg/dL level, bedside mathematical dose calculation 15 and intravenous (IV) insulin. This article com- 7 pares bedside mathematical dose calculation 10 with computerized management of glucose 20 control and reviews concepts of engineer- 14 4 ing-type control mathematics applied to the 5 Safety axis (% of patients)

Incidence of Hypoglycemia* 11 complex problem of controlling elevated 5 Healthy pancreas Computer 17 13 blood glucose levels. 0 60 80 100 120 140 160 180 200 Performance axis

36 Executive Summary Conference Report 7th Invited Conference: Computerized Management of Tight Glycemic Control “The challenge to imitate a healthy pancreas”

Hypoglycemia is a concern in the majority control circuits every day, from thermostats patients with diabetes scheduled for CABG of the reported protocols for insulin infusion and cruise controllers to the auto-pilots that and receiving D10W at 1 mL/kg/hr to either a dosing. None of these protocols achieve the guide passenger planes across the continent. paper protocol or to a computerized protocol same results as the human pancreas, which The task of controlling an insulin infusion (EndoTool®). In this small study the computer- regulates blood glucose without incidence of to regulate elevated blood glucose, which ized method led to better control with a high- severe hypoglycemia. The protocols shown in strongly influences patient outcomes, is left ly significant correlation both in the operating Figure 1 are simple two-point control systems to the bedside determination of a healthcare room (p = 0.001) and in the recovery room that use the current and last blood glucose provider. (p < 0.0001) without severe hypoglycemia levels in a linear relationship to calculate the (< 40 mg/dL). Several other studies2,3 that By applying a computerized, control- next IV insulin dose. It would seem relatively used retrospective control data have shown mathematics approach, more than two dose/ intuitive that the benefits of tight glucose that the application of computer technology response data points can be used to precisely control could be offset by the side-effects can reduce errors and provide more consis- estimate the next dose of IV insulin. This of hypoglycemia. Control of blood glucose tent dosing of IV insulin. type of calculation would be very difficult to should therefore have minimal hypoglycemia perform at the bedside. Applying higher-level Results as one of the primary goals. mathematics to the regulation of the insulin When the mathematical approach used Computerizing Insulin Dose dosing is likely to significantly improve glu- by EndoTool® is applied to glycemic con- Calculations cose control outcomes and result in the main- trol in critical care, glucose control is much tenance of normal blood glucose. The appli- Compared to paper protocols, there are more aligned with the behavior of the human cation of this type of complex mathematics is many advantages to the use of a comput- pancreas (Figure 1). This software protocol only possible with a computer. This sophisti- erized protocol: reduced errors, improved has been the standard of care at one major cated approach is now only available in one protocol consistency, improved compliance hospital for more than four years3. The distri- FDA–approved proprietary software product and discipline, a database for audits, qual- bution of all of their blood glucose results is called the EndoTool® Glucose Management ity assurance and improvement and a plat- illustrated in Figure 2. The incidence of severe System. form for subcutaneous conversion. Another hypoglycemia is less than 1 per 1000 readings, advantage of a computerized insulin-dosing Evidenced-based Medicine with more than 60% of these low readings system is enabling the use of sophisticated associated with either blood glucose determi- A few studies have compared the bedside mathematics that typically is not available at nation more than 30 minutes late or when no mathematics approach to the computerized, the bedside. Basing dose calculations on a insulin was being infused during the previous control mathematics-based protocol for gly- physiologic insulin-dosing relationship is one period. cemic control. Saagar1 randomly assigned 40 such example. Insulin clearance is a function of the glomerular filtration rate, which, in turn, is related partially to the blood glucose Figure 2. Distribution of All Glucose Readings Using Computer Control level. Higher glucose levels are associated with increased insulin clearance; thus, the 20% insulin dose is not linearly related to the blood 18% n > 200,000 dose calculations glucose level. The use of a nonlinear, physi- 16% in over 5,500 patients ologic insulin dosing function should lead to 14% improved glycemic control. 12%

Another example of the use of sophisti- 10% cated mathematics to improve glycemic con- 8% trol is the application of control mathematics. 6% Control mathematics is a scientific discipline 4% 1.1% 2% 0.5% that originated in the field of engineering 0.2% 0.1% and has evolved to become a specialty of 0% computer science. Millions of people rely on Blood Glucose Range, mg/dL

5 10 15 20 25

Executive Summary Conference Report 37

20 19 15 *Hypoglycemia is de ned as BG 40 mg/dL

15

7

10 20

14 4 5

11 5 Healthy pancreas Computer 17 13 0 60 80 100 120 140 160 180 200 Performance axis 7th Invited Conference: Computerized Management of Tight Glycemic Control “The challenge to imitate a healthy pancreas”

Discussion Conclusions References

The potential for a computerized approach The control of elevated blood glucose 1. Saager L, Collins GL, Burnside B, et al. A randomized study in diabetic patients undergoing cardiac sur- to improve glycemic control should be intui- levels in critically ill patients is a complex gery comparing computer-guided glucose man- tive given the potential for errors and the problem. The use of IV insulin is ideal for agement with a standard sliding scale protocol. J Cardiothoracic and Vascular Anesthesia 2007; article complexity of some paper-based glycemic glycemic control because of its short half-life. in press. control protocols. Critical care nurses are over- Computerized calculation of IV insulin dosing 2. Davidson PC, Steed RD, Bode BW. Glucommander: whelmed by new protocols that have to be for glycemic control may be a methodology a computer-directed intravenous insulin system integrated into the care plans. Removing the that can: shown to be safe, simple, and effective in 120,618 h of operation. Diabetes Care 2005; 28:2418-23. need for of bedside calculations by using a • Imitate the human pancreas’ control of computerized system that reduces the work- 3. Dunn K, Miller E, Cochran S, et al. Optimized glu- glucose cose management with EndoTool® intravenous load for the caregiver should be a component insulin dosing. American Diabetes Association, of computerized glycemic control programs. • Optimize patient outcomes with minimal 67th Scientific Session, June 2007; Abstract 0463-P. Controlling the blood glucose level promptly hypoglycemia. can help reduce the frequency of point-of- Computerized management of elevated care determinations and associated costs, and blood glucose in critically ill patients can can lead to control similar to that achieved reduce human errors and improve work flow. by the human pancreas, optimizing patient A database generated by this method also outcomes and reducing the caregiver’s work- has the potential to enhance patient care by load. Available alarms and quality assurance documenting previously unavailable informa- reports can improve compliance with the tion to help clinicians in the assessment and glycemic-control protocol. treatment of critically ill patients with hyperg- lycemia.

38 Executive Summary Conference Report 7th Invited Conference: Analysis of Variation in Insulin Protocols

PROCEEDINGS Analysis of Variation in Insulin Protocols Guy W. Soo Hoo, MD, MPH; Pulmonary and Critical Care Section; West Los Angeles Healthcare Center; VA Greater Los Angeles Healthcare System

Key points hyperglycemia and the merits of tight glu- cose control (TGC) defined as serum glucose • Existing insulin infusion protocols vary greatly in the details of implementation and can between 80-110 mg/dL have been well docu- result in great variations in insulin dose recommendations. mented4-8. Universal acceptance of guidelines • Protocols developed with one group of patients may require validation when used for other for TGC is tempered by limited scientific sup- patients, but ease of use and efficacy in patients are important features of any protocol. port that remains confined to the experience 8,9 • A trial using a protocol may be necessary in deciding which one to use. from one medical center .

• Insulin protocol innovations include the use of nomograms, pre-printed tables and internet- Many of these protocols have been incor- based or computer-based protocols with automatic dose calculations. porated into best practice models referred to as "bundles," with ventilator or sepsis bun- • Involving and empowering nursing staff is critical for the success of any protocol. dles the most common. Greater benefit may • Different protocols may be required for different patients in the same hospital because one result from the use of bundled management protocol may not fit all. compared to the interventions alone. Many organizations have taken the lead in this approach, with the Institute for Healthcare Protocol Development cols developed for other patients may com- Improvement (IHI) conducting one of the Managing critically ill patients has dramat- promise patient comfort or safety. Protocols most visible campaigns advocating ventilator ically changed over the past decade. Many require flexibility and adaptation for suc- and central line bundles. Severe sepsis guide- large clinical trials have identified optimal cessful application in any patient population. lines have also garnered much attention. approaches or in areas where there Successful implementation requires a critical The first guidelines arose from collaboration was once a high degree of variation. There mass of organizational and administrative between the Society of Critical Care Medicine, are now established protocols that have support, especially for protocols that require European Society of become part of the routine care of critically acceptance and use by several healthcare and International Sepsis Forum, who col- ill patients. Protocols reduce variation in care providers. laborated on the "Surviving Sepsis" cam- 10 and promote the best available therapeu- Examples of widely accepted protocols paign with subsequent support from IHI. 1 tics or best practices . Adherence to these used in intensive care units (ICU) include Although questions have been raised about protocols should improve patient care and ventilator management of acute lung injury the objectivity of the recommendations, the 2 patient outcomes . Protocols are intended (ALI) and acute respiratory distress syndrome guidelines are highly visible within the medi- 11 to complement care, however, and are not a (ARDS), weaning patients from ventilators, cal community . The sepsis bundle includes substitute for sound clinical judgment. sedation, analgesia, transfusions and more recommendations for adequate glycemic 1,3. Intensive control defined as a serum glucose < 150 Protocols are not a "cookbook" approach recently antimicrobial stewardship insulin therapy (IIT) to control hyperglycemia mg/dL achieved by administering continuous to patient care. Protocols may not necessarilly in critically ill patients is another area of great infusions of insulin and glucose. The group translate well from the research setting to interest. Adverse outcomes associated with acknowledged the limited scientific support general patient use. Blind adoption of proto- for this recommendation. Others have been

Executive Summary Conference Report 39 7th Invited Conference: Analysis of Variation in Insulin Protocols more circumspect with their endorsement of infusion protocols, selected from 24 proto- sion rate. While most changes to the insulin intravenous (IV) insulin infusions and treat- cols. Significant variation was noted among infusion rate are fixed, some protocols require ment of hyperglycemia12. Nevertheless, insu- these protocols in all aspects of protocol adjustment using a multiplier. These adjust- lin infusion protocols have become common. implementation, beginning with the initial ments may or may not require mathematical insulin dose, subsequent dosing adjustments calculations with each change in the insulin Insulin Infusion Protocols to other logistical matters. dose. The target goal glucose was within the We surveyed a convenience sample of 80-180 mg/dL range; some advocating an Major Areas of Variation surrounding institutions and found that all 80-110 mg/dL goal, and others with a range had insulin infusion protocols. The protocols Major areas of variation are summarized of 120-180 mg/dL. Variations in the protocols varied, ranging from the use of published in Table 1. Notable differences included the are further outlined in Table 2 from Wilson, et protocols, to slightly modified protocols, to threshold for initiating an insulin infusion, the al13. others entirely unique to the institution. The initial use of bolus insulin and subsequent While most of the protocols were nurse- number of protocols and differences between doses of bolus insulin. Bolus insulin treat- driven with limited physician input, van den protocols is significant. The task of identify- ment resulted in larger doses administered Berghe reported active physician guidance14. ing the most appropriate protocol can be a earlier in the course of an insulin protocol. All except the Furnary protocol required a challenge for those trying to implement a Subsequent adjustments to the insulin rate continuous glucose source, parenteral or protocol in their own institution. A protocol were based on a variety of factors, including enteral, during continuous insulin infusion15. may be successful in one institution, but the direction and rate of change in serum Adjustments to hypoglycemic episodes also there is no guarantee of its efficacy when glucose, insulin resistance with the actual varied with differences in the magnitude of used in another hospital with different nurs- change either fixed or calculated based on adjustment as the patient approached hypo- ing staff and a different patient population. the actual glucose value or the insulin infu- It is impractical to pilot test every available Table I. Areas of Variation in Insulin Infusion Protocols protocol and developing a new protocol may INSULIN INFUSION RATE be time-consuming. Presence or absence of pre-existing diabetes The diversity found in our survey of local Initial hyperglycemic threshold ( > 150-200 mg/dL) hospitals prompted a more detailed review Initial bolus insulin dose (calculated [formula] versus fixed [pre-determined]) of published protocols13. The analysis started with a search of published protocols using Subsequent bolus insulin Some with larger doses earlier in the protocol the PUBMED search engine and the terms "intravenous insulin" and "insulin protocols." Changed in insulin dose (calculated vs fixed dose) Based on direction of change in glucose (decrease, no change, increase) The focus was on protocols used for criti- Velocity of change (30-50 mg/dL/hr) cally ill patients. Glucose-insulin-potassium Dose adjustment for insulin resistance (GIK) protocols, although similar, were not Basis of change in insulin dose (rate, infusion) included, since the basis of these protocols Unit of change in insulin dose (fixed dose or calculated using a multiplier) was not glucose control but a reduction in Number of steps required (1-3); some requiring calculations to change the dose (8 of 12) free fatty acids and involved a different group Target glucose (80 – 180 mg/dL; with a wide range of acceptable values) of patients. The focus was on paper-based LOGISTICS OF IMPLEMENTATION protocols, given their wide availability and ease of use. Computerized protocols were Nurse-driven, with varying physician input Adherence to protocol or allowance for changes in protocol ("protocol violations") not reviewed given limited access to the pro- grams. Differences in frequency of glucose monitoring

Based on a review of protocols identified Differences in requirement of a constant glucose source (parenteral or enteral infusions) it became clear that some protocols had Adjustments for hypoglycemia (variable levels of adjustment after hypoglycemia) many versions. In these instances, the latest Time to goal (variable; from 2-24 hours) published protocol was used for review and Duration within goal (not consistently reported) analysis. The review focused on twelve insulin

40 Executive Summary Conference Report 7th Invited Conference: Analysis of Variation in Insulin Protocols glycemia and variable adjustments to the insulin rates doses. In one protocol, a patient and American Diabetes Association16,17. They insulin rate when it was restarted. Time to received almost 100 units of insulin before outlined necessary requirements for inpa- goal and duration of time in the goal range reaching the target of 80-110 mg/dL. Almost tient management, including administrative were variably reported, ranging from 2-24 60% of the insulin was infused after a serum support, multi-disciplinary committees, with hours to goal and duration in the goal range glucose of 200 mg/dL with 15 units/hour analysis of current practices and barriers to in the 40-60% range. the highest infusion rate. Among the other implementation. Key components included protocols, the recommended doses of insulin standardized order sets or algorithms and The variation between protocols would varied widely, ranging from a total of 27-115 appropriate metrics to evaluate the impact predictably produce different recommenda- units, and from 4-21 units/hr. of their programs. They felt the best insulin tions and results. While comparison of the protocols would include consideration of cur- details of the protocols was possible, compari- No Single Approach rent and previous glucose levels, adjustments son of the performance of the different proto- It is clear that no single approach or pro- of the rate of glucose change, and the cur- cols was impossible. Therefore, comparison of tocol can anticipate and accommodate the rent insulin infusion rate. Recognition and protocols was made using a computer simula- great variety of critically ill patients. Serum adjustment for hypoglycemia was also cru- tion using the results from an actual patient glucose levels are affected by a host of cial, and each protocol in use would require treated with the van den Berghe protocol. The influences, including infection, renal insuf- validation and an ongoing assessment of key assumption of the simulation was that the ficiency, catabolic stress and erratic nutri- efficacy and safety. IIT also requires more changes in glucose occurred as a result of the tion that lead to dynamic and fluctuating nursing time, another consideration when insulin dose recommendations of the proto- insulin requirements that would probably be choosing a protocol, since time spent on the col being evaluated. The limitations of such best addressed by frequent glucose monitor- protocol, whether in monitoring or calculat- a simulation exercise are well acknowledged, ing and a continuous insulin infusion. The ing changes in the infusion rate, eventu- since the actual changes in glucose would increasing use of IV insulin infusions for glu- ally affects other patient care areas. The index differ with each protocol. This approach does cose control has led to recommendations patient described had 20 glucose determina- however provide insight into the differences by the American College of Endocrinology tions during the initial portion of his infusion, among the protocols and their recommended

Table 2. Areas of Variation in Insulin Infusion Protocols13

Bolus Changes insulin infusion based Basis of changes in Steps for insulin on damages inglucose insulin rate insulin adjustment Author Target Initial Add Direction Velocity Resistance R or I U±% n/calculations Time to goal glucose Y/N glucose (mg/mL) Bode 100-150 Y* N N Y Y R U 3/N NR Boord 120-180 N N N N N R U 1/Y NR Chant 90-144 N Y Y Y N R U+% 2/Y 15h Davidson < 100 N N N N Y R Multiplier 3/Y 7.5-10.5h Fumary 100-150 Y Y Y Y Y R U+% 2/Y NR Goldberg 100-139 Y N Y Y N R+1 U+% 3/Y 9.0h Kanji 80-110 N N Y Y Y R U+% 2/Y 11.3±7.9h Krinskey < 140 N N N N N R U 1/N NR Marks 120-180 N N N N N R U 1/N NR Van den Berghe 80-110 N N N Y N R U+% 2/Y 12-24h Watts 120-180 N Y N N N R U 1/N 8h Zimmerman 101-150 Y Y N N N R+1 U+% 2/Y 2.1h

Copyright © 2007 American Diabetes Association from Diabetes Care®, Vol. 30, 2007; 1005-1011 Reprinted with permission from The American Diabetes Association.

Executive Summary Conference Report 41 7th Invited Conference: Analysis of Variation in Insulin Protocols

which would translate into 100 minutes spent References 9. Van den BG, Wilmer A, Hermans G, et al. Intensive on glucose management alone, allotting five insulin therapy in the medical ICU. N Engl J Med 1. Holcomb BW, Wheeler AP, Ely EW. New ways to 2006;354(5):449-61. minutes for each glucose check. reduce unnecessary variation and improve out- comes in the intensive care unit. Curr Opin Crit Care 10. Dellinger RP, Carlet JM, Masur H, et al. Surviving Summary 2001; 7(4):304-11. Sepsis Campaign guidelines for management of severe sepsis and septic shock. Crit Care Med 2004; Existing insulin infusion protocols vary 2. Meade MO, Ely EW. Protocols to improve the care of 32(3):858-73. critically ill pediatric and adult patients. J Amer Med greatly in the details of implementation. The Assn 2002;288(20):2601-3. 11. Eichacker PQ, Natanson C, Danner RL. Surviving differences in protocols can result in great sepsis--practice guidelines, marketing campaigns, 3). Barlam TF, DiVall M. Antibiotic-stewardship practic- and Eli Lilly. N Engl J Med 2006; 355(16):1640-2. variations in insulin dose recommendations. es at top academic centers throughout the United 12. Angus DC, Abraham E. Intensive insulin thera- Protocols developed with one group of States and at hospitals throughout Massachusetts. Infect Control Hosp Epidemiol 2006;27(7):695-703. py in critical illness. Am J Respir Crit Care Med patients may require validation when applied 2005;172(11):1358-9. 4. Capes SE, Hunt D, Malmberg K, et al. Stress hypergly- to other patients. The ease of use and efficacy caemia and increased risk of death after myocardial 13. Wilson M, Weinreb J, Soo Hoo GW. Intensive insulin in patients are important features of each infarction in patients with and without diabetes: a therapy in critical care: a review of 12 protocols. Diabetes Care 2007;30(4):1005-11. protocol. A fair assessment of a protocol may systematic overview. Lancet 2000;355(9206):773-8. not be possible without a treatment trial. 5). Capes SE, Hunt D, Malmberg K, et al. Stress hyper- 14. van den Berghe G, Bouillon R, Lauwers P. Intensive glycemia and prognosis of stroke in nondiabetic insulin therapy in critically ill patients; supplemen- New innovations are emerging, including the and diabetic patients: a systematic overview. Stroke tary material. N Engl J Med 2002;346:1587-8. 2001; 32(10):2426-32. use of nomograms, pre-printed tables, inter- 15. Furnary AP, Wu Y, Bookin SO. Effect of hyperglyce- net-based or computer-based protocols with 6. Finney SJ, Zekveld C, Elia A, et al. Glucose control mia and continuous intravenous insulin infusions automatic dose calculations. Irrespective of and mortality in critically ill patients. J Amer Med on outcomes of cardiac surgical procedures: the Assn 2003;290(15):2041-7. Portland Diabetic Project. Endocr Pract 2004;10 the protocol chosen, involving and empow- (Suppl 2):21-33. ering nursing staff is critical for its success. 7. Krinsley JS. Association between hyperglycemia and increased hospital mortality in a heteroge- 16. American College of Endocrinology and American There may not be one protocol suitable for neous population of critically ill patients. Mayo Clin Diabetes Association Consensus statement on all patients, and different protocols may be Proc 2003;78(12):1471-8. inpatient diabetes and glycemic control: a call to action. Diabetes Care 2006; 29(8):1955-62. required for different patients in the same 8. Van den BG, Wouters P, Weekers F, et al. Intensive insulin therapy in the critically ill patients. N Engl J 17. American College of Endocrinology and American hospital. One protocol may not fit all. Med 2001;345(19):1359-67. Diabetes Association consensus statement on inpatient diabetes and glycemis control. Endocr Pract 2006;12(Suppl 3):4-13.

42 Executive Summary Conference Report 7th Invited Conference: Improving ICU Quality and Safety–Implications for Tight Glycemic Control

PROCEEDINGS Improving ICU Quality and Safety: Implications for Tight Glycemic Control Sean Berenholtz, MD MHS FCCM, Johns Hopkins University,Quality and Safety Research Group, Baltimore, MD

Key points and mortality and improvements in patient satisfaction, rates of nosocomial infections, • Structured approaches to improve the reliability of care and teamwork climate in intensive glucose control and sepsis care. care units (ICUs) have improved patient safety and outcomes. For example, this model has been used • The link between teamwork climate and tight glucose control (TGC) has not been to reduce catheter-related bloodstream investigated; however, TGC is clearly a team effort. infections (CR-BSIs), a completely prevent- • Lessons learned from quality and safety research to improve infection control and able complication that currently results in post-surgical care may well be applicable to TGC. one to eleven patient deaths every day. Studies show that CR-BSIs can be reduced by the following steps, ensuring that patients only about 50% of the time1. To improve reli- The Johns Hopkins Quality and Safety receive five evidence-based interventions as ability of care:2 Research Group (QSRG), which manages sev- described in Center for Disease Control (CDC) eral large collaboratives including more than • Pick a clinically important area—some- guidelines3: 250 intensive care units (ICUs), has a strong thing commonly seen in the ICU or associ- • Apply five best practices focus on quality and safety issues. Studies ated with significant morbidity and mor- in these ICUs have shown that a culture of tality. – Remove unnecessary lines safety, particularly a teamwork climate, is a – Perform hand hygiene predictor for many clinically important out- • Identify “What should we do?”—what evi- comes. Fortunately, teamwork and the safety denced-based therapies should be pro- – Use maximal barrier precautions climate can be improved. vided for patients? – Use chlorhexidine for skin antisepsis The link between teamwork climate and • Measure performance—translate those – Avoid femoral lines tight glycemic control (TGC) has not been therapies into behaviors that can be mea- investigated but TGTC clearly requires team- sured validly and feasibly with data col- • Decrease complexity, e.g., create a “line work. In this article the Johns Hopkins model lected concurrently at the bedside and cart” that has all the necessary supplies, so for improving the reliability of care, the four analyze whether those behaviors are that providers do not have to go to eight critical components of all change efforts, being done. different places to gather their supplies. and the Johns Hopkins Comprehensive Unit- • Focus on improving the systems of care • Create redundancy, e.g., create a checklist based Safety Program (CUSP) are discussed. (decreasing complexity, creating redun- to be completed by the nurse at the bed- Improving Reliability of Care dancy and learning from mistakes) to side and then empower the nurses to stop improve performance. the procedure if a violation is noted. The Johns Hopkins QSRG model for improving the reliability of care that has been • Evaluate whether improvement in mea- Use of this model reduced the rate of developed over the past five years is based on sures has been achieved. CR-BSIs in the Johns Hopkins ICU from about literature showing that patients receive evi- Many individual hospitals have used this the 50th percentile, the National Nosocomial dence-based therapies they should receive model and seen reductions in complications Infection Surveillance benchmark, to less than one per thousand catheter days. Because

Executive Summary Conference Report 43 7th Invited Conference: Improving ICU Quality and Safety–Implications for Tight Glycemic Control safety efforts focused on systems change, accidents happened on the first day a flight evaluation of sentinel events and root cause this dramatic reduction has been maintained crew worked together. In healthcare, there analyses have shown that in the vast majority in an academic medical center for the past is little training on teamwork. In labor and of instances, somebody knew something was three years. This approach can be broadly delivery, intensive care units and the operat- not comfortable speaking up, or they spoke generalized. In Michigan, for example, within ing room, survey data show that physicians up and their concerns were not acknowl- three months of implementing all the above uniformly overrate communication and col- edged. interventions in 136 ICUs across the state, the laboration compared to nurses5. Teamwork also is a predictor for other clini- median rate of CR-BSIs went to zero4. A highly validated safety culture assess- cally important outcomes such as wrong-site Technical and Adaptive Components ment instrument is the safety attitude ques- surgeries, decubitus ulcers, delays in starting of Change tionnaire (SAQ)6. The SAQ evaluates a variety in the operating room, bloodstream infec- of domains within the ICU—job satisfaction, tions, post-operative sepsis, post-operative One of the most important lessons from teamwork climate, safety climate, percep- infections, post-operative bleeding, pulmo- our experience and the literature is the four tions of management, stress recognition and nary embolism/deep vein thrombosis, venti- critical components for all change efforts. working conditions—as perceived by front- lator-associated pneumonia, nursing turnover • Engage. Providers and front-line staff line nurses, physicians, respiratory therapists, and absenteeism5. The link between safety or need to be engaged. More than just pro- pharmacists, aides, secretaries and other teamwork and tight glucose control has not viding data back to providers, what can staff. been investigated; however, TGC is clearly a we do to touch their hearts or create team effort. The SAQ definition of teamwork is “the that imperative for change? Telling stories perceived quality of collaboration between CUSP is one example. Many hospitals identify personnel in this unit.” Examples include patients who had adverse events and then Another important lesson from our collab- “Disagreements in the ICU are appropriately feed those stories back to staff. orative efforts is that teamwork and a safety resolved (i.e., not who is right, but what is best climate can be improved, as shown by results • Educate. Provide relevant educational for the patient)” and “Our doctors and nurses achieved by CUSP7. Data from the general material. Many providers are not able, work together as a well coordinated team.” surgical ICU (SICU) and the Weinberg oncol- interested or willing to read a 72-page CDC Safety climate is defined as “perceptions of ogy ICU (WICU) at Johns Hopkins showed dra- guideline on the prevention of CR-BSIs, so strong or proactive commitment to patient matic improvements in the culture of safety. distill the information down into discreet safety in this unit,” e.g., “I would feel safe being The SICU improved from one-third to 70% of behaviors known to be important. The treated in this ICU” and “Medical errors are respondents reporting a good safety culture biggest gap between current performance handled appropriately in this ICU6.” pre- and post-CUSP, and the WICU improved and best evidence is not that providers Baseline SAQ results from Michigan from one-third to almost 90% (Figure). In disagree with the evidence, but that they ranged from an ICU in which only 15% of staff Michigan, after two years there has been an do not know the evidence exists. agreed they have good teamwork to another incremental improvement of approximately • Execute. Provide templates or examples of ICU in which almost 90% agreed2. When sur- 2% to 10% in teamwork and safety culture in how to create redundancy or standardized vey results were correlated with subsequent ICUs (Figure)2. Culture change takes time. care within the ICU. Many of ICU teams efforts to reduce CR-BSIs rates, the results To track changes in SAQ results over time, want examples they can adapt to their were striking. Of ICUs in the lowest tercile in hospitals are classified as ‘needs improve- local environment. teamwork climate, 21% went five months or ment’ if less than 60% of the providers say more without a CR-BSI, compared to 44% in • Evaluate. Provide valid tools to evaluate that they have a good teamwork or safety the highest-tercile ICUs. The strongest pre- whether care has improved. culture. From 2004 to 2006 CUSP implemen- dictor of an ICU’s ability to reduce its CR-BSI tation decreased the percent of ICUs across Context of Care rate was the answer to a single question: “Do Michigan that “need improvement” from 84% caregivers feel comfortable speaking up if Teamwork is a critically important element to 41% for safety climate and from 82% to they perceive a problem with patient care?”2 in the culture of safety. An aviation industry 47% for teamwork climate. study showed that most commercial aviation Perhaps not surprising when we consider that

44 Executive Summary Conference Report 7th Invited Conference: Improving ICU Quality and Safety–Implications for Tight Glycemic Control

The CUSP iterative process includes the Providing an opportunity to create struc- “I support you” compared with “I’m there for following steps7: tured communication between the ICU you three times a week or two times a week.” attending physician and the charge nurse can Leadership needs to be engaged. 1. Evaluate culture of safety lead to valuable improvements with regard to Learning from mistakes by asking key ques- 2. Educate staff on science of safety staffing teamwork and communication. tions is also a critical component of CUSP10. 3. Identify defects Another tool is shadowing another person What happened? Why? What will you do to 4. Assign executive to adopt unit in the ICU. A physician might shadow a respi- reduce probability that it will happen again? ratory therapist or a nurse. Medical students How do you know risk is reduced? 5. Learn from one defect per month and can shadow physicians on rounds to observe Finally, we have been working with teams implement teamwork tools how effectively they communicate with bed- to develop strategies to track progress in side nurses, residents, other attending physi- 6. Re-evaluate culture improving patient safety. A safety scorecard cians and providers. Their observations can Another important lesson learned is that can be developed to track safety and team- change the way physicians communicate. there may be at least a fourfold over-report- work on measures such as the number of ing of teamwork and safety culture by senior A culture check-up tool is a structured BSI/1000 patient days, percent of patients executives compared to front-line staff. One approach to assessing improvement in safety receiving the ventilator bundle of evidence- 11 approach is to have a senior executive “adopt culture . Results can be fed back to and be based practices, percent of months in which an ICU8.“ The executive comes to that ICU, used by teams so they can focus on a specific a unit learned from a defect, the number of meets with the staff, perhaps to focus on question in the culture survey and develop units in which 60% of staff report positive some of the safety defects and learns to bet- strategies to improve that. The team check- teamwork and safety climate and average ter understand the clinical improvement pro- up tool adds science to quality improvement score in each ICU. The scorecard, including cesses and what can senior executives can do by identifying explicit barriers and successful safety and teamwork findings, can help feed to help improve teamwork, safety and culture strategies. Teams can use this tool to discuss results back to staff and increase awareness of to try to fix defects. these barriers and improvement strategies safety throughout the organization, including with senior executives. senior leadership12,13. Another important step is to learn from defects and implement teamwork tools to Leadership support is critically important Summary for quality improvement. However, there is a prevent identified mistakes from happening Front-line staff and senior leadership need difference between a senior executive saying, again. to view safety as a science and focus on Teamwork tools. One teamwork improve- ment tool is the implementation of daily Figure. Safety Climate in the WICU and SICU Pre-Post CUSP goals9. Setting daily goals is a powerful tool to me 3 improve communication and teamwork used Ti 100 in hundreds of ICUs. Another tool is morn- me 3 90 WICU ost CUSP ing briefings10. Before starting rounds in the Ti 80 te SICU ICU, the ICU physician meets with the charge ost CUSP 70 SICU P nurse to ask three questions: 60 e CUSP e CUSP WICU P ty clima

• Was there anything that happened last fe 50

40 SICU Pr

night that I need to know about? WICU Pr 30 • Are there any flow issues within the ICU, 20 ing good sa

admissions or discharges you’re concerned rt 10 about, i.e., where should I start rounds?

% of respondents within a clinical area repo 0 • Are there any anticipated problems today? Safety climate in surgical ICU (SICU) and Weinberg ICU (WICU) at The Johns Hopkins Hospital over time and pre- and Staffing issues are commonly identified as post- implementation of the Comprehensive Unit-based Safety Program (CUSP). Each yellow bar represents the per- a problem area. cent of providers in an individual ICU, clinical area or ward that reported a good safety climate on the Safety Attitudes Questionnaire (SAQ).

Executive Summary Conference Report 45 7th Invited Conference: Improving ICU Quality and Safety–Implications for Tight Glycemic Control systems to ensure patients receive the thera- References 8. Pronovost PJ, Weast B, Bishop K, et al. Senior execu- tive adopt-a-work unit: a model for safety improve- pies they should. Both the technical and the 1. McGlynn EA, Asch SM, Adams J, et al. The quality ment. Jt Comm J Qual Saf 2004 Feb;30(2):59-68. adaptive components of change must be of health care delivered to adults in the United States. N Engl J Med 2003;348:2635-45. 9. Pronovost P, Berenholtz S, Dorman T, et al. addressed. Culture trumps strategy, and effi- Improving communication in the ICU using daily cient, structured approaches must be used to 2. Health Services Research, 2006;41(4 Part II):1599. goals. J Crit Care 2002;18(2):71-5. learn from mistakes and improve safety cul- 3. Guidelines for the Prevention of Intravascular 10. Thompson D, Holzmueller C, Hunt D, et al. A morn- Catheter-Related Infections. Center for Disease ture. Efforts to improve glucose control within ing briefing: setting the stage for a clinically and Control. CDC August 9, 2002 / 51(RR10);1-26. operationally good day. Jt Comm J Qual Patient Saf. the ICU would be remiss if they do not explic- http://www.cdc.gov/mmwr/preview/mmwrhtml/ 2005 Aug;31(8):476-9. itly address culture and the prior beliefs of rr5110a1.htm (Accessed January 7, 2008) 11. Sexton BJ, Paine LA, Manfuso J, et al. A check-up for the ICU staff. Fortunately, tools such as CUSP 4. Pronovost P, Needham D, Berenholtz S, et al. safety culture in "my patient care area". Jt Comm J An intervention to decrease catheter-related can now be used to help improve culture. The Qual Patient Saf 2007 Nov;33(11):699-703, 645. blood stream infections in the ICU. N Engl J Med ultimate goal is to help teams be able to say 2006;355:2725-32. 12. Pronovost P, Holzmueller CG, Needham DM, et al. How will we know patients are safer? An organiza- that a patient is less likely to be harmed this 5. Unpublished data, personal communication, Dr. tion-wide approach to measuring and improving year as opposed to last. Bryan Sexton at The Johns Hopkins University, safety. Crit Car Med 2006;34(7):1988-95. 2007. 13. Pronovost PJ, Berenholtz SM, Needham DM. A 6. Colla J, Bracken A, Kinney L, et al. Measuring patient framework for healthcare organizations to develop safety climate: a review of surveys. Qual Saf Health and evaluate a safety scorecard. J Amer Med Assn Care 2005;14:364-6. 2007;298(17):2063-5. 7. Pronovost P. Improving care throughout Maryland: Introducing CUSP. Jt Comm J Qual Patient Saf 2006 Feb;32(2):102-8.

46 Executive Summary Conference Report 7th Invited Conference: Specialized Nutrition Support and Glycemic Control

PROCEEDINGS Specialized Nutrition Support and Glycemic Control Kalman Holdy, MD, ABPNS, Medical Director of Clinical Nutrition, Sharp Memorial Hospital, San Diego, CA; Clinical Professor of Medicine, University of California, San Diego

Key points of hospitalized patients. One of these con- cerns is the possibility that inadequate calorie • Hyperglycemia associated with specialized nutrition support (SNS) may compromise intake will accelerate malnutrition because intended treatment goals. Thus, clinicians need to consider physiologic glycemic of the “hypermetabolism” of illness and the management as a primary goal. nutrient needs of patients will be unmet dur- • A new approach is to: ing illness. Most illnesses do not accelerate

– Provide SNS to all patients based on nutrition assessment and severity of illness, with total energy expenditure significantly and glycemic control given priority over provision of adequate energy. short-term underfeeding is well tolerated by most non-critically ill patients1. – After glycemic control is achieved, usually in 24 to 48 hours, advance nutrition while In contrast, the harm from overfeeding is maintaining glycemic control. now well recognized. The benefit of meeting – The Nutrition and Metabolic Support Service at Sharp Memorial Hospital has observed metabolic needs at the expense of creating that this approach minimizes hyperglycemia for patients receiving both parenteral and hyperglycemia is doubtful at best and harm- enteral nutrition. ful at worst. From the Leuven studies2. it is – For patients who are eating, a “glycemic control diet”—which provides controlled clear that in critically ill patients, providing calories, adequate protein and the same carbohydrate with each meal—allows more SNS while permitting hyperglycemia results predictable prandial insulin dosing, compared to the conventional 1800-2000 calorie in worse outcomes than maintaining eug- 3 diet, is well accepted by patients and leads to better prandial glycemic control. lycemia (Figure) . No comparable study is available in patients who are not dependant on SNS, but the harmful effects of hyperg- lycemia related to diet appear clear (Table The benefits of tight glycemic control a reasonable approach regarding hospital 1). Hyperglycemia during feeding results (TGC) are being embraced for the care of nutrition is to provide nutrition in a manner in catabolism, is immunenosuppresive and critical and acute care patients. Successful that fosters glycemic control as first priority proinflamatory, and delays gastric empty- glycemic management is integrally related and achieves “goal” calories as a secondary ing. Postprandial hyperglycemia is recog- to the nutrition provided to a hospitalized priority. This prioritized approach is increas- nized as a major element in the long-term patient. A new look should be taken at hospi- ingly used with SNS and should be applied to complications of diabetes. Based on these tal nutrition with physiologic glycemic man- all hospital nutrition, including oral diets. observations, nutrition-related hyperglyce- agement as a primary goal. The glycemic Nutrition Goals and Fears mia appears to be counterproductive to the effects of specialized nutrition support (SNS) basic goals of hospital nutrition. for hospitalized patients (parenteral [PN] and/ Nutrition goals for hospitalized patents or enteral nutrition [EN]) have been exam- include the provision of essential nutrients Another concern about hospital nutrition ined more than the effects of oral diets, and to maintain energy, metabolic homoeosta- which affects feeding recommendations is lessons learned from the glycemic effects of sis, organ function and functional capacity. that hospital feeding reverses the catabolic SNS should be applied to patients taking oral There are several concerns about the cur- effects of illness. A large body of evidence diets. Based on the currently available data, rent recommendations for the nutrition care indicates that no single or combination of

Executive Summary Conference Report 47 7th Invited Conference: Specialized Nutrition Support and Glycemic Control macronutrients prevents illness-related hospital setting7. The American Diabetes grams of glucose (references)—the equivalent catabolism4. This is not to say that prolonged Association position statement regarding the carbohydrate (CHO) contained in about starvation in illness is desirable or can be medical nutrition therapy make 63 recom- five cans of common soft drinks. tolerated indefinitely5. Feeding that induces mendations8. Approximately 50% of these When viewed in this perspective, such recommendations are counter-intuitive and Figure. Surgical ICU Mortality with SNS2 not what would be recommended even for healthy individuals. Two hundred grams of 12% clear-liquid CHO will induce hyperglycemia, 80-110 mg/dL 10% 180-200 mg/dL especially if prescribed without regard to weight. Typically, a patient with a body mass 8% index (BMI) of 30 consumes about 250 grams CHO with mixed glycemic index9, much lower 6% than the glycemic index of 200 grams CHO in 4% clear-liquid diet. Giving 200 grams clear-liquid

rcent of patients CHO to newly hospitalized patients, often Pe 2% with poorly known insulin sensitivity and often irregular intake, makes prandial insulin 0% administration difficult. As a result, it is overly ICU HOSP simplistic to just recommend “adequate insu- Adapted from Van den Berghe, et al. New Engl J Med 2001;345:1359-67. lin” to control the glycemic excursion related to meals. hyperglycemia adds to the catabolic effects are Grade A-B recommendations; the oth- Glycemic Control and Nutrition: of illness, rather than achieving the desired ers are mostly expert opinion. This strongly A New Approach anabolic effect. indicates the lack of adequate research in A rational approach to nutrition for hospi- this area. The American Association of Clinical Current Recommendations talized patients that avoids hospital-related Endocrinologists (AACE) position statement malnutrition is to base nutrition intake on the The question is whether we first meet about hospital nutrition states that adequate degree of malnutrition and seventy of illness. estimated nutrition needs at the expense nutrition intake must be assured and that Such an approach permits under-feeding, of hyperglycemia, or stabilize glycemia first calories restriction is not the way to main- when it is safe, and ensures that adequate and then synchronously increase energy tain glycemic control; rather, adequate insulin nutrition is provided during prolonged hospi- intake and insulin as needed while main- should be used. The sequence mentioned talization and severe illness. taining glycemic control? Current recom- above, however, is not addressed in these rec- mendations stress the former based on an ommendations. In fact, the recommendation Our Nutrition and Metabolic Support application of outpatient approaches to the is that all clear-liquid diets should contain 200 Service has used this approach to SNS. We have minimized hyperglycemia both in patients receiving PN and EN5. We stabilize Table 1. Harmful Effects of Hyperglycemia Related to Nutrition glycemic control first, then start with EN when the blood glucose level is < 200 mg/dL or PN • Impaired protein synthesis at < 15 cal/kg/day, and advance to measuring • Catabolic resting energy expenditure [REE] or synchro- nously estimating increasing insulin doses as • Immunosuppresive and proinflamatory effects needed. For patients who are eating, we have • Increased mortality and complications in the critically ill developed a “glycemic control diet” (Table 2), • Delayed gastric emptying which provides controlled calories, adequate protein and the same CHO with each meal, to allow more predictable prandial insulin

48 Executive Summary Conference Report 7th Invited Conference: Specialized Nutrition Support and Glycemic Control dosing. Our unpublished observation is that all eating patients. Needless to say, we avoid References this diet is well accepted by patients and glucose in clear-liquid diets; in fact, we dis- 1. Holdy K. Monitoring energy metabolism with indi- leads to prandial glycemic control with fewer courage clear-liquid diets for most patients. rect calorimetry: instruments, interpretation, and clinical application. Nutr Clin Pract. 2004;19:447-54. excursions compared to starting with the Summary conventional diet of 1800 to 2000 calories for 2. Van den Berghe G, et al. Intensive insulin therapy in mixed medical/surgical intensive care units: benefit Nutrition should be provided to all patients versus harm. Diabetes. 2006;55:3151-9. based on nutrition assessment and severity of Table 2. Glycemic Control Diet 3. Van den Berghe G, Wouters P, Weekers F, et al. illness, with glycemic control given priority Intensive insulin therapy in critically ill patients. N • High protein ~ 70 gms over adequate energy provision. After glyce- Engl J Med 2001;345:1359-67. • Limited energy ~ 1200 cal/day mic control is achieved, usually in 24 to 48 4. Plank LD and Hill GL. Energy balance in critical ill- ness. Proc Nutr Soc. 2003;62: 545-52. • Controlled carbohydrate hours, nutrition should be advanced while maintaining glycemic control. This principle 5. Bistrian BR. Clinical dilemmas: clinical epistemol- • High fiber ogy, or what do we do till the PRCT comes? Nutr should be applied both to eating patients and Clin Pract. 1995;12:264-5. • No HS snack to those receiving SNS. 6. Schafer RG, et al. Translation of the diabetes nutri- • Three day mandatory RD review tion recommendations for health care institutions. Diabetes Care. 1997;20:96-105.

7. Boucher JL, et al. Inpatient management of diabe- tes and hyperglycemia: implications for nutrition practice and the food and nutrition professional. Am Diet Assoc. 2007;107:105-11.

8. American Diabetes Association Standards of Medical Care in Diabetes. Diabetes Care. 2007; 30:S4-41S.

9. Monnier L, et al. The loss of postprandial glyce- mic control precedes stepwise deterioration of fasting with worsening diabetes. Diabetes Care. 2007;30:263-9.

Executive Summary Conference Report 49 7th Invited Conference: Examining Medication Errors Associated with Intravenous Insulin

PROCEEDINGS Examining Medication Errors Associated with Intravenous Insulin John P. Santell, MS, RPh, FASHP, Director, Practitioner Programs and Services, United States Pharmacopeia, Rockville, MD

Key points safety experts has identified the following problems with insulin. • Insulin therapy is fraught with patient safety concerns and the potential for medication errors. • A hand-written “U” for units being misread as a number • USP MEDMARX® data show that: − Compared with other medication errors, intravenous (IV) insulin errors are six times • The many types of insulin products more likely to result in harm. (approximately 26) − The majority of IV insulin errors occur during administration by nurses. • Similarity between brand/generic names − The leading causes of IV insulin errors are performance deficit, procedure/protocol and packages (e.g., Humalog Mix 75/25® not followed, and communication. and Humulin 70/30® and Novolog Mix 70/30® and Novolin 70/30®) • The following are frequent problems with IV insulin therapy: − IV pump programming errors • Wide access to insulin as floor stock in most hospitals/health systems − Confusing IV insulin with another IV piggyback − Order incorrectly entered by pharmacy • Many dosing schedules and ‘sliding-scales’ within an individual facility − Patient tampering with IV pump − Staff unfamiliar with glucose protocol leading to inadequate monitoring, • Non-standard, compounded intravenous unclear control orders (IV) solutions and infusion rates − Incomplete documentation USP MEDMARX® Data Findings – • Policies, procedures, or protocols for tight glycemic control should be developed with IV Insulin

consideration of the potential for increasing the opportunities for harmful medication Analysis of data submitted to USP’s errors and adverse drug events. MEDMARX® program during the five-year period from January 2002- December 2006 While the risks and benefits of intensive between people, processes, technology and provides additional information about the insulin therapy (IIT) to manage blood glucose products. severity, origin, types and causes of IV insulin levels continues to be debated, it is impor- Medication errors involving insulin have errors. During this period, 1,298 IV insulin tant to recognize that patient risks associ- been reported for many years, but despite errors were reported with approximately 9.3% ated with insulin use go beyond episodes of this, their common occurrence continues1-5. (n = 121) of these causing harm to the patient. hypo/hyperglycemia. Medication errors that According to the most recent MEDMARX® The average percentage of harm for error can lead to adverse drug events (ADEs) are report published by the United States reports submitted to MEDMARX® has been another issue that should be included in the Pharmacopeia (USP), insulin, in all its dosage approximately 1.5%, indicating that an error discussion about the safety and efficacy of forms, has been the most commonly report- involving an IV insulin product is six times insulin therapy. Medication errors are pre- ed product involved in errors overall and the more likely to result in harm (Categories E-I) ventable events that arise from the interplay leading product involved in harmful errors6. (Table 1) compared to other . Analysis of these reported events by patient

50 Executive Summary Conference Report 7th Invited Conference–Examining Medication Errors Associated with Intravenous Insulin

Previous studies have shown that the error (20.5% general errors vs. 7.4% for insu- Most patient safety experts agree that majority of medication errors are with either lin errors), drug prepared incorrectly (4% vs. there are many causes for medication errors. prescribing or administration of medicines7,8. 6.2%), wrong administration technique (1.4% Among the more than sixty different causes The majority (56%) of errors involving IV insu- vs. 4.9%) and wrong route (1.5% vs. 4%). This tracked in the MEDMARX® program, the lead- lin originate during administration, followed suggests that compared with prescribing IV ing causes associated with IV insulin errors were performance deficit, procedure/protocol Table 1. Severity of IV Insulin Errors not followed, and communication (Table 3). Performance deficit is often cited in combina- Error Category n % tion with procedure/ protocol not followed, Potential Error indicating a logical connection between A 51 3.9 them. Both of these leading causes were cited Intercepted Error more often for IV insulin error events com- B 277 21.3 pared to the general data set (45.4% vs. 39% Non-harmful Error for performance deficit and 31.5% vs. 17.5% C 479 36.9 for procedure/protocol not followed). D 370 28.5 Contributing factors such as distractions, Harmful Error inexperienced staff and workload increase E 108 8.3 are often cited in reports when performance F 9 0.7 deficit or procedure/protocol not followed H 3 0.2 are listed as error causes. This may explain, I 1 0.1 in part, their high ranking among the many Total 1,298 100 possible causes that a reporter may select. For complete definitions of the NCC MERP Error Category Index, see nccmerp.org Other causes reported more frequently with IV insulin errors were monitoring inadequate/ by 17% in dispensing, 14% in transcribing and insulin therapy, there are more problems with lacking, calculation error and improper use of 9% in prescribing activities. This distribution preparing IV insulin infusions, either in the IV pumps. These findings identify areas where differs from that of overall errors in the gener- pharmacy or by nursing staff, with admin- safety improvements are needed. Discussions al MEDMARX® dataset: i.e., errors originating istering the infusion (wrong administration on implementing new policies/procedures in administering activities are approximately technique), e.g., programming and using an for tighter glycemic control should exam- 31%; dispensing, 23%; and transcribing and IV pump and wrong route errors, e.g., admin- ine errors associated with failing to follow prescribing, approximately 22%. Reflecting istering long-acting insulin intravenously. the medication-process stages where errors a originated, nursing staff were most frequently Table 2. Most Frequently Reported Types of Error Involving IV Insulin involved with IV insulin errors (68%), fol- Error Type N % lowed by pharmacy staff (22%) and prescrib- Wrong Dose 523 41.0 ers (9%). Omission Error 278 21.8 The three most frequently reported types Unauthorized/Wrong Drug 119 9.3 of error were wrong dose, omission and unau- Prescribing Error 94 7.4 thorized/wrong drug, which together com- Drug Prepared Incorrectly 79 6.2 prised nearly 70% of all error-type selections Wrong Administration Technique 63 4.9 (Table 2). These error types are also among Wrong Time 54 4.2 the leading types of error in the general Wrong Route 51 4.0 MEDMARX® data set. There were, however, Extra Dose 42 3.3 differences between the general data set Wrong Patient 41 3.2 and the sub-set of IV insulin errors for several USP’s MEDMARX® program tracks 14 different types of error. other types of errors, including prescribing Only the 10 most frequently reported involving IV insulin are shown.

Executive Summary Conference Report 51 7th Invited Conference–Examining Medication Errors Associated with Intravenous Insulin policies/procedures and insufficient patient clear prior totals and to enter the new volume • Patient tampering with IV pump causing monitoring to avoid introducing new error that was to be infused. Shortly after the new an increased infusion rate opportunities. bag was hung, a nurse noticed that the infu- • Staff unfamiliar with glucose protocol sion pump was incorrectly set at 150mL (i.e., Selected Insulin Error Reports leading to inadequate monitoring, unclear 150 units) per hour. The infusion was stopped control orders Case #1: An insulin infusion was ordered and the patient was given orange juice and for an ICU patient. The infusion was started closely monitored for the next three hours. • Incomplete documentation on medication at the wrong rate with subsequent bolus and If the total volume of the bag (100mL) had administration record leading to unclear rate changes not administered as ordered by been infused at the rate of 150mL/hour, it or omitted rate information, when infu- the physician. Fasting blood sugar was 27mg/ would have taken only 40 minutes for the sion started, etc dL, and the patient was found unresponsive patient to receive 100 units of insulin, poten- • General IV pump programming errors and diaphoretic. Dextrose 50% IV was ordered tially causing irreversible brain damage and/ and administered and the patient remained in or death from cerebral edema and insulin Conclusion the ICU for a prolonged period of time. shock. Insulin therapy is fraught with safety con- Case #2: An insulin infusion was ordered Common Error Scenarios cerns and the potential for medication errors. and started pre-operatively on a patient under- Data submitted to USP’s MEDMARX® program A review of several hundred reported error going kidney transplant. Post-operatively, the can help identify where safety risks exist and events identified the following frequently patient was transferred to the ICU without how current practices contribute to error occurring problems: the insulin drip. After this was discovered, it events. Any discussion of implementing new was determined the patient’s blood glucose • Incorrect infusion rates (generally by a policies, procedures or protocols for tight was 443 mg/dL and significant electrolyte factor of 10) as a result of incorrectly pro- glycemic control should proactively evaluate abnormalities. Dialysis was reinstituted on the gramming the IV pump (e.g., 60 units/hr vs their potential for increasing the opportuni- patient, who also required a lengthened ICU 6 units/hr) ties for medication errors and ADEs. stay. • Mix-ups with another IV piggyback (e.g., References Case #3: A patient with diabetes in the ICU anesthesiologist infusing insulin thinking 1. Cohen MR. Medication errors. Check and double- was receiving an IV infusion of regular insulin it was the antibiotic) check all insulin doses. Nursing 1983;13(4):32. 1unit/mL at a rate of 10 units/hour titrated 2. Cohen MR. Lantus or lente insulin? The confusion • Order incorrectly entered by pharmacy builds. Nursing 2003;33(9):12. per sliding scale. After changing to a new bag leading to incorrect concentration pre- of insulin, the IV pump was reset manually to 3. Abbreviations Will Get U In Trouble. ISMP, 2007. pared and infused (Accessed October 1, at .)

4. Insulin Errors: A Common Problem. 2003. (Accessed October 1, 2007, at http://www.usp.org/pdf/EN/ Table 3. Most Frequently Reported Causes of Error Involving IV Insulin patientSafety/capsLink2003-07-01.pdf.)

Cause of Error n % 5. Hicks R, Santell JP, Cousins DD, et al. MEDMARX® Performance/ Human Deficit 583 45.4 5th Anniversary Data Report: A Chartbook of 2003 Findings and Trends 1999-2003. Rockville: Procedure / Protocol not followed 405 31.5 USP Center for the Advancement of Patient Communication 172 13.4 Safety;2004. Knowledge deficit 159 12.4 6. Hicks R, Becker SC, Cousins DD. MEDMARX® Data Computer entry 140 10.9 Report: A Chartbook of Medication Error Findings from the Perioperative Settings from 1998-2005. Documentation 113 8.8 Rockville: USP Center for the Advancement of Monitoring inadequate / lacking 111 8.6 Patient Safety;2006. Calculation error 109 8.5 7. Bates DW, Cullen DJ, Laird N, et al. Incidence of Pump, improper use 104 8.1 adverse drug events and potential adverse drug events. Implications for prevention. ADE Prevention Transcription inaccurate/missing 100 7.8 Study Group. J Amer Med Assn 1995;274(1):29-34.

USP’s MEDMARX® program tracks 67 different types of error. 8. Leape LL, Bates DW, Cullen DJ, et al. Systems analy- Only the 10 most frequently reported involving IV insulin are shown. sis of adverse drug events. ADE Prevention Study Group. J Amer Med Assn 1995;274(1):35-43.

52 Executive Summary Conference Report 7th Invited Conference: The Portland-Vancouver Regional Inpatient Glycemic Control Collaborative

PROCEEDINGS The Portland-Vancouver Regional Inpatient Glycemic Control Collaborative Chris Hogness, MD, MPH, Southwest Washington Medical Center, Vancouver, WA

Key points Center for Patient Safety and Quality collabor- ative effort to reduce central-line-associated • A regional focus on providing optimal inpatient care to diabetic patients is clearly needed. blood stream infections. • State collaboratives supporting chronic disease management, including diabetes, in the Perhaps less commonly, local or regional outpatient setting have become increasingly common over the past decade. health care delivery organizations are directly • Locally based, inpatient quality collaboratives, in which hospitals directly competing bringing themselves together to share qual- for market share agree to cooperate and transparently share their work in quality ity improvement work. The announcement improvement, are an emerging phenomenon. in May 2007 that Adventist, Wellmont and

• In a regional inpatient glycemic control collaborative convened by Southwest Washington Novant health systems were launching a col- Medical Center (SWMC) in September 2006, participating hospitals share glucometric data laborative effort to reduce medical error by to assist members in identifying and learning from best practices and to stimulate regional “creating metrics and identifying best prac- improvements in inpatient glycemic control. tices that can serve as a template for promot- ing patient safety at hospitals nationwide” is • The work of a task force convened by the Society of (SHM) to develop one example of what may become a more practical recommendations for glucometrics in the hospital has been helpful. widespread trend2.

• The open sharing of different approaches to inpatient glycemic control has benefited all There is much to gain and little to lose for institutions in the collaborative. competing hospitals to collaborate in quality improvement work. Hospitals who are fur- A focus on providing optimal inpatient State Collaborative on Diabetes and thest along in developing a particular quality care to diabetic patients in our region is clear- Cardiovascular Disease from 1999 through improvement program may vault themselves ly needed. From 1994 to 2004 the percentage 20051. into a position of regional quality leadership of adults with diabetes in Washington State in convening a collaborative of local hospitals Locally based, inpatient quality collabora- increased from 4% to 6%. In 2002, diabetes- to share their work. Hospitals less far along tives, in which hospitals directly competing related hospitalizations in Washington State the path can accelerate their progress by par- for market share agree to cooperate and cost $1.1 billion. One of the more important ticipating in a regional quality improvement share their work in quality improvement, are sequellae of diabetes is cardiovascular dis- collaborative and learning from the work of an emerging phenomenon. In some cases ease, which was responsible for nearly 4 out peer organizations. Ultimately the patients this work has been coordinated by an out- of 10 hospitalizations in Washington State in in the region benefit—and it is sometimes side intermediary (such as a state health 20021. worth emphasizing that one never knows department, academic institution, or qual- in which hospital one may be personally State collaboratives supporting chronic ity improvement advocacy group), either a patient (including those of competing disease management, including diabetes, in nationally, as in the Institute for Healthcare organizations). This is what gives the quality the outpatient setting have become increas- Improvement’s 100,000 Lives and 5 Million improvement dictum to “steal shamelessly ingly common over the past decade. For Lives Campaigns, or locally, as in the Michigan and share senselessly” its very rational foun- example, more than 100 outpatient health- Health and Hospital Association’s Keystone dation. care facilities participated in the Washington

Executive Summary Conference Report 53 7th Invited Conference: The Portland-Vancouver Regional Inpatient Glycemic Control Collaborative

The Portland-Vancouver Regional glucometer-value denominator (e.g., percent ing, collaborating institutions agreed to inves- Inpatient Glycemic Control of glucometer values in different areas of the tigate each instance of a glucose less than 40 Collaborative hospital within specified ranges) as opposed mg/dL and to report back to the group on this to a patient-denominator (e.g., percent of process at our fall meeting. We feel that meet- Following a one- to two-year period of patient-monitored days for which all or all ing at three-month intervals works well for us intense focus on inpatient glycemic con- but one values are in a specified range). Some at this time. trol, Southwest Washington Medical Center hospitals were in the initial stages of getting (SWMC) convened a regional inpatient gly- We have done this work without budgeted access to their data or had not yet begun to cemic control collaborative initially attended financial support. This may be a limiting factor do so. by nine hospitals in the Portland-Vancouver as the group considers taking on more struc- area in September 2006. Hospital person- At our June 2007 meeting, we agreed on tured, regional quality improvement activities. nel who attended the first meeting included the following metrics with which to compare One possibility for collaboratives to obtain physicians (hospitalists, surgeons and endo- data: funding may be to approach insurance com- crinologists), nurse managers, diabetic edu- panies or other potential donors to support • Patient-monitored day = a day on which a cators and pharmacists from Oregon Health the development of a more well-resourced, patient has at least two glucometer values and Sciences University, Kaiser Sunnyside, the formalized collaborative focused on assisting recorded. Portland VA hospital, Providence Portland, all hospitals in the region to achieve optimal Providence St. Vincent, Legacy Emanuel, • Patient-controlled day for ward patients = institutional glycemic control. Legacy Salmon Creek, Adventist and SWMC. a patient-monitored day in which no more National networking with benchmark We met in December 2006, March 2007 and than one glucometer value is outside of a institutions has also been useful to many of June 2007, and plan to continue meeting range of 70-180 mg/dL. us in the collaborative. The range of prac- with every three months for the foreseeable • Patient-controlled day for intensive care tices reported in the SHM Workbook for future. Over this period other hospitals from unit (ICU) patients = a patient-monitored Improvement on inpatient glycemic control the Providence and Legacy systems joined day in which no more than one glucom- has been helpful in this regard4. Other oppor- the group. The participating health system eter value is outside of a range of 70-150 tunities for national collaboration may emerge farthest from the Portland-Vancouver metro- mg/dL. from a national training conference for quality politan area is Asante Health Systems in Bend, In addition, for our ICU patients, many or improvement teams working on inpatient gly- Oregon. most of who are on insulin infusions, often cemic control hosted by SWMC and Oregon All participating hospitals are committed with hourly glucose checks, we will also com- Health and Sciences University in Vancouver, to transparently sharing glucometric data for pare percent of glucometer values (glucom- Washington in October 2007. This program purposes of inter-hospital comparison across eter denominator as opposed to patient-mon- targets the range of people involved in inpa- the region. We believe that this will assist us itored day denominator) within the range of tient glycemic control, including pharmacists, in identifying and learning from best prac- 70-150 mg/dL. This issue is addressed well in CDEs, nurses, quality improvement personnel, tices, stimulating regional improvements in the SHM document referenced above3. information systems staff, hospitalists, endo- inpatient glycemic control. The work of a task crinologists, surgeons, critical care physicians, As a collaborative, openly sharing our force convened by the Society of Hospital hospital administrators and others. different approaches to inpatient glycemic Medicine (SHM) to develop practical recom- control has benefited all participating institu- References mendations for glucometrics in the hospital 3 tions. SWMC has shared the process we used 1. Washington State Diabetes Collaborative Website: has been helpful . http://www.doh.wa.gov/cfh/WSC/default.htm to develop a pharmacist/clinical diabetic edu- accessed 7/6/2007. Participating hospitals came to the col- cator (CDE)-centered glycemic control team 2. Health Care Advisory Board newsletter, May 2007. laborative with a wide range of experience with other institutions thinking of developing in inpatient glycemic control and varying 3. Society of Hospital Medicine Workbook for similar teams. We also have shared our work in Improvement titled: Improving Glycemic Control, degrees of initial access to glucometric data. educating nursing in glycemic control and in Preventing Hypoglycemia, and Optimizing Care of the Some, including SWMC, had already exam- developing insulin infusions, e.g., the recently Inpatient with Hyperglycemia and Diabetes; 27-43. http://www.hospitalmedicine.org/AM/Template. ined their hospitals’ glucometrics using a 4 published SWMC protocol . At our June meet- cfm?Section=Quality_Improvement_Resource_

54 Executive Summary Conference Report 7th Invited Conference: The Portland-Vancouver Regional Inpatient Glycemic Control Collaborative

Rooms&Template=/CM/ContentDisplay. cfm&ContentID=11878

4. Hogness C, Finneman L, Chandler P, et al. Effective implementation of a new 6-column insulin infu- sion protocol in the ICU. The Hospitalist 2006 December;10(12):42-4. http://media.wiley.com/ assets/1146/32/TH1206.pdf.

Executive Summary Conference Report 55 7th Invited Conference: Building Transitions from the ICU to the Ward for the Hyperglycemic Patient: One Piece of the Puzzle

PROCEEDINGS Building Transitions from the ICU to the Ward for the Hyperglycemic Patient: One Piece of the Puzzle Greg Maynard, MD, MS, Associate Clinical Professor of Medicine and Chief, University of California at San Diego, Division of Hospital Medicine, San Diego, CA

Key points transitions can be accomplished reliably with the use of a step-by-step method that we are • A rational, step-wise approach can ensure successful transitions from intravenous to adopting at our institution. subcutaneous insulin regimens. Step 1. Calculate the amount of insulin • Some variation of this step-wise approach to transition should be developed into a the patient requires over 24 hours. For standardized protocol for every institution. example, an insulin infusion rate of 3 units/hr • Multiple methods can enhance reliability of protocol use, including: would lead to an insulin requirement of 72

– Educate staff about the protocol. units in 24 hours (Table 1).

– Design the protocol guidance into the orders used at the point of care. Step 2. Decrease this dose to ensure a safety cushion. Just as insulin require- – Make the protocol the “default” method for making the transition. ments increase with increasing physiologic – Automate or delegate special teams to assist in the transition. stress and illness, insulin requirements can be reduced as the patient improves. One meth- – Measure results obtained after protocol implementation to further refine and od to insert a safety cushion is to simply take revise the protocol. the most recent stable-infusion rate times 20 • The Society of Hospital Medicine has a web-based “Glycemic Control Resource Room”1 (instead of 24). In the above example, this designed to assist Glycemic Control teams with all aspects of building and effectively yields an estimate of 60 units. Note that while implementing a series of linked glycemic control/hypoglycemia prevention protocols. you have introduced a safety cushion, your estimate of insulin requirement is still twice Case Study Mismanagement of transitions like this the insulin dose RW took as an outpatient. may lead to prolonged hyperglycemia, hypo- RW is a 60-year-old hyperglycemic male Step 3. Assess whether the insulin glycemia or, in the case of an insulin-depen- in the intensive care unit (ICU) with multiple requirement was furnishing basal insulin dent patient deprived of basal insulin, dia- medical problems. He is on full-dose, con- needs, or basal and nutritional needs. In betic ketoacidosis (DKA). Safe and effective tinuous tube feedings at 40 mL/hr, and he has been on intravenous (IV) insulin infusion at a Table 1. Step-wise Approach: Transitioning from IV to SC Insulin stable rate of 3 units/hr over the last 6 hours. • Calculate how much IV insulin the patient has been requiring. Modify down for Prior to falling ill, his baseline A1c was 8.7%, safety cushion. and his outpatient regimen was 30 units of 70/30 mixed insulin and an oral hypoglyce- • Was this insulin supplying basal requirements, or basal and nutritional mic agent. The problem: maintaining good requirements? Translate into the subcutaneous regimen. glycemic control safely as RW transitions to a • Consider any nutritional changes that may be implemented at the time of the general medical ward that does not support transition off of the drip. IV insulin infusion, while continuing enteral tube feeding. • Make sure SC insulin is given before discontinuation of the IV insulin.

56 Executive Summary Conference Report 7th Invited Conference: Building Transitions from the ICU to the Ward for the Hyperglycemic Patient: One Piece of the Puzzle this case, the insulin was furnishing both basal and nutritional needs. Table 2. Enteral to PO • What if, instead of continuing enteral nutrition on the floor, you opt to stop Step 4. Allow for nutritional changes enteral nutrition and start the patient on a mechanical soft diet? that also may occur during the transition away from infusion insulin. Very often, a • Glargine 30 units = Basal patient’s nutritional intake and plan also may • RAA 10 units q AC = Nutritional/Prandial change at the same time you transition to (If you expect them to eat a full meal!) subcutaneous (SC) insulin. In the case of RW, however, the patient is to continue on the • If PO intake suspect at first, use CHO counting, or empirically reduce nutritional same full-dose enteral tube feedings. RAA dose and give the dose just after the meal instead of just before the meal.

Step 5. Apply this knowledge to create • Correction dose RAA insulin also needed. a SQ basal-bolus insulin regimen for the their full nutritional requirement, so the treat- The Society of Hospital Medicine has built patient. For this patient, 60 units represents ment team needs to take this into account. a web-based “Glycemic Control Resource the total daily dose (TDD) of insulin, the dose Solutions may include carbohydrate counting Room”1 designed to assist Glycemic Control of insulin required for a patient taking in a for the nutritional component, empiric reduc- teams with all aspects of building and effec- full nutritional load. Using the 50:50 rule, half tion of the nutritional insulin requirement, tively implementing a series of linked glyce- of this TDD (or 30 units) should generally or withholding the nutritional insulin dose mic control/hypoglycemia prevention proto- be supplied in the hospital as a long-acting, until just after the patient eats, to allow for cols. This invaluable resource takes a broad peakless, basal insulin such as insulin glargine a better estimate of nutritional intake and view of the inpatient hyperglycemia /diabetes or the more recently available insulin detemir. the expected carbohydrate-induced glycemic issues and addresses everything from how to For a patient on continuous enteral tube excursion. feeding, the remaining 30 units can be sup- plied in split doses of rapid-acting analogue An inpatient with hyperglycemia fre- Table 3. Performance Improvement insulin (RAA-I) (5 units q 4 hours) or, alterna- quently makes many transitions across sev- Principles tively, as split doses of regular insulin (7 units eral care-giving teams and locations during • Obtain institutional buy -in/ q 6 hours). a single inpatient stay, and the transition support. from insulin infusion to a SC regimen is but Step 6. Make sure the SC insulin is given • Form a multidisciplinary team. one piece of the puzzle. How can we improve PRIOR to stopping the insulin infusion! the reliability of care provided for patients • Set achievable but aggressive Build this premise into your order sets and in such a rapidly changing environment, as goals. protocols. the patient undergoes changes in nutritional • Use metrics that are reliable, Correction-dose regular insulin could be intake, insulin requirements and varied access practical. given along with the nutritional doses as to insulin infusion? The answer generally lies needed. in standardizing the care and monitoring • Pay attention to ease of use of process via a series of institution-specific pro- orders/protocols. What if RW had indicated he was ready to tocols. These protocols should be linked and try an oral diet, and you wanted to discon- • Identify "best practice"; reference each other in a consistent way. The tinue enteral tube feedings at the same time standardize choices. insulin infusion protocol, for example, should you transferred him to the wards (Table 2)? • Layer interventions and methods contain a hypoglycemia protocol, standards Step # 4 would mandate that you take this to enhance reliability. for monitoring glucose values, guidance for into account. His basal insulin dose would transitioning the patient to SQ insulin and a remain the same, but his nutritional require- • Fail faster: ongoing feedback and direct reference to an institution-specific SC ment estimate would be better delivered as refinement. insulin regimen. The guidance from the pro- a RAA-I in three divided doses with meals • Education: always necessary, but tocols should be designed into order sets and (10 units of RAA-I q AC). Patients just start- rarely sufficient if used alone. workflow. ing a diet are not always reliable in taking in

Executive Summary Conference Report 57 7th Invited Conference: Building Transitions from the ICU to the Ward for the Hyperglycemic Patient: One Piece of the Puzzle build an effective team, to garnering institu- the structured order sets and documentation computerized, closed-loop insulin infu- tional support and building the business case forms. At Level 3, you can expect measurable sion protocols, autocalculation of the for glycemic control efforts, to enhancing improvement, and reliability increases to 65% ideal suggested basal–bolus SC insulin reliable care via effective implementation of to 85%. regimen has been demonstrated with protocols and structured order sets. The prin- good effect. At Level 4, a variety of other performance- ciples of performance improvement (Table 3) improvement and high-reliability design • Measurement of results obtained after are illustrated throughout with examples that techniques are layered onto Level III efforts protocol implementation to further refine apply to inpatient glycemic control issues. including education. and revise the protocol, and to provide “Hierarchy of Reliability” feedback to the providers. Examples would include: An important concept that has emerged At Level V, real-time data are used to moni- • Making the protocol the “default” method from this work is the “Hierarchy of Reliability," tor the transition process. Individual patients for making the transition. i.e., the success in ensuring that best practice who do not attain an optimal transition is followed is a function of how far up this − For example, orders to discontinue are identified in near real-time, and reme- hierarchy your team has gone (Table 4). insulin infusion would require the use dial action occurs concurrently. Furthermore, of an IV to SQ insulin order set, which these outliers are scrutinized carefully to fur- At Level 1 Reliability, there is no standard- would lead the provider through the ther refine the protocol and the manner in ization, and achieving a reliable transition logical steps of the transition protocol. which it is implemented. Reliability of over from IV to SQ insulin would depend entirely Providers may “opt out” of the sug- 90% can be attained at this level. on the knowledge and vigilance of individual gested doses and regimens called for practitioners. At Level 2, a protocol or instruc- Summary in your protocol, but they must go tions for making the transition may exist, but through your protocol steps and explic- It is very common for a patient to transi- they are not well integrated into the flow itly choose to opt out, since your proto- tion from IV insulin infusion to an SQ insu- of work, and thus guidance remains largely col is the default method. lin regimen. Unfortunately, this transition is unavailable for the major of patients. At Level often mismanaged, with resultant loss of gly- 3, we start to see some serious improvement • Automation or delegation of a special cemic control or avoidable hypoglycemia. in reliability. Protocols exist for IV and SQ team to assist in the transition. Improvement teams need to devise best- insulin. Furthermore, the guidance provided − Going one step further, nurses, phar- practice protocols for insulin management by the protocols (for calculating insulin doses, macists, or some other specially trained and transitions of care. The guidance derived choosing regimens for different nutritional team would be empowered to activate from these protocols should be incorporated scenarios, glucose monitoring, when to call and carry out the protocol, assisting the into the workflow, usually as prompts and a physician, hypoglycemia care and desirable provider in making this transition in a guidance built into structured order sets and glycemic targets, etc) is available to front- standardized way. In institutions with documentation forms. Other performance- line workers in the form of cues built into improvement techniques and high-reliability design measures can dramatically improve Table 4. Hierarchy of Reliability the chances that a patient will benefit from Level Predicted the best practices available. Extensive guid- success rate ance for inpatient glycemic control teams 1 "State of Nature" −no standardization 40% are available at no charge from the Society of 2 Decision support exists but not linked to orderwriting, or Hospital Medicine Glycemic Control Resource 1 prompts withing orders but no decision support 50% Room .

3 Protocol well -integrated (into orders at point of care) 65-85% Reference 1. www.hospitalmedicine.org, follow Quality/Safety 4 Protocol enyhanced (by other QI and high reliability strategies) 90% links. 5 Oversights identified and addressed in real time 95+%

58 Executive Summary Conference Report 7th Invited Conference: Glucose Control and (Continuous) Glucose Monitoring in Critical Illness

PROCEEDINGS Glucose Control and (Continuous) Glucose Monitoring in Critical Illness Christophe De Block, MD Ph, Department of Diabetology, Antwerp University Hospital, and the Antwerp Metabolic Research Unit, Antwerp University, Belgium

Key points Aggressive treatment of hyperglycemia with insulin may, however, be limited by an • Stress hyperglycemia occurs in 50% to 80% of critically ill patients and has become an increased risk of hypoglycemia. Recognition important therapeutic target in critically ill patients. of hypoglycemia in a patient who is receiv- • Aggressive treatment of hyperglycemia with insulin may be limited by an increased risk ing sedatives and analgesics and/or neu- of hypoglycemia. romuscular blocking agents in the ICU is

• A factor that may contribute to hypoglycemia is insufficient frequency of glucose problematic, since the hypoglycemic state monitoring. may go unrecognized for a critical period before treatment. The German VISEP trial was • Recent evidence suggests that continuous monitoring of glucose levels may help to signal stopped prematurely because no differences glycemic excursions and eventually to optimize titration of insulin therapy in the ICU. in mortality and frequent hypoglycemia were found in the intensive insulin therapy arm4. Stress hyperglycemia occurs in 50% to Clinical Impact of Hyperglycemia and Interestingly, no adverse clinical outcome 80% of critically ill patients and has become Benefits of Insulin Therapy in the ICU associated with hypoglycemia was reported an important therapeutic target in critically The studies of Van den Berghe et al. in any study. One factor that may contribute ill patients. Hyperglycemia is associated with showed that ITT to achieve normoglycemia to hypoglycemia is insufficient frequency of adverse outcome, not only after myocardi- reduced the incidence of acute renal failure glucose monitoring. al infarction, coronary artery bypass graft, and accelerated discharge from the ICU and stroke, and trauma, but also in the intensive Management of Hyperglycemia the hospital1,2. In contrast to the surgical ICU, care unit (ICU). Intensive insulin therapy (IIT) To achieve strict glycemic control in in the medical ICU, in-hospital mortality was to maintain strict normoglycemia has been critically ill patients, the implementation of reduced only among patients staying for shown to reduce morbidity and the length of insulin infusion protocols based on frequent three days or longer. Most likely, the benefi- stay in the ICU and hospital. IIT may be associ- glucose monitoring is required. First, precipi- cial effects of IIT require time to be realized, ated with an increased risk of hypoglycemia. tating causes of stress hyperglycemia should because insulin therapy is aimed not at cur- The inherent clinical perturbations of criti- be identified and treated. Second, the patient ing disease but at preventing complications. cally ill patients result in frequent changes in population in which insulin therapy may be The impact of hyperglycemia on mortality insulin needs that require extensive efforts of benefit should be clearly defined. Third, varies depending on the background severity from nurses. Current insulin titration is based consensus should be obtained regarding the and type of illness. upon discontinuous glucose measurements, glycemia target level. Fourth, glycemic excur- which may miss rapid changes in glycemia. Whether attainment of strict normoglyce- sions should be carefully monitored, prefer- Continuous monitoring of glucose levels may mia or administration of insulin is the decisive ably on a continuous base, and a validated, help to signal glycemic excursions and even- factor that explains the clinical benefits of IIT easily implementable insulin infusion proto- tually to optimize titration of insulin therapy is open for discussion. A linear correlation col should be provided. in the ICU. between the degree of hyperglycemia and What level of glycemia should be the the risk of death, which persisted after cor- target? The Society for Critical Care Medicine rection for insulin dose and severity of illness, 3 recently recommended maintaining a blood has been demonstrated . glucose level of < 150 mg/dL in patients with

Executive Summary Conference Report 59 7th Invited Conference: Glucose Control and (Continuous) Glucose Monitoring in Critical Illness severe sepsis5. The target glycemia in the meals and do not have signs of infections, need) did not worsen accuracy15. Instead, Leuven studies was 80-110 mg/dL1,2. Krinsley provided that an infusion of ≤ 3 units/hr is such variables would affect the process of observed the lowest in-hospital mortality in sufficient to maintain normoglycemia. subcutaneous glucose recovery, resulting in critically ill patients with a mean blood glu- a calibration issue, rather than in a sensor Current Methods to Evaluate Glycemic cose of 80-99 mg/dL6, whereas Finney et al. performance issue. A lag time of < 10 min is Control in the ICU observed a mortality benefit with a specula- clinically acceptable since online adjustment tive upper limit of 145 mg/dL in cardiotho- Glucose indices. To objectively assess glu- of insulin dose occurs every hour and should racic surgery patients7. Data are difficult to cose control in acutely ill patients, the magni- be based on immediate detection by CGM of interpret because of the diverse clinical set- tude and duration of hyperglycemia should unacceptable rates of change (> 25 mg/dL/ tings, the varying methods of insulin admin- be evaluated. Indices of glucose regulation hour). istration, and different targets of glycemic that have been used in acutely ill patients CGM accuracy improves with an increasing control. The American Diabetes Association are admission glucose regulation, maximum number of calibration points15,16. Calibration and the American College of Endocrinology and mean glucose. However, these indices should also be performed in times of glucose sta- have published guidelines recommending are based on either a single measurement or bility (< 10% change in glucose over 9 minutes in-hospital IIT to maintain preprandial blood on a subset of measurements, and thus are for the GlucoDay®, and a rate of change in glu- glucose levels at ≤ 110 mg/dL in critical not indicative of overall glycemia and give no cose < 2 mg/dL/min for the CGMS and Freestyle care patients8. Postprandial glycemia should indication of blood glucose variability. Navigator® Continuous Glucose Monitor). be kept < 180 mg/dL in any hospitalized Continuous Glucose Monitoring Only a few studies used CGM in critically ill patient. Systems (CGMS) in the ICU patients15,17-20. In a pilot study using continu- Glucose control and monitoring in the ICU. Rationale for use. A continuous display of ous glucose monitoring (GlucoDay®) in the Insulin requirements vary widely in patients blood glucose levels seems to be essential for medical ICU, rapid changes in glycemia were depending on insulin sensitivity, caloric optimal titration of insulin therapy in the ICU. noted immediately, whereas these were noted intake, the nature and fluctuating severity of Besides giving an indication of overall glyce- much later (~1-3 hours) when only intermit- the underlying illness and the administration mia, it shows the variability and fluctuations tent blood glucose determinations were used. of medications. The analysis of the correct of blood glucose concentrations which may Hyperglycemia was present in 74% of medical amount of insulin to be administered requires affect patient outcome. ICU patients, and target glycemia (80-110 mg/ a relatively high degree of skill, and frequent dL) was reached only 22% of the time, which expert assessment will be needed as the clini- Technical requirements and validation for reveals the inadequacy of current insulin pro- cal situation changes9,10. the ICU. Data on the reliability of CGMS in diabetic patients cannot be automatically tocols to optimize glycemia and suggests the A standardized protocol that prompts applied to a different situation such as the potential of an accurate continuous glucose 15 users to initiate an insulin drip for critically ill ICU, where many variables can affect CGMS monitoring system in this setting . Similar patients to maintain normoglycemia should performance (edema, hypotension, vasoac- results were reported by Goldberg et al. using be developed. Goldberg et al. proposed a tive drugs, etc.). Necessary requirements for the CGMS device (Medtronic MiniMed Inc., 17 comprehensive, validated insulin infusion a CGMS include immediate availability of Northridge, CA) in a medical ICU . The pres- 9 protocol (IIP) that took into account the the measurement result, a high frequency of ence of edema or hypotension, and the use of current and previous blood glucose level to measurements, and fast sensor signal stabil- vasopressors did not affect sensor accuracy. calculate the rate of glycemic change, and ity after application and over time11. There were no serious adverse events report- the current insulin infusion rate. IIPs add ed during the use of the CGMS. Vriesendorp significantly to the work of managing ICU Current CGMS measure glucose in the et al., who used the GlucoDay® device during patients. Another obstacle to implementing interstitial fluid. Changes of glucose concen- and after surgery, encountered a high techni- IIPs is the fear of hypoglycemia in patients trations in interstitial fluid lag behind those cal failure rate18, which was mainly attributed being treated. The transition from intrave- in the blood by a few seconds to up to 15 to breaking of the microdialysis fiber during nous (IV) to subcutaneous (SC) insulin should minutes. The lag time seems to be consistent, transfer from the surgical bench to the ICU be an integral part of any insulin infusion irrespective of increments/decrements in gly- bed. Using CGMS, Baird et al. observed that 12-14 protocol. This transition can be considered cemia and insulin levels . In the ICU, the acute and final infarct volume change and in extubated patients who are taking regular hemodynamic alterations we encountered outcome were negatively affected in patients (hypotension, shock, vasopressor/inotropic with mean blood glucose levels ≥126 mg/ dL19. 60 Executive Summary Conference Report 7th Invited Conference: Glucose Control and (Continuous) Glucose Monitoring in Critical Illness

Chee et al. conducted a study to determine to develop a low-risk monitoring and control 8. ACE/ADA Task Force on Inpatient Diabetes. American College of Endocrinology and American if the CGMS device could be used in real-time system that allows health care providers to Diabetes Association consensus statement on inpa- to control glycemia in five critically ill patients. maintain strict glycemic control in ICUs using tient diabetes and glycemic control. Diabetes Care A closed-loop control system was constructed a SC-IV closed-loop system. 2006;29:1955-62. to use CGMS in a real-time manner, coupled 9. Goldberg PA, Siegel M, Sherwin RS, et al. Conclusion Implementation of a safe and effective insulin with a proportional integral (PI) control algo- infusion protocol in a medical intensive care unit. rithm based on a sliding scale approach for Stress hyperglycemia is highly prevalent Diabetes Care 2004; 27:461-67. in the ICU and is associated with adverse automatic IV insulin infusion. They concluded 10. Clement S, Braithwaite SS, Magee MF, et al., on behalf that the automatic sliding scale approach of outcome. IIT to achieve normoglycemia may of the Diabetes in Hospitals Writing Committee. Management of diabetes and hyperglycemia in hos- closed-loop glycemic control is feasible in reduce mortality and morbidity. Identification pitals. Diabetes Care 2004;27:553-91. ICU patients, but the algorithm needs refine- of the hyperglycemic patient with timely, 11. Koschinsky T, Heinemann L. Sensors for glucose cost-effective, and comprehensive evaluation ment and the sensor accuracy needs to be monitoring: technical and clinical aspects. Diabetes improved20. and risk stratification may facilitate appropri- Metab Res Rev 2001;17:113-23. ate implementation of therapies and proce- 12. Steil GM, Rebrin K, Hariri F, et al. Interstitial fluid How to use data obtained with CGM? The dures that may enhance outcome. Current glucose dynamics during insulin-induced hypogly- vast amount of data collected during CGM caemia. Diabetologia 2005;48:1833-40. insulin titration is based upon discontinuous must be presented in an understandable way glucose measurements, which may miss rapid 13. Boyne MS, Silver DM, Kaplan J, et al. Timing of so that the physician can interpret it ade- changes in interstitial and venous blood glucose changes in glycemia. Recent evidence sug- measured with a continuous subcutaneous glucose quately. First, the CGM System should display gests that continuous monitoring of glucose sensor. Diabetes 2003;52:2790-4. the actual (real-time) glucose measurement, levels may help to signal glycemic excursions 14. Rossetti P, Porcellati F, Fanelli CG, et al. Evaluation and a warning alarm should be available if the and eventually to optimize titration of insulin of the accuracy of a microdialysis-based glucose actual glucose value is outside a predefined sensor during insulin-induced hypoglycaemia, its therapy in the ICU. recovery, and post-hypoglycemic hyperglycemia in target value. Second, CGM provides trend humans. Diabetes Technol Ther 2006;8:326-37. information, making it possible to predict References 15. De Block C, Manuel-y-Keenoy B, Van Gaal L, Rogiers the course of glucose changes for over lon- 1. Van den Berghe G, Wouters P, Weekers F, et al. P. Intensive insulin therapy in the intensive care ger time periods. Third, CGM data not only Intensive insulin therapy in critically ill patients. N unit. Assessment by continuous glucose monitor- Engl J Med 2001; 345:1359–67. ing. Diabetes Care 2006;29:1750-6. highlight the cumulative hyper- and hypo- 2. Van den Berghe G, Wilmer A, Hermans G, et al. glycemia, but also show glucose fluctuations. 16. Diabetes in Research Children Network (DirectNet) Intensive insulin therapy in the medical ICU. N Engl Study Group. Evaluation of factors affecting CGMS J Med 2006; 354:449–61. The use of CGMS in critically ill patients calibration. Diabetes Technol Ther 2006; 8:318-25. looks promising. If further developed as a 3. Van den Berghe G, Wouters PJ, Bouillon R, et al. 17. Goldberg PA, Siegel M, Russell RR, et al. Experience Outcome benefit of intensive insulin therapy in the with the continuous glucose monitoring system in "real-time" glucose sensor, CGMS technology critically ill: insulin dose versus glycemic control. a medical intensive care unit. Diabetes Technol Ther could ultimately prove clinically useful in the Crit Care Med 2003; 31:359-66. 2004;6:339-47.

ICU by providing alarm signals for impend- 4. Brunkhorst FM, Kuhnt E, Engel C, et al., and the 18. Vriesendorp T, DeVries J, Holleman F, et al. The use ing glycemic excursions, rendering IIT easier German Competence Network Sepsis (SepNet). of two continuous glucose sensors during and after Intensive insulin therapy in patients with severe sep- surgery. Diabetes Technol Ther 2005;7:315-22. and safer. The development of a closed-loop sis and septic shock is associated with an increased control system, with an accurate CGMS and rate of hypoglycaemia – results from a randomized 19. Baird TA, Parsons MW, Phanh T, et al. Persistent post- multicenter study (VISEP). Infection 2005; 33(Suppl stroke hyperglycemia is independently associated computer-assisted titration of insulin dose 1):19. with infarct expansion and worse clinical outcome. based on glucose measurements, could per- Stroke 2003; 34:2208-14. 5. Dellinger RP, Carlet JM, Masur H, et al. Surviving mit tight glycemic control without increas- sepsis campaign guidelines for management of 20. Chee F, Fernando T, van Heerden PV. Closed-loop ing the workload of the nursing staff. Plank severe sepsis and septic shock. Crit Care Med 2004; glucose control in critically ill patients using con- 32:858-73. tinuous glucose monitoring system (CGMS) in real et al. observed that, compared with rou- time. IEEE Trans Inf Technol Biomed 2003;7:43-53. tine protocols, treatment according to a fully 6. Krinsley JS. Association between hyperglycemia and increased hospital mortality in a heterogenous 21. Plank J, Blaha J, Cordingley J, et al. Multicentric, ran- automated model-predictive-control (MPC) population of critically ill patients. Mayo Clin Proc domized, controlled trial to evaluate blood glucose algorithm resulted in a significantly higher 2003;78:1471-8. control by the model predictive control algorithm versus routine glucose management protocols in percentage of time within the target glycemic 7. Finney SJ, Zekveld C, Elia A, Evans TW. Glucose con- intensive care unit patients. Diabetes Care 2006; range (80-110 mg/dL)21. The European com- trol and mortality in critically ill patients. J Amer Med 29:271-6. Assn 2003; 290:2041-7. munity-funded CLINICIP (Closed Loop Insulin Infusion for Critically Ill Patients) project aims

Executive Summary Conference Report 61 7th Invited Conference: Glucose Sensor Technology: Current State and Future Trends

PROCEEDINGS Glucose Sensor Technology: Current State and Future Trends Timothy S. Bailey, MD, FACE, FACP, Advanced Metabolic Care and Research, Escondido, CA

Key points branes, these sensors are now available com- mercially to assist in the outpatient manage- • General agreement now exists that normalization of blood glucose levels through insulin ment of diabetes. administration is desirable for all hyperglycemic patients. Current CGMS require initial calibration • Delivering insulin safely (to avoid hypoglycemia) and effectively (to achieve the increasingly to an external reference, either a laboratory strict glucose targets) requires meticulous monitoring of blood glucose levels, and glucose value or, more commonly, a point-of-care sensors are a critical component of an effective glucose control system. (POC) glucose-monitor result. If the latter • Compared with standard, intermittent glucose checks, continuous glucose monitoring reference is used for calibration, the lower provides far more frequent checks, which can more promptly detect important trends in precision of POC monitors may contribute blood glucose levels. to CGMS inaccuracy. In addition, most elec-

• The Food and Drug Administration limits current glucose-sensor technology to adjunctive trodes exhibit drift over time and require re- use, because of a time-lag effect caused by various factors. calibration. Thus, stability is also an important glucose-electrode characteristic. • Other approaches to blood glucose measurement being explored include extracorporeal methods, indwelling venous catheters, and non-invasive optical technologies—all of which Today the most prevalent approach is require further testing before they can be used in critical-care populations. interstitial fluid sensing, done either by a “needle”-type sensor or a microdialysis cath- • Ultimately, closed-loop, continuous glucose monitoring systems are likely to replace all eter inserted subcutaneously. With these sys- of the labor-intensive protocols currently in use and will be a key factor in improving tems, when glucose values change rapidly, hyperglycemia therapy. there may be differences between sensor val- ues and a reference method. For example, Normalization of blood glucose levels is more frequent monitoring. Hyperglycemic with the microdialysis method, the length now generally considered desirable for all patients present complex clinical challeng- of the catheter and, thus, the time required hyperglycemic patients. Achieving this goal es, and less-frequent monitoring can fail to for the fluid to arrive at the sensor creates a frequently requires the use of insulin, par- detect important trends promptly. delay (usually in minutes) that must be taken ticularly in hospitalized patients. Delivering into account. Device software that averages Current State insulin safely (to avoid hypoglycemia) and values to improve accuracy also may cause effectively (to achieve the increasingly strict In principle, an effective continuous glu- delays. Finally, an “alternate-site-testing-like glucose targets) requires meticulous moni- cose monitoring system (CGMS) uses a glu- effect” has been described. This term refers toring of blood glucose levels. Glucose sen- cose electrode with the following characteris- to a variable time lag that was first reported sors are a critical component of an effective tics: 1) specificity for glucose, 2) sensitivity for with standard, intermittent, POC monitoring, glucose control system. glucose, 3) a favorable signal-to-noise ratio, when sampling sites other than fingertip (e.g., 4) accuracy, and 5) precision. Today, most forearm) were tested. Because of this time-lag Compared with the current standard of glucose electrodes are glucose-oxidase (GO) effect, FDA clearance for current glucose-sen- care that uses intermittent glucose checks, based. Using additional innovative technolo- sor technology is limited to adjunctive use. continuous glucose monitoring offers far gies such as selectively permeable mem-

62 Executive Summary Conference Report 7th Invited Conference: Glucose Sensor Technology: Current State and Future Trends

Future Trends Summary

Several companies have used an extracor- Current developments in glucose moni- poreal approach that withdraws venous blood toring sensors are promising and clinical tri- from the patient for measurement. Earlier als are underway. The future of emerging, techniques (Biostator® glucose-controlled entirely non-invasive technologies, such as insulin-infusion system, Miles Laboratories) optical or spectroscopic systems is still uncer- did not return the blood to the patient and tain; however, other minimally invasive and could lead to anemia. Newer methods return invasive techniques are promising. They have the blood to the patient. the potential to decrease the staff workload inherent in frequent glucose monitoring and, Indwelling venous catheters are also therefore, will be key to enabling nurses to being studied, but these have the potential implement hyperglycemia therapy. Reduced for thrombus, embolism and infection. All of blood glucose variability and rates of hypogly- these more-invasive methods may eliminate cemia that result from more frequent glucose the delays seen with less-invasive monitor- monitoring, in turn, are likely to be associated ing and have the potential to simplify the with better outcomes and greater patient control system algorithms. Entirely non-in- safety. vasive approaches (mostly optical methods utilizing near-infrared spectroscopy) are also being explored. Due to the complexity in dis- tinguishing the glucose signal, progress has been slow.

Most sensor research has involved patients who are not acutely ill. Critically ill patients in the intensive care unit (ICU) are more com- plex. They exhibit hemodynamic instability and variable perfusion, and usually are treated with many medications. For this reason, clini- cal trials must be performed in this population prior to adopting the technology in this set- ting.

Ultimately, glucose sensors will routinely serve as the input systems to closed-loop glu- cose control therapy. Once the control system algorithms have been clinically validated and safety concerns have been addressed, these closed-loop systems are likely to replace the labor-intensive protocols currently in use.

Executive Summary Conference Report 63 7th Invited Conference: Assessing the Accuracy and Confounding Factors in Critical Care Glucose Monitoring

PROCEEDINGS Assessing the Accuracy and Confounding Factors in Critical Care Glucose Monitoring Nam K. Tran, Nicole L. Gentile, Victor J. Abad, Richard F. Louie, and Gerald J. Kost. Point-of-Care Testing Center for Teaching and Research, Department of and Laboratory Medicine, University of California-Davis, School of Medicine

Key points Factors Affecting Accuracy • As the use of point-of-care (POC) glucose monitoring systems (GMS) for tight glycemic Exogenous and endogenous factors control (TGC) becomes more common in critically ill patients, the accuracy of the affect the accuracy of GMS. Endogenous fac- information from instruments is important. tors include specific physiological elements

such as hematocrit, blood pO2, blood pres- • Physiologic factors that can affect accuracy include hematocrit, blood pO2, blood pressure, sure, sample source (venous, arterial, capil- sample source (venous, arterial, capillary), pH, temperature and the blood matrix (water, lary), pH, temperature and the blood matrix lipid, cellular, and protein contents).1 (water, lipid, cellular and protein contents)1. • Other factors affecting accuracy include drug interferences, user error and harsh Exogenous factors include drug interferenc- environmental conditions. es, user error and harsh environmental condi- • There are many methods to evaluate the accuracy of POC glucose meter readings, including tions. In this article we highlight factors that national and international standards, and illustrative and statistical methods to determine if may cause errors in glucose monitoring. a GMS is comparable to a reference instrument. Hematocrit. The amount of red blood cells • Manufacturers must continue to develop measures to adjust for factors that affect accuracy, relative to plasma has an inverse effect on and healthcare professionals should be aware of the limitations of the GMS they use and measured glucose. Dacombe et al. found that evaluate accuracy. increases in hematocrit cause decreases in glucose readings2. Conversely; decreases in • New range-specific analytical methods are needed to assess the accuracy of GMS at hematocrit were found to increase glucose hypoglycemic, normoglycemic and TGC ranges. values. The mechanisms that cause this effect may be diffusion-related or mechanical in nature. A high hematocrit can decrease the There is a need to address factors affect- sors and rely on amperometry or photometry amount of glucose that diffuses into the bio- ing the accuracy of point-of-care (POC) glu- to measure glucose. The blood sample types sensor. The large amount of red blood cells in cose monitoring systems (GMS). Several stud- used in different devices may also vary. Most high-hematocrit samples may also mechani- ies have shown that factors such as pO2 systems are designed to accept capillary, cally impede the biosensor and reduce glu- and hematocrit affect GMS accuracy. More arterial or venous blood samples, which make cose readings. Low-hematocrit samples can recently, new instruments that adjust for them convenient tools for tight glycemic cause increased diffusion of glucose towards these types of factors have shown improved control (TGC). With the frequent use of POC the biosensor, thereby leading to higher read- performance characteristics of the devices.1 glucose monitoring for TGC in critically ill ings on the GMS. Some instrument manufac- Factors affecting accuracy may cause differ- patients, there is a need for accurate instru- turers address the hematocrit problem by ent effects based on the detection method ments that are resistant to the factors affect- pre-lysing red blood cells before testing or by used by a specific POC GMS. GMS commonly ing accuracy. The purpose of this article is to compensating for hematocrit effects using use enzymes such as glucose dehydrogenase provide a brief overview of these factors and software algorithms1. (GD) or glucose oxidase (GO) in their biosen- how to evaluate GMS accuracy.

64 Executive Summary Conference Report 7th Invited Conference: Assessing the Accuracy and Confounding Factors in Critical Care Glucose Monitoring

12 pO2 effects. Critically ill patients may be lular components may impede photometric tocols . Environmental conditions have also receiving oxygen therapy and blood pO2 lev- systems by increasing the turbidity of the become more important with the use of els may be higher than normal. GO-based sys- sample. Such matrix effects may also alter POC instruments under field conditions (e.g., tems originally had inaccuracies at extreme the viscosity and hence reduce the glucose disaster areas, war zones, rural areas). POC pO2 levels. This was common in systems that diffusion rate in both amperometric and pho- instruments, including GMS, were used on relied on reactions that generated a hydrogen tometric systems. airborne hospitals to evacuate survivors after peroxide intermediate. Hydrogen peroxide the Asian Tsunami of 200413. Kost et al. have Drugs. Drugs that can oxidize or reduce then disassociated into hydrogen ions, oxy- suggested that conditions in post-Tsunami glucose biosensor electrodes can also inter- gen and electrons (used to measure glu- Thailand and post-Hurricane Katrina New fere with glucose results. Tang et al. deter- cose)3. As pO increased (>100 mmHg), the Orleans were unacceptable for POC GMS use 2 mined the effect of 30 different drugs on disassociation reaction would become less because of high humidity and temperatures. six different glucose meters8. High doses favorable and thus cause lower glucose read- This has been verified by a study that tested of ascorbic acid were shown to yield lower ings. Recent GO-based systems no longer use the robustness of POC instruments under glucose values on GO-based systems. an oxygen intermediate, thereby minimizing certain environmental stresses14. Acetaminophen caused increased glucose or eliminating this confounding factor. The values in all tested GD-based systems and in Accuracy Assessment opposite may also be true, where low pO 2 some GO-based systems. High doses of dop- (<40 mmgHg) can affect glucose results by Methods are available to evaluate the amine increased glucose values in GD-based 15% when using GO-based systems4. accuracy of POC glucose meters. These systems. Mannitol increased glucose values include national and international standards, Hypotension and hypoperfusion. Sylvain in GO-based systems. Icodextrin has been as well as illustrative and statistical meth- et al. showed GD-based systems have sig- shown to generate falsely high glucose read- ods to determine if a GMS is comparable to nificantly different results (p < 0.001) when ings because the drug being metabolized a reference instrument. Illustrative methods testing capillary whole blood samples from to maltose, which is indistinguishable from such as mountain plots, Clarke Error Grid, hypotensive patients compared to a reference glucose with GD-based instruments9. and Bland-Altman plots are commonly used. analyzer5. GO-based systems showed similar pH and temperature. Since most GMSs These illustrative methods may be coupled results. In contrast, venous samples tested on use enzymes to catalyze the conversion of to standards such as the Clinical Laboratory GMS did not show statistically significant dif- glucose into a detectable signal, pH and Standards Institute (CLSI) and International ferences when compared to venous samples temperature may play a role in altering the Organization for Standardization (ISO) 15197 tested using laboratory analyzers. enzyme kinetics, thereby causing errors. criteria15,16. Examples of statistical methods Sample source. Capillary blood is the most Kilpatric et al. found significant deviation to evaluate accuracy include the Student’s common sample used for POC GMS. Several in glucose measurements for samples with t-test, analysis of variance (ANOVA), and least studies have shown variations in accuracy pH < 6.95 and > 7.8510. More recently, a new squares linear regression. when comparing capillary versus venous 5mg/dL median absolute difference standard Illustrative methods. These methods allow blood samples6. Recent GMS showed greater has been proposed. This standard is based on for quick determination of biases, trends accuracy with arterial samples compared to a locally smoothed median absolute differ- and errors. The mountain plot, also called an capillary blood. In contrast, older GMS mod- ence curve, a non-parametric mathematical empirical cumulative distribution plot, allows els showed similar results between the two method to identify trends and biases within for easy identification of the central 95% of sample types. Results derived from venous large datasets20. Temperature can also play a the data and provides clearer differentiation blood samples exhibited improved accuracy role. Oberg et al. suggest that cold tempera- of two distributions (e.g., GMS versus refer- compared to capillary blood for GD-based tures may also produce discrepancies but the ence)3. It is generated by computing the per- GMS. effects of fever are currently unknown4,11. centile for each ranked difference between Specimen matrix. Water, lipid, protein Other factors. User error and environ- the GMS and the reference. The plot shows and cellular content may also affect accu- mental conditions can also affect accuracy. the median bias between the two methods, racy. Glucose is more soluble in water and Studies have shown that the potential for while the tails of the plot show the propensity high plasma water content can cause higher operator error still exists despite advances for the new method to deviate from the refer- glucose readings7. Lipid, proteins and cel- in more user-friendly instruments and pro- ence method.

Executive Summary Conference Report 65 7th Invited Conference: Assessing the Accuracy and Confounding Factors in Critical Care Glucose Monitoring

The Clarke Error Grid divides a scatter plot Statistical methods. Statistical tests can Conclusions (GMS versus reference) into five zones (A, B, provide p-values or correlation coefficients. Manufactures must continue to develop C, D, and E)17. Data points falling into zones A The Student’s t-test is an example of one sta- measures to adjust for confounding fac- and B are considered acceptable, while data tistical test.3 There are three kinds of t-tests: tors, and healthcare professionals need to falling into zone C indicate the GMS’ results one-sample, two-sample, and the t-test for be aware of the limitations of the GMS they may prompt unnecessary corrections. Zone D paired differences. For the purposes of com- use. There are methods to evaluate accu- represents dangerous failure to detect a glu- paring a GMS versus a reference analyzer, the racy that are useful but may be misleading. cose level and Zone E represents results caus- Student’s t-test for paired differences is used, Instruments may show good correlation to ing erroneous treatment. There has recently since results from the same blood sample a reference standard, but this may not mean been the introduction of the continuous glu- are being compared. The Student’s t-test for that the results show agreement when using cose error grid that allows a user to evaluate paired differences assumes that the samples other methods such as linear regression. As the accuracy of continuous glucose monitors. are not independent (e.g., using the same instruments accommodate factors affecting The five zones are retained but the shape of blood samples) and the data are distributed accuracy through biosensor and software the error grid allows the investigator to also normally. design, there may be a shift towards newer, evaluate the glucose rate (e.g., blood glucose ANOVA is another statistical method, stricter methods to evaluate the accuracy of as a function of time)18. which compares the differences of three or GMS. Given the heterogeneity of physiologi- The Bland-Altman plot is a bias plot (bias more groups of data. This is very useful for cal conditions present in critically ill patients, versus reference)19. This method is useful in comparing variations between different GMS factors affecting accuracy may play a much that it shows biases, trends, and errors. When test strip lots. Both the Student’s t-test and larger role in trying to achieve TGC. Accuracy coupled to the CLSI or ISO 15197 standards, ANOVA generate a p-value, where p < 0.05 is at the low-ranges, especially for these plots serve to identify if an instrument usually considered statistically significant21. and neonates, are also important consider- meets the acceptance criteria. ations. High accuracy near and within the TGC Least squares linear regression, commonly ranges is also necessary for reliable glucose Accuracy standards. The CLSI and ISO referred to as linear regression, generates a monitoring in TGC protocols. Therefore, new 15197 standards are often cited in literature best-fit line onto a scatter plot (GMS versus range-specific analytical methods are needed when comparing laboratory instruments3,15,16. reference analyzer)3. The correlation coeffi- to assess the accuracy of GMS at hypoglyce- For example, the ISO 15197 standard uses the cient (r), coefficient of determination (r2), and mic, normoglycemic and TGC ranges. measurement of glucose concentrations from equation of the line are generated on this capillary blood samples and provides proce- plot. The r2-value is commonly used, with References 2 dures to verify and validate performance. The r ranging from 0 to 1, where 0 indicates a 1. Dungan K, Braithwaite SS, Chapman J, et al. Glucose ISO criteria requires 95% of the data points non-linear relationship and 1 is a perfect fit. measurement: confounding issues in setting tar- gets for inpatient management. Diabetes Care to have a bias of ±15 mg/dL at reference Manufacturers strive to attain a very high 2007;30:403. glucose levels of < 75 mg/dL. For reference r2-value, because it indicates high correlation 2. Dacombe CM, Dalton RG, Goldie DJ, et al. Effect of values ≥75 mg/dL, the bias must be within between GMS and the reference. However, packed cell volume on blood glucose estimations. ±20% of the reference value. In contrast, the it must be noted that least squares linear Arch Dis Child 1981;56:789. CLSI standard requires the bias to be ±15 mg/ regression is very susceptible to the weight- 3. Tran NK, Promptmas C, Kost GJ. Biosensors, min- dL for values < 100 mg/dL. For values ≥100 ing effects of data points at extreme values. iaturization, and noninvasive techniques. In: Cook KW, Lehmann C, Schoeff L, Williams R. “Clinical mg/dL, the bias must be within ±20% of the For example, if a single value at a high refer- Diagnostic Technology: The Total Testing Process, reference. Originally the ISO standard was ence range falls on the regression line, it may Preanalytical, Analytical, and Post-Analytical Phases, Volume 3 of 3.” Washington DC: AACC; Chapter 7, commonly used outside the United States. provide a high r2-value. Removal of this data 2006, p. 145-84. More recently, an October 2006 draft Food point may then reveal the less-than-satisfac- 4. Oberg D, Ostenson CG. Performance of glucose and Drug Administration (FDA) guideline tory nature of a dataset. Therefore, investiga- dehydrogenase—and glucose oxidase-based describes the use of the ISO 15197 criteria for tors must be aware of the potential for data blood glucose meters at high altitude and low temperatures (letter). Diabetes Care 2005;28:1261. GMS20. to have poor agreement but still produce relatively high correlations.

66 Executive Summary Conference Report 7th Invited Conference: Assessing the Accuracy and Confounding Factors in Critical Care Glucose Monitoring

5. Sylvain HF, Pokorny ME, English SM, et al. Accuracy 10. Kilpatrick ES, Rumley AG, Smith EA. Variations in 16. International Organization for Standardization. of fingerstick glucose values in shock patients. Am J sample pH and pO2 affect ExacTech meter glucose www.iso.ch (Accessed July 15, 2007). Crit Care 1995;4:44. measurements. Diabet Med 1994;11:506. 17. Clarke WL: The original Clarke error grid analysis 6. Kuwa K, Makayama T, Hoshino T, et al. Relationships 11. Haupt A, Berg B, Paschen P, et al. The effects of (EGA). Diabertes Technol Ther 2005;7:776. of glucose concentrations in capillary whole blood, skin temperature and testing site on blood glu- 18. Kovatchev BP, Cox DJ, Godner-Frederic LA, et al. venous whole blood and venous plasma. Clin Chim cose measurements taken by a modern blood Evaluating the accuracy of continuous glucose- Acta 2001;307:187. glucose monitoring device. Diabetes Technol Ther monitoring sensors. Diabetes Care 2004;27:1922. 2005;7:597. 7. D’Orazio P, Burnett RW, Fogh-Anderson N, et al. 19. Bland JM, Altman DG. Statistical methods for The International Federation of 12. Bergenstal R, Pearson J, Cembrowski GS, et al. assessing agreement between two methods of Science Division Working Group on Selective Identifying variables associated with inaccurate self- clinical measurement. Lancet 1986;327:307. Electrodes and Point of Care Testing: Approved monitoring of blood glucose: proposed guidelines IFCC recommendations on reporting results for to improve accuracy. Diabetes Edu 2000;26:981. 20. Kost GJ, Tran NK, Abad VJ, et al. Evaluation of point- blood glucose. Clin Chem 2005;1573. of-care glucose testing accuracy using locally- 13. Kost GJ, Tran NK, Tuntideelert M, et al. Katrina, smoothed median absolute difference curves. Clin 8. Tang Z, Du X, Louie RF, et al. Effects of drugs on glu- the tsunami, and point-of-care testing: optimizing Chim Acta 2007. [in press] cose measurements with handheld glucose meters rapid response diagnosis in disasters. Am J Clin and a portable glucose analyzer. Am J Clin Pathol Pathol 2006;126:513. 21. Total Produce Life Cycle for Portable Invasive Blood 2000;113:75. Glucose Monitoring Systems. Draft FDA Guidance 14. Sumner S, Louie R, Vo L, et al. Environmental Limits Document, October 24, 2006. http://www.fda. 9. U.S. Food and Drug Administration: FDA reminders of POCT: Relevance to Disaster Readiness. AACC gov/cdrh/oivd/guidance/1603.html Accessed on for falsely elevated glucose readings from use of National Meeting, San Diego, CA. Clin Chem 2007; 7/15/07. inappropriate test method, 2005. www.fda.gov/ Presentation Number, E-18. cdrh/oivd/news/glucosefalse.html. (Accessed on 22. Daniel WW. Biostatistics: a foundation for analysis 15. Clinical Laboratory Standards Institute. www.nccls. July 17, 2007) in the health sciences. 8th ed. Hoboken, JN: Wiley, org (Accessed July 15, 2007). 2005.

Executive Summary Conference Report 67 7th Invited Conference: Glucose Sensor Augmented Insulin Delivery in the Hospital–Open and Closed-Loop Methods

PROCEEDINGS Glucose Sensor Augmented Insulin Delivery in the Hospital: Open and Closed-Loop Methods Jeffrey I. Joseph, DO, Director, Artificial Pancreas Center, Jefferson Medical College of Thomas Jefferson University, Department of Anesthesiology, Philadelphia, PA

Key points The average time required for a nurse or technician to acquire a blood sample and • Frequent monitoring is required to effectively control blood glucose levels with a low measure the concentration of glucose using incidence of hypoglycemia; however, current manual methods of blood glucose monitoring a point-of-care meter has been estimated to are labor-intensive, costly, prone to error and expose the caregiver to potentially infectious be 4.7 minutes23. Thus, one to two hours of a blood. hospital employee’s time per day is required • A safe, effective and user-friendly medical device that automatically and continuously to monitor the blood glucose concentration monitors the concentration of glucose in hospitalized patients is greatly needed. of a patient being managed with IIT.

• Continuous glucose monitoring systems are being developed for hospital use that measure IIT methods have been plagued by an glucose in the blood or interstitial tissue fluid. unacceptably high rate of hypoglycemia24. All clinical trials published to date report a • Although a fully closed-loop automated system seems desirable for the hospital setting, higher incidence of hypoglycemia in the IIT safety and regulatory issues need to be overcome. group compared to the conventional treat- • A major step forward for in-hospital blood glucose control would be a system with a bedside ment group25-30. The fear of hypoglycemia monitor that organizes all important clinical data for glucose control in one real-time and the increased morbidity and mortality display and a computer algorithm to recommend an appropriate insulin infusion dose associated with hypoglycemia remains the according to glucose trend data. major barrier to the clinical application of IIT protocols in the hospital setting.

The prospective studies by Van den mal and human suggest that even mild to A safe, effective and user-friendly medical Berghe et al. focused attention on the impor- moderate hyperglycemia can adversely affect device that automatically and continuously tance of glycemic control in medical and immune function and cell survival following monitors the concentration of glucose in hos- surgical patients requiring intensive care1,2,3. ischemia19-21. In contrast, one recent study pitalized patients is greatly needed. Real-time Retrospective studies of patients in the inten- documented an increased risk for stroke and glucose monitoring will provide important sive care unit (ICU)4-13 revealed a marked death when a patient’s high glucose levels trend information: absolute concentration decrease in morbidity, mortality, length of stay were rapidly lowered to normal levels during (mg/dL, mmol/L), direction of change (stable, and cost when blood glucose is controlled in general anesthesia and cardiac surgery22. increasing or decreasing), and rate of change the near-normal range with intensive insulin (stable, slow or fast). The trend information Need for Frequent Monitoring therapy (IIT) methods. Recent clinical trial can be used by the bedside nurse to titrate evidence suggests a correlation between Frequent monitoring is required to effec- the delivery of insulin and/or glucose to high blood glucose variability and increased tively control blood glucose levels with a low maintain blood glucose in the desired range. morbidity and mortality14-15. Although not incidence of hypoglycemia. Current manual The risk for hypoglycemia will be minimized all trials using IIT for glucose control have methods of blood glucose monitoring are and or eliminated. demonstrated an improvement in clinical labor-intensive, costly, prone to error, and Continuous glucose monitoring (CGM) outcome16-18, the majority of studies in ani- expose the caregiver to potentially infectious systems are being developed for use in hos- blood. pitals that measure glucose in the blood or

68 Executive Summary Conference Report 7th Invited Conference: Glucose Sensor Augmented Insulin Delivery in the Hospital–Open and Closed-Loop Methods interstitial tissue fluid (ISF). Some glucose Figure 1 sensors are inserted directly into the subcuta- neous tissue or blood stream and other CGM systems automatically deliver a sample of ISF or blood to a glucose sensor external to the body31-32.

Blood Glucose Monitors

The Via® Blood Chemistry Monitor for Glucose (VIA) shown in Figure 1 was devel- oped by VIA Medical Corporation (San Diego, CA) in 1991 to automate the process of blood glucose monitoring in the hospital VIA Blood Chemistry Monitor for Glucose attached to vascular catheter in peripheral vein. Vascular catheter connected setting. This device received Food and Drug to sterile tubing, flow-through glucose sensor and Isolyte-glucose calibration solution. Bedside monitor contains user Administration approval to measure the con- controls, bi-direction infusion pump, air bubble sensor, pressure sensor, data display, auto-sampler and printer. Sensor housing and cable are attached to the patient’s left arm with tape. centration of glucose as frequently as every 5 minutes for 72 hours using blood sampled ized patients with cardiac and renal disease. tions, but vessel thrombosis and clot for- from a radial artery catheter, a peripheral Glucose measurements using a VIA glucose mation within the catheter lumen remain a venous catheter or the proximal port of a cen- monitor are shown in Figures 2 and 3. clinical problem38-40. The ability to frequently tral venous catheter. The system automati- sample blood from a central venous cath- cally delivers a sample of patient blood to an All external flow-through sensors have eter (CVC) over time using the VIA has not external flow-through glucose sensor using a problems with blood vessels and indwelling been validated. The lumen of the CVC will be bi-directional infusion pump33-35. catheters when blood samples are obtained exposed to static blood for 60 minutes per frequently over an extended period of time. The sensor measures the blood glucose day when sampling once every 20 minutes, Repeated sampling from the peripheral vein concentration using glucose oxidase to pro- possibly leading to catheter obstruction (50 of an ICU patient can be a problem because duce hydrogen peroxide. Each sample is seconds x 3 samples/hr x 24 hours = 3,600 of low flow, vessel wall collapse, obstruction automatically infused back into the patient, seconds). The VIA sample can be contaminat- from a valve, vessel thrombosis and catheter avoiding blood loss and caregiver exposure ed with glucose-free or glucose-containing occlusion due to clot, fibrous tissue and kink- to bodily fluids. VIA sensors tested in-vivo solutions being infused through the tubing. ing36-37. Attaching the VIA to a radial artery demonstrate high sensitivity, accuracy (R2 The sample can also be contaminated with catheter will overcome some of these limita- = 0.997) over the physiological range (30 to fluids infused through an adjacent CVC port. 600 mg/dL), specificity for glucose and lack of sensitivity to changes in oxygen and hemat- Figure 2. Type 1 Diabetic–Leg Ischemia 300 ocrit. 500

450 Reference 250 The calibration solution is continuously 400 VIA Blood Glucose Monitor infused through the tubing and sensor at 350 200 a rate of 5 mL/hour. Before each blood glu- 300 cose measurement, the monitor performs 250 150 a one-point calibration using the Isolyte- 200 150 100 glucose solution (83 mg/dL) as the reference. Blood Glucose (mg/dL)

Blood Glucose (mg/dL) 100 The acquired blood sample remains within 50 50 Reference the sensor and tubing for approximately 50 0 VIA Blood Glucose Monitor

15-Feb-0216-Feb-02 17-Feb-02 18-Feb-02 19-Feb-02 20-Feb-02 21-Feb-02 22-Feb-02 23-Feb-02 24-Feb-02 25-Feb-02 0 seconds. The sample is then flushed back 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00000:00 02:00 4:00 06:00 08:00 10:00 12:00 14:00 Date into the bloodstream with 6.0 mL of Isolyte- Time (HH:MM) glucose solution. The volume of fluid infused Glucose control in hospitalized patient with type 2 diabetes requiring 80 units of insulin/day. Note glucose variability using fingerstick measurements and intermittent subcutaneous insulin injections ( ). VIA glucose monitor used to by VIA may be excessive for many hospital- adjust IV infusion of regular insulin during anesthesia and major ( ).

Executive Summary Conference Report 69

Blood glucose concentration Plasma Insulin (mU/I) 300 12.0 Meals Insulin infusion rate 250 10.0

200 8.0

150 6.0 oncentration, mg/dL (mU/I)

100 4.0

50 2.0

Glucose (Insulind) C 0 0.0 8:24 9:36 10:48 12:00 13:12 14:24 15:36 16:48 18:00 Time (HH:MM) 7th Invited Conference: Glucose Sensor Augmented Insulin Delivery in the Hospital–Open and Closed-Loop Methods

ISF sensors overcome the limitations of blood- Figure 3. Glucose Sensor Augmented Insulin Delivery based glucose sensors, such as lack of vascu- Open-loop lar access, vessel occlusion, sample contami- 300 nation and infection. 500

450 Reference 250 Recent studies have used an analysis of 400 VIA Blood Glucose Monitor multiple simultaneous glucose sensor output 350 200 signals to improve accuracy and robustness 300 250 150 in the clinical setting. An array of sensors 200 implanted in more than one location can be 300 100 150 500 used to ensure a steady stream of glucose data Blood Glucose (mg/dL)

Blood Glucose (mg/dL) 100 450 Reference with accuracy superior to an individual250 sensor 50 50 Reference 400 VIA Blood Glucose Monitor implanted in one location. Miniaturization of 0 VIA Blood Glucose Monitor 350 200 15-Feb-0216-Feb-02 17-Feb-02 18-Feb-02 19-Feb-02 20-Feb-02 21-Feb-02 22-Feb-02 23-Feb-02 24-Feb-02 25-Feb-02 0 the sensor and electronics is required to make 06:00 08:00 10:00300 12:00 14:00 16:00 18:00 20:00 22:00000:00 02:00 4:00 06:00 08:00 10:00 12:00 14:00 Date this concept practical in the clinical setting42. 250 Time (HH:MM) 150 200 Closed-loop System Close-up of glucose measurements using VIA glucose monitor attached to a peripheral vein of a diabetic patient under- 100 going major vascular surgery. Fingerstick150 glucose measurements were obtained once hourly (reference). IV insulin infu- The Biostator commercialized by Miles Blood Glucose (mg/dL) sion was titrated by an experiencedBlood Glucose (mg/dL) 100 anesthesiologist. 50 Laboratories in the 1970s demonstrated50 Re theferenc e Interstitial Tissue Fluid Monitors bedside glucose testing should equate to less 0 feasibility of continuous blood glucose moniVIA Blood- Glucose Monitor 15-Feb-0216-Feb-02 17-Feb-02 18-Feb-nursing02 19-Feb-02 20-time.Feb-02 21-Feb-02 22-Feb-02 23-Feb-02 24-Feb-02 25-Feb-02 0 Many are quick to discount the clinical util- toring and closed-loop feedback06:00 control08:00 10:00 of 12:00 14:00 16:00 18:00 20:00 22:00000:00 02:00 4:00 06:00 08:00 10:00 12:00 14:00 Date ity of needle-type glucose sensors inserted The major advantage of the needle-sen- glucose levels. Although a fully closed-loop Time (HH:MM) into the subcutaneous tissue of hospitalized sor technology is the ease of insertion into automated system seems desirable for the patients. Issues with tissue edema, physi- the subcutaneous tissue. Although typically hospital setting, safety and regulatory issues ological differences between blood and ISF implanted in the loose connective tissue of need to be overcome. Sensor inaccuracies, Blood glucoseglucose, concentration variable time delays following a step the abdomen, insertion in the thigh, arm, time delays in glucose sensing, time delays Plasma Insulin (mU/I) 300 12.0 chest (Figure 5) and buttock are also pos- Meals change in glucose, variable changes in tissue in IV insulin pharmacodynamics and loss of Insulin infusionoxygen rate concentration and blood flow and sible. 250 10.0 potential interfering substances have been Figure 4. Bedside Data Display Glucose Sensor, Insulin, Meals cited as concerns. 200 8.0 Preliminary data in hospitalized medical Profile for 5D16A0BA 150 6.0 and surgical patients, however, has encour- oncentration, mg/dL (mU/I) Blood glucose concentration Plasma Insulin (mU/I) 100 aged further research. 4.0Comparison of glucose 300 12.0 Meals sensor data with reference arterial, venous Insulin infusion rate 50 and capillary glucose data2.0 have yielded mean 250 10.0 absolute relative difference (MARD) and cor-

Glucose (Insulind) C 0 0.0 200 8.0 8:24 9:36 10:48 12:00 13:12 14:24 relation15:36 coefficient16:48 18:00values that are clinically Time (HH:MM) acceptable31,32,41,42. 150 6.0

Current needle-type sensors provide an oncentration, mg/dL (mU/I) 100 4.0 acceptable estimate of ISF glucose trends. Bedside nurses will be able to combine the 50 2.0 CGM trend data with intermittent reference

blood sample glucose measurements to Glucose (Insulind) C 0 0.0 8:24 9:36 10:48 12:00 13:12 14:24 15:36 16:48 18:00 improve the titration of insulin (Figure 4). This Time (HH:MM) combined approach should improved overall glucose control while decreasing the risk of Flat-panel display visualizing VIA glucose monitor trend data ( ), intravenous insulin infusion trend data ( ) and actual venous blood insulin levels (ELISA method). Type 1 diabetes patient consumed breakfast at 8:50, lunch at 12:50 hypoglycemia. Fewer blood samples and less and exercised on a bicycle at 16:00.

70 Executive Summary Conference Report 7th Invited Conference: Glucose Sensor Augmented Insulin Delivery in the Hospital–Open and Closed-Loop Methods real-time data are problems when used in the Conclusion Key to the clinical success of glucose sensor real-world setting. Successful application of a augmented insulin delivery is a CGM system There is great clinical need in the hospital CGM system that is accurate and robust in the that works well in a broad range of patient for a continous glucose monitoring system hospital setting will overcome the major tech- populations and hospital environments. that is safe, accurate, robust and user-friendly. nical obstacles of a fully closed-loop system. Although interstitial fluid and blood-based Although our ultimate goal is the development glucose sensors have shown great promise in the hospital setting, ongoing research is Figure 5. Continuous ISF glucose monitoring needed to optimize their clinical use43,44. 1440 measurements/day vs. 4 References

1. Van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin therapy in the critically ill patients. N Engl J Med. 2001 Nov 8;345(19):1359-67.

2. Van den Berghe G, Wouters PJ, Bouillon R, et al. Outcome benefit of intensive insulin therapy in the critically ill: Insulin dose versus glycemic control. Crit Care Med. 2003 Feb;31(2):359-66.

3. Van den Berghe G, Wilmer A, Hermans G, et al. Intensive insulin therapy in the medical ICU. N Engl J Med. 2006 Feb 2;354(5):449-61.

4. Krinsley JS, Grover A. Severe hypoglycemia in criti- cally ill patients: Risk factors and outcomes. Crit Care Med. 2007 Oct; 35(10):2262-7.

5. Krinsley JS. Association between hyperglycemia and increased hospital mortality in a heteroge- neous population of critically ill patients. Mayo Clin Proc. 2003 Dec;78(12):1471-8. Medtronic MiniMed needle-type CGM inserted into right upper arm (3 sensor array) and anterior chest (3 sensor array) of type 2 diabetes patient undergoing major surgery. Medtronic Vascular Glucose Monitoring System (VGMS) inserted 6. Krinsley JS. Effect of an intensive glucose manage- through central venous catheter into superior vena cava. ment protocol on the mortality of critically ill adult patients. Mayo Clin Proc. 2004 Aug;79(8):992-1000. Bedside Monitor of a closed-loop artificial pancreas capable of 7. Furnary AP, Zerr KJ, Grunkemeier GL, et al. controlling glucose levels in a narrow range Continuous intravenous insulin infusion reduces A major step forward for in-hospital blood the incidence of deep sternal wound infection in safely and automatically, our immediate goal glucose control would be a bedside monitor diabetic patients after cardiac surgical procedures. is to provide the bedside nurse with real-time Ann Thorac Surg. 1999 Feb;67(2):352,60; discussion that displays the real-time CGM glucose trend 360-2. glucose, insulin and meal information that is data, insulin delivery data and enteral/paren- organized and analyzed in a clinically useful 8. Furnary AP, Gao G, Grunkemeier GL, et al. Continuous teral nutrition delivery data. Organizing all of insulin infusion reduces mortality in patients with way. Alarms will alert the caregiver when glu- the important clinical data for glucose control diabetes undergoing coronary artery bypass graft- cose levels exceed a programmable threshold ing. J Thorac Cardiovasc Surg. 2003 May;125(5):1007- in one real-time display would help the bed- 21. and increase or decrease at a high rate of side nurse titrate insulin, glucose and meals. change. Smart computer algorithms will rec- 9. Bochicchio GV, Sung J, Joshi M, et al. Persistent A computer algorithm similar to the com- hyperglycemia is predictive of outcome in criti- ommend the most appropriate management mercially available GlucoScout or EndoTool® cally ill trauma patients. J Trauma 2005 May;58(5): of the patient over a wide range of clinical 921-4. could be used to recommend an appropriate situations. Nurses will utilize this real-time 10. Finney SJ, Zekveld C, Elia A, et al. Glucose control insulin infusion dose based on glucose trend and mortality in critically ill patients. JAMA. 2003 information to optimize the delivery of insulin data. Advanced algorithms will consider IV Oct 15;290(15):2041-7. in relation to the clinical needs of the patient. insulin kinetics/dynamics, meal models and 11. Ouattara A, Lecomte P, Le Manach Y, et al. Poor Glucose levels will be better controlled and real-time parameter adjustments for acute intraoperative blood glucose control is associated the risk for hypoglycemia will be minimized or with a worsened hospital outcome after cardiac changes in insulin sensitivity, drugs and major surgery in diabetic patients. Anesthesiology. 2005 eliminated. stress (steroids, catecholamines, sepsis and Oct;103(4):687-94. cardiopulmonary bypass).

Executive Summary Conference Report 71 7th Invited Conference: Glucose Sensor Augmented Insulin Delivery in the Hospital–Open and Closed-Loop Methods

12. Umpierrez GE, Isaacs SD, Bazargan N, et al. 24. Cryer PE. Hypoglycaemia: The limiting factor in 38 Cousins TR, O'Donnell JM. Arterial cannulation: A Hyperglycemia: An independent marker of in-hos- the glycaemic management of the critically ill? critical review. AANA J. 2004 Aug;72(4):267-71. pital mortality in patients with undiagnosed diabe- Diabetologia. 2006 Aug;49(8):1722-5. 39. Martin C, Saux P, Papazian L, et al. Long-term arterial tes. J Clin Endocrinol Metab. 2002 Mar;87(3):978-82. 25. Braithwaite SS, Buie MM, Thompson CL, et al. cannulation in ICU patients using the radial artery 13. Clement S, Braithwaite SS, Magee MF, et al. Hospital hypoglycemia: Not only treatment but also or dorsalis pedis artery. Chest. 2001 Mar;119(3):901- Management of diabetes and hyperglycemia in prevention. Endocr Pract. 2004 Mar-Apr;10 Suppl 6. hospitals. Diabetes Care. 2004 Feb;27(2):553-91. 2:89-99. 40. Wallach SG. Cannulation injury of the radial artery: 14. Egi M, Bellomo R, Stachowski E, et al. Variability of 26. Fischer KF, Lees JA, Newman JH. Hypoglycemia in Diagnosis and treatment algorithm. Am J Crit Care. blood glucose concentration and short-term mor- hospitalized patients. causes and outcomes. N Engl 2004 Jul;13(4):315-9. tality in critically ill patients. Anesthesiology. 2006 J Med. 1986 Nov 13;315(20):1245-50. 41. Hipszer B, Furlong K, Lessin J, et al. Continuous Aug;105(2):244-52. 27. Stagnaro-Green A, Barton MK, Linekin PL, et al. Glucose Monitoring in the Perioperative Period. 15. Hirsch IB, Brownlee M. Should minimal blood glu- Mortality in hospitalized patients with hypoglyce- American Society of Anesthesiology, Abstract October cose variability become the gold standard of gly- mia and severe hyperglycemia. Mt Sinai J Med. 1995 2006. cemic control? J Diabetes Complications. 2005 May- Nov;62(6):422-6. 42. Hipszer B, Chervoneva I, Gratch D, et al. The Utility Jun;19(3):178-81. 28. Thomas AN, Boxall EM, Twamley HW. Evaluation of Simultaneous Glucose Sensor Measurements. 16. GluControl study: Comparing the effects of two of short-term consequences of hypoglycemia Diabetes Techology Society. Abstract November glucose control regimens by insulin in intensive in an intensive care unit. Crit Care Med. 2007 2007. care unit patients [homepage on the Internet]. Apr;35(4):1218,9; author reply 1219. 43. Joseph JI. Anesthesia and Surgery in the Diabetic U.S. National Institutes of Health [cited November 29. Vriesendorp TM, DeVries JH, van Santen S, et al. Patient, Chapter 31, Textbook of Type 2 Diabetes, 2007]. Available from: http://clinicaltrials.gov/show/ Evaluation of short-term consequences of hypogly- Martin Dunitz Publisher, 2007. NCT00107601. cemia in an intensive care unit. Crit Care Med. 2006 44. Joseph JI. Future Management Approaches: New 17. Mitchell I, Finfer S, Bellomo R, et al. ANZICS Clinical Nov;34(11):2714-8. Devices in the Management of Diabetes, Chapter Trials Group Glucose Management Investigators. 30. Wexler DJ, Meigs JB, Cagliero E, et al. Prevalence of 36, Textbook of Type 2 Diabetes, Martin Dunitz Management of blood glucose in the critically ill in hyper- and hypoglycemia among inpatients with Publisher, 2007. australia and New Zealand: A practice survey and diabetes: A national survey of 44 U.S. hospitals. inception cohort study. Intensive Care Med. 2006 Jun; Diabetes Care. 2007 Feb;30(2):367-9. 32(6):867-74. 31. Goldberg PA, Segal M, Russel RR et al.Experience 18. DeBrouwere R. Tight intraoperative glucose control with the continuous glucose monitoring system in a does not improve outcome in cardiovascular sur- medical intensive care unit. Diabetes Techology and gery. Journal of Cardiothoracic & Vascular Anesthesia Therapeutics 2004;6:339-47. 2000;14(4):479-81. 32. Vriesendorp T, DeVries J, Holleman F et al. The use 19. Golden SH, Peart-Vigilance C, Kao WH, et al. of two continuous glucose sensors during and Perioperative glycemic control and the risk of infec- after surgery. Diabetes Technology and Therapeutics tious complications in a cohort of adults with diabe- 2005;7:315-22. tes. Diabetes Care. 1999 Sep;22(9):1408-14. 33. Ganesh A, Hipszer B, Loomba N, et al. Evaluation of 20. Pomposelli JJ, Baxter JK,3rd, Babineau TJ, et al. Early the VIA® Blood Chemistry Monitor for Glucose in postoperative glucose control predicts nosocomial Healthy and Diabetic Volunteers (In Press) Journal infection rate in diabetic patients. JPEN J Parenter of Diabetes Science and Technology 2008 Enteral Nutr. 1998 Mar-Apr;22(2):77-81. 34. Lucisano JY, Edelman SV, Quinto BD, et al. 21. Lazar HL, Chipkin SR, Fitzgerald CA, et al. Tight Development of a biosensor-based, patient- glycemic control in diabetic coronary artery bypass attached blood glucose monitoring system. Proc graft patients improves perioperative outcomes and Am Chem Soc. 1997;76:256. decreases recurrent ischemic events. Circulation. 2004 Mar 30;109(12):1497-502. 35. VIA blood chemistry monitor for glucose VIA V-GLU 1 Operator’s manual. San Diego, CA: VIA Medical 22. Gandhi G, Nuttall G, Abel M, et al. Intensive intra- Corporation; 1998. operative insulin therapy versus conventional glu- cose management during cardiac surgery. Annals of 36. Kagel EM, Rayan GM. Intravenous catheter com- Internal Med 2007:Vol146 (4)233-43. plications in the hand and forearm. J Trauma 2004 Jan;56(1):123-7. 23. Aragon D. Evaluation of nursing work effort and perceptions about blood glucose testing in tight 37. Mohler M, Sato Y, Bobick K, et al. The reliability of glycemic control. Am J Crit Care 2006 Jul;15(4): blood sampling from peripheral intravenous infu- 370-7. sion lines. complete blood cell counts, electrolyte panels, and survey panels. J Intraven Nurs. 1998 Jul- Aug;21(4):209-14.

72 Executive Summary Conference Report 7th Invited Conference: The Impact of Intensive Insulin Therapy on Nursing

PROCEEDINGS The Impact of Intensive Insulin Therapy on Nursing Daleen Aragon, PhD, CCRN, FCCM, Director, Advanced Practice Nursing and Research, Orlando Regional Healthcare, Orlando, FL

Key points time and activities involved in hourly gly- cemic control activities, nursing perception • Intensive insulin therapy (IIT) results in an increase in nursing workload. of TGC, and the estimated costs of glycemic • Approximately two hours of direct nursing time per patient may be devoted to glycemic control. Another purpose of the study was to control activities in a 24-hour period. get input from nurses about potential use of

• Approximately $250,000 in costs was associated with IIT in one intensive care unit in one continuous IV methods to monitor blood glu- year for nursing activities, blood glucose testing and supplies. cose, since many companies are developing this technology1.

The study setting was six ICUs with a As tight glycemic control (TGC) with inten- at the bedside voiced concerns when they total of 58 beds in a level I trauma center in sive insulin therapy (IIT) has become more attempted to work glycemic control mea- the southeastern United States. The study common in the intensive care unit (ICU), a sures into their hourly routines. Many indi- was descriptive and exploratory in nature. change in nursing activities has evolved to cated that this was difficult to accomplish Time and motion methods were used to incorporate this initiative into patient man- because of the time and effort required. observe nurses performing glycemic control agement. Intravenous (IV) insulin is being activities. Surveys were distributed to 122 Impact of TGC on Nursing used more often to control blood glucose, nurses that used a five-point Likert scale, since subcutaneous insulin is less effective A study was initiated to evaluate the structured questions about perceptions of 1 because of variable absorption in the critical- impact of TGC on nursing workload . The glycemic control and open-ended questions ly ill patient. Administering IV insulin requires purposes of the study were to evaluate the for comments. Information on the number more intensive blood glucose monitoring to avoid dangerous glycemic excursions and to Table 1. Survey Results1 meet target glucose levels. Hourly or more (1-5 scale – 3 = moderate agreement) frequent checking of blood glucose and 5 4.83 frequent adjustments and titration of insu- 4.32 4.5 lin typically are required, particularly when 4 3.73 beginning IV insulin therapy and when the 3.5 patient has not reached blood glucose target 2.98 3 0.7 levels. 2.38 62% 2.5 2.18 59% 1.96 0.6 Critical care nurses already have frequent 2 0.5 1.5 monitoring, measurement and treatment 38% 0.4 activities on an hourly basis. The additional 1 26% tasks required to manage blood glucose sig- 0.5 0.3 nificantly increase nursing workload. 0 0.2 believes hourly BG don’t too much prefer not concern automatic When glycemic control initiatives were helps barrier mind work on drip over FS system 0.1 outcomes doing needed implemented at our organization, nurses 0 Good for Worr y for Work E ort Monitoring patient care patients issues Executive Summary Conference Report 73

4000 3500 3000 CVRR 2500 ICU 2000 BTICU 1500 CCU 1000 500 0 t v c y t Oc No De Jan Feb Mar Apr Ma June July Aug Sept Oc 7th Invited Conference: The Impact of Intensive Insulin Therapy on Nursing 5 4.83 4.32 4.5 of point-of-care3.73 (POC) blood glucose tests, 4 Table 2. Answers to Open-ended Questions1 costs for materials and supplies, and nursing 3.5 Survey (n-66) 2.98 salaries was obtained from existing hospital 3 0.7 databases2.38 1. 62% 2.5 2.18 59% 1.96 0.6 2 Survey Results 0.5 1.5 38% 1 Sixty-six of 122 nurses (54%) completed 0.4 26% 0.5 the survey. Key findings are shown in Tables 0.3 1-4. Nurses strongly agreed that glycemic 0 0.2 believes hourly BG don’t too much controlprefer no tis coanncer importantn automati caspect of patient helps barrier mind work on drip over FS system 0.1 outcomes doing care that affected outcomes.needed They also felt 0 glycemic control activities required exten- Good for Worr y for Work E ort Monitoring sive time and increased workload. They also patient care patients issues were concerned about the number of finger- sticks required in5 some patients to obtain for one hourly round from4.83 testing to docu- and length of stay associated with improv- 4.32 blood glucose4. readings5 for IIT. They strongly mentation, one patient on IV insulin requires ing glycemic control. Other possible financial 3.73 indicated that more4 automatic and continu- approximately two full hours of direct nursing benefits of glycemic control could be realized ous sampling 3.technological5 methods were time during a 24-hour period. Based on nurs- through reduction in morbidity and mortality, 2.98 needed and that3 they would be willing to ing salaries, the number of POC blood glucose improved patient outcomes,0.7 fewer complica- 2.38 62% devote an IV 2.access5 site for that purpose.2.18 tests, costs of testing supplies, quality control tions and improved resource utilization in the 59% 1.96 0.6 When asked, 90%2 of the nurses indicated that tests and average nursing time, the costs to critically ill patients. 0.5 4000 1.5 rather than delegating activities to ancillary perform glycemic control activities in our 38% 3500 1 Elements that can facilitate0.4 the implemen- clinical technicians, they preferred to per- hospital’s ICU was conservatively estimated 26% 3000 0.5 CVRR tation of glycemic control0. efforts3 include hav- form their own blood glucose measurement to be $250,000 per year (Table 4). The steady 2500 0 ICU ing an excellent, passionate leader to keep the to have better control over the process and increase in the number of POC blood glucose 0.2 2000 believes hourly BG don’t too much prefer not concern automatic initiative going; forming a multidisciplinary greater certainty of thehelp resultss barrier before titratingmind wotestsrk overon thedrip yearover (Figure) FSBTICsystemU further confirmed 0.1 outcomes doing needed team to guide and lead the directive; and IV insulin.1500 Other survey comments were that the increased time spentCC Uin measuring blood 0 1000 obtaining input from potential stakeholders nurses felt the extra workload associated with glucose as IIT was being administered1. Good for Worr y for Monitoring in the project. Nurses at the front line need to Work E ort glycemic control500 competed with other tasks patient care patients issues 0 Discussion be included in planning efforts for glycemic performed on ant hourlyv basis, particularlyy t Oc No Dec Jan Feb Mar Apr Ma June July Aug Sept Oc control, so that they have an opportunity to when nurses were assigned more than one Is all the work worthwhile? Studies con- provide input into the process and streamline patient being given IV insulin. Having to focus ducted by Krinsley2 and by Van den Berghe3 their work. on these time-sensitive glucose control activ- both showed decreased patient care costs ities sometimes took them away from other things they could be doing for the patient1. Figure. POCT BG tests 1 yr1 Nursing Work Time and Costs n = 85,530 4000 Up to 32 steps that could be part of hourly 3500 glycemic control activities were observed 3000 CVRR during the study. There was considerable 2500 ICU variation in how nurses performed blood glu- 2000 BTICU cose testing. Nurses sometimes took short- 1500 CCU cuts such as omitting hand hygiene or not 1000 wearing gloves to decrease the time needed 500 to do these procedures. With an average time 0 t v y t of 4.72 minutes (range 3.13-8.15 minutes) Oc No Dec Jan Feb Mar Apr Ma June July Aug Sept Oc

74 Executive Summary Conference Report 7th Invited Conference: The Impact of Intensive Insulin Therapy on Nursing

Conclusion Table 3. Cued Responses1 While nurses felt that glycemic control was Checked y/n what applied) % N important for the patient, they also felt that Too much work 24 16 there was considerable extra work involved. Takes too much time 44 29 The costs associated with glycemic control efforts were substantial. Providing nurses Is a waste of time 6 4 with a continuous and/or non-invasive meth- Easier if automated 86 56 od for blood glucose measurement was well Like doing it 1.5 1 accepted by nurses as a potential solution to Is not difficult to do 38 25 the increased work associated with POC test- Normal part of patient care 45 30 ing. Should be done by other than RN 15 10 References Willing to dedicate an IV line if automated and displayed 76 50 1. Aragon, D. Evaluation of nursing work effort and Who performs BG monitoring for patients perceptions about blood glucose testing in tight on IV insulin infusions? RNs—90% glycemic control. Am J Crit Care 2006;15:370-7. Clinical technicians —10% 2. Krinsley JS, Jones RL. Cost analysis of intensive glycemic control in critically ill adult patients. Chest 2006;129:644-50. Table 4. Findings: Summary1 3. Van den Berghe G, Wouters PJ, Kesteloot K, et al. Analysis of healthcare resource utilization with • About 2 hours direct nursing care to perform glycemic control related activities intensive insulin therapy in critically ill patients. Crit Care Med 2006; 34:612-6. • Nurses believed glycemic control was beneficial for patient outcomes

• Increases the number of BG tests and calibrations done • Estimated cost of around $250,000 of nursing time and supplies for one hospital

Executive Summary Conference Report 75 7th Invited Conference: Nursing Education and Intensive Insulin Therapy

PROCEEDINGS Nursing Education and Intensive Insulin Therapy Carol S. Manchester, MSN, APRN, BC-ADM, CDE, Diabetes Clinical Nurse Specialist University of Minnesota Medical Center, Fairview University of Minnesota Children’s Hospital, Fairview, Minneapolis, Minnesota; Adjunct Faculty, University of Minnesota School of Nursing, Minneapolis, Minnesota

Key points led by an endocrinologist and an advanced practice nurse. Composed of individuals who • Core competencies related to TGC and diabetes care include the proper use of insulin and had a passion for diabetes and championed other diabetes medications, medical nutrition therapy, point-of-care testing, hypoglycemia tight glycemic control (TGC), the team had treatment and concomitant therapies. ultimate accountability and responsibility for • Establishment of an interdisciplinary Acute Care Diabetes Advisory Committee was the first all aspects of diabetes care. Complementary step in a quality improvement process. knowledge and skills allowed the team to

• A thorough assessment of each professional discipline’s current level of knowledge and skills build upon each other’s strengths in devel- is essential to the development of a strategic plan for glycemic-management education. oping an educational program that would disseminate the requisite knowledge and • Challenges include nurses’ varied work schedules, “information acquisition fatigue,” budget training to ensure professional competency constraints and need for administration to understand and support the concept of training throughout the care system. to competency rather than to a specific period of time. Core Competencies • An educational plan for diabetes management and tight glycemic control can include Core competencies related to TGC and employee orientation, tests of medication knowledge, preceptor guidelines and curriculum, diabetes care include the proper use of insu- annual competency testing for all disciplines, continuing education and remedial support. lin and other diabetes medications, medi- • A creative and innovative approach to education is welcomed by staff and more effective for cal nutrition therapy, point-of-care testing, sustained professional growth. hypoglycemia treatment and concomitant therapies. Educational methods to success- fully promote competency include case Ensuring the professional competency of tion. However, the information applies to all scenarios with exercises that require critical the entire clinical staff is essential to suc- healthcare disciplines, including pharmacy, thinking and problem-solving skills to answer cessfully delivering evidence-based, safe, nutrition and medicine. the questions correctly, simulation exercises, effective, respectful and appropriate diabe- Inpatient Glycemic Management clinical rounding on all shifts, observation tes care. In this article, key elements of a Improvement and shadowing experiences and cross-disci- successful inpatient glycemic management pline training. Tables 1 and 2 provide exam- educational program are reviewed, including The University of Minnesota Medical ples of insulin competencies and a question core competencies, knowledge assessment, Center (UMMC), Fairview, initiated an inpa- designed to promote critical thinking skills. development of a strategic educational plan tient glycemic management improvement and continuing education. project in August, 2000. The average daily Assessing Level of Knowledge census is 1,000 and 47% of the inpatient Since nursing comprises the largest num- Before developing a strategic plan for population develop hyperglycemia. ber of staff members at the frontline of care glycemic-management education, the team delivery and the nurse is critical to the suc- The first step in quality improvement conducted a thorough assessment of each cess of diabetes care improvement initia- process was to establish an interdisciplinary professional discipline’s current knowledge tives, this article focuses on nursing educa- Acute Care Diabetes Advisory Committee, co- and skills and reviewed and inventoried the

76 Executive Summary Conference Report 7th Invited Conference: Nursing Education and Intensive Insulin Therapy available tools and reference materials. At cation was developed that included critical Professional Development UMMC 73% of the nursing staff held a bach- information and core competencies, catego- After competency was achieved, the next elor’s degree or higher, whereas at a commu- rized as “must know” and “nice to know” infor- stage was continuing education on the basics nity affiliate, the overwhelming majority held mation, and evaluated learning outcomes. of diabetes management, special-population an associate’s degree. Knowing the staff’s The educational plan for diabetes manage- needs, research and advanced practice taught training and educational background, par- ment and TGC initiatives included employee across the continuum of care. Methods to ticularly with regard to critical thinking skills orientation, tests of medication knowledge, facilitate learning included the development and clinical judgment, was essential to deter- preceptor guidelines and curriculum, annual and use of case studies and case scenarios, mining the appropriate educational content competency testing for each discipline (phy- grand round presentations, web-based learn- and teaching methods. sician, nurse, dietitian and pharmacist), pro- ing modules, journal clubs, resource toolkits To determine the current knowledge level fessional development and continuing edu- for each patient care unit, medication charts, of the staff, the team developed an interdis- cation, and resource or remedial support. For pocket cards and staff participation in presen- ciplinary tool that evaluated critical thinking facilities affiliated with an academic health tations. A creative and innovative approach skills related to insulin, basal/bolus insulin center, the plan also included a review of the to education was welcomed by staff and has therapy, carbohydrates and hypoglycemia. School of Nursing’s curriculum and develop- proven more effective for sustained profes- Only 44% of the staff achieved a passing score ment of strong graduate student preceptor- sional growth. of 90%. Education was needed to ensure ships. Summary competency before any new protocols, order Central nursing’s new employee orienta- sets and practices could be implemented. Lessons learned from a successful inpa- tion was a perfect opportunity to impart the tient glycemic management improvement Development of a Strategic Plan for necessary knowledge and understanding of program suggest that the following can help Education glycemic management. A review of diabetes ensure professional competency in diabetes classifications, insulin basal-bolus manage- The challenge of providing education to management to improve quality outcome ment, the rationale for TGC, acute-care glyce- 1,800 staff nurses with varied work sched- measures, enhance customer satisfaction and mic targets and nutrition therapy with consis- ules was compounded by budget constraints increase staff retention: tent carbohydrates established the standard related to “non-productive” education time of practice for nurses entering the institution. • Establish an interdisciplinary Acute Care and the administration’s need to under- By incorporating relevant diabetes scenarios Diabetes Advisory Committee. stand and support the need for this training. on the new employees’ test, their level of Another challenge was “information acquisi- • Identify core competencies related to dia- knowledge could be assessed before they tion fatigue”–the constant introduction of betes management and TGC. left formal orientation. Preceptors on nurs- new information, equipment, processes and ing units also had their knowledge validated • Assess staff's current level of knowledge. change that leads to an inability to process before they modeled diabetes care for new and internalize the information. Based on • Ensure that the administration under- staff. the assessment of knowledge, a plan for edu- stands and supports the concept of train- ing to assure competency.

Table 1. Core Competencies for Insulin • Develop a plan for education that includes critical information and core competen- • Types, actions, peaks, duration of insulins cies, differentiates between “must know” • Intravenous insulin and “nice to know” information and evalu- • Basal-bolus subcutaneous insulin management ates learning outcomes. • Insulin storage and administration • Develop an educational program that • Skills-utilization of pens, devices, vials and syringes includes employee orientation, medica- • Carbohydrate counting with prandial bolus tion-knowledge tests, preceptor guide- • Correction scale insulin lines and curriculum, annual competency • Hypoglycemia prevention testing, continuing education, resource or remedial support and, where applicable, a

Executive Summary Conference Report 77 7th Invited Conference: Nursing Education and Intensive Insulin Therapy

School of Nursing curriculum review and Table 2. Example of Competency Questions strong graduate student preceptorships.

• JP is a 58-year-old male with Type 2 DM S/P a right knee replacement. JP is one day • Use creative and innovative approaches post-op and has experienced some nausea through the night. JP did not receive his for continuing education on the basics of NPH insulin last evening because he had been NPO. This morning, his blood glucose diabetes management, special-population pre-breakfast is 388 mg/dL. JP’s orders include: needs, research and advanced practice − Clear liquid diet across the continuum of care. − Low consistent carbohydrate (1200-1500 calories, 3 CHO units per meal) − Glucose monitoring ac meals and HS − NPH insulin 22 units daily at 2000 − Aspart insulin prandial bolus-fixed; 3 units aspart subcutaneously per meal − Aspart insulin correction bolus with each meal based on pre-meal blood glucose: · BG 120-149 mg/dL give 2 units Aspart SC · BG 150-199 mg/dL give 3 units Aspart SC · BG 200-249 mg/dL give 4 units Aspart SC · BG 250-299 mg/dL give 7 units Aspart SC · BG 300-349 mg/dL give 10 units Aspart SC · BG > 350 mg/dL give 12 units Aspart SC Questions: • Plan a 3 CHO clear liquid breakfast tray for JP. • How much aspart insulin will JP receive this morning? • When will the aspart be administered? • Should JP have received his NPH last evening?

BG = Blood glucose

78 Executive Summary Conference Report 7th Invited Conference: Applying Glucometrics to Tight Glycemic Control

PROCEEDINGS Applying Glucometrics to Tight Glycemic Control Jacqueline Thompson, MAS, RN, CDE, Director, Diabetes Service Line, Sharp Healthcare, San Diego, CA

Key points sis of the data including mean and stan- dard deviation and reported “extreme” data • "Glucometrics” is the assessment of glucose control that includes four measures: glycemic (blood glucose < 60 and > 300 mg/dL). levels, efficacy of control, rate of adverse events and the impact on clinical outcomes. Glucometrics were developed for use with • Our goal was to improve glycemic control without increasing the risk of hypoglycemia. all patients admitted to the intensive care

• At baseline 33% of patients had poor glucose control, defined as at least one daily episode unit (ICU) that show the distribution of blood where blood glucose < 65 mg/dL or > 200 mg/dL, and 18.2% experienced extreme blood glucose levels. Extreme values have been glucose excursions ( > 300mg/dL). modified to blood glucose > 180 mg/dL or < 50 mg/dL. • Glucometrics were developed for patients in the intensive care unit (ICU). Extreme values were modified to blood glucose > 180 mg/dL or < 50mg/dL. Results were shared each month at all levels of the organization from executive • Results were shared on a monthly basis at all levels of the organization. leadership, who receive a report summariz-

• From 2002 to 2006, a 30% reduction in out-of-control days was realized, suggesting that ing all Sharp hospitals, to individual hospitals, the use of Glucometrics can help improve clinical performance and patient outcomes. which receive results benchmarked against all other hospitals within Sharp, and to indi- vidual patient care units. More information Although glycemic target ranges have and point-of-care (POC) blood glucose data can be obtained from the data-rich reposi- been recommend for hospitalized acute- and were obtained from Sharp’s clinical informa- tory to evaluate population subsets such as intensive-care patients, no consensus exists tion system for calendar year 2002. Patients trauma, cardiac and renal transplant patients as to how a single hospital or hospital system were identified by DRG 250, which is a sec- or physicians and medical groups. can gauge its clinical performance compared ondary diagnosis of diabetes. We included all to others. “Glucometrics” is a way of assessing laboratory and bedside blood glucose values During the four-year period from 2002 to glucose control that includes four measures: and double-checked all glucose levels >600 2006, a 30% reduction in out-of-control days glycemic levels, efficacy of control, rate of mg/dL for coding error or missed diagnosis. has been realized from using Glucometrics to adverse events and the impact on clinical It was found that 33% of patients had poor evaluate glucose management of inpatients. outcomes. control, defined as at least one episode where This suggests that providing timely and accu- blood glucose was < 65 mg/dL or > 200 mg/ rate clinical performance data to frontline At Sharp, we initially set out to improve dL. It was also noted that 18.2% experienced clinical teams can help improve clinical per- the glycemic control of patients with a sec- extreme blood glucose excursions, which was formance and patient outcomes. ondary diagnosis of diabetes without increas- defined as blood glucose > 300 mg/dL. ing the risk for hypoglycemic events—the concern most frequently expressed by Sharp refined its Glucometrics for our medical staff with regard to achieving the acute-care setting to include a “Well- improved glycemic control. To determine our managed Day” (adequately monitored and baseline for glycemic control, all laboratory not out of control), added statistical analy-

Executive Summary Conference Report 79 Roundtable

Intensive Insulin Therapy for Tight Glycemic Control

Philip J. Schneider, MS, FASHP Editor

The value of a conference of this type is not only the outstanding information presented in the lectures, but also discussing the application of this information among the distinguished presenters and invited guests. While consensus seems to be emerging among experts in this field, much still needs to be done to reach agreement on issues like the impact on hospital costs, a reasonable target for treatment outcome, methods, the administrative aspects needed to achieve tight glycemic control, the risks and consequences of hypoglycemia and the best method to measure blood glucose. What follows are some of the comments from the discussions at this roundtable about these six topics. Participants shared their insights about how intensive insulin therapy might actually be used to achieve tight glycemic control and debated issues where consensus has yet to be reached.

80 Executive Summary Conference Report Roundtable

What is the impact the need to start hemodialysis or hemo- What is a reasonable target for of intensive insulin therapy filtration. You can get your patient off blood glucose control? of the ventilator more quickly. It might (IIT) on hospital costs? The range of 80-110 mg/dL has been even save time if you have a good pro- My administrative colleagues will ask, touted as a goal, but if you look at the tocol and also if you could work with “What is this going to cost? What’s the clinical experience to date, patients are continuous monitoring. upfront cost?” I’m not talking about hos- not in that range for very long—in some Christophe De Block, MD pital length of stay reductions and all cases less than half the time, or perhaps those other important things, because We were interested in implementing 40%. the administrators, at least at my institu- continuous monitoring for tight glyce- Guy Soo Hoo, MD tion, don’t seem to be particularly fazed mic control at our hospital, so I did look If patients don’t have diabetes, our or impressed by those data. They want into the cost. I was quoted a cost of $5 glycemic range is about 80-130 mg/L; to know, “How much is this machine or per day/per bed. The cost is not based on 80 fasting and maximum 130. We think device or computer going to cost me? whether your patient is using the device, patients shouldn’t be outside the nor- What’s the capital investment up front?” it’s based on the cost of your beds. mal range of blood glucose variation I need that answer if I’m going to con- Jacqueline Thompson, MAS, RN, CDE that non-diabetics have within a 24-hour vince them to buy. Otherwise they’re We looked into that, too. The server period. going to say, “Sorry, too much money.” alone was $12,000, and then we looked Kalman Holdy, MD John P. Kress, MD at the cost per ICU bed. For our institu- There is huge inter-patient variation I’d be happy to answer “What is this tion, the cost was about $50,000. in how people respond. When a patient going to cost?” One ventilator-associated Kris Hedges, MBA is really sick, blood sugar varies enor- pneumonia for you…for the whole hos- The savings is on the order of $20 to mously. pital. $30 per bed, when you look at the reduc- Simon Finfer, MB, BS, FRCP, FRCA, FJFICM W. Patrick Burgess, MD, PhD tion in the number of infections and We have a commercially available IIT might be time-saving, because if length of stay. computerized program for insulin infu- you achieve tight glycemic control, you W. Patrick Burgess, MD, PhD sion dosing and keep our patients at can avoid 40% of acute renal failures and goal about 60% of the time. We get our patients to goal in about six hours at 80-110 mg/dL. Judith Jacobi, PharmD, FCCM, FCCP, BCPS

Executive Summary Conference Report 81 Roundtable

What factors complicate glycemic by standardizing the site where blood for minor; it’s not two to three hours a day control in critically ill patients? glucose testing is obtained, we were able per patient at all. A lot of the work is to obtain the goal of 80-150 mg/dL in done by nursing partners, so the time to One factor is how the pharmacy pre- 80% of tested blood glucose values, with do a glucose, get it back to the nurse and pares a product. If the insulin infusion is a hypoglycemia rate, i.e., blood glucose input it would be probably three to five always mixed in 5% dextrose, a 50-mL < 40 mg/dL, less than 1%. minutes. bag of 5% dextrose isn’t going to make Rhonda S. Rea, PharmD Bruce W. Bode, MD a big difference. But how many patients also get antibiotics or other medications The glycemic control diet is really We have a training tape on the web that are provided in dextrose-based solu- based on the source of hyperglycemia that teaches the nurses how to use bed- tions? in the majority of patients who are ill. side testing, so that an institution can Judith Jacobi, PharmD, FCCM, FCCP, BCPS In the majority of cases, the reason that train their new nurses by using the web- patients are hyperglycemic is not exog- based tutorials. We find that program- Another factor is stopping TPN imme- enous nutrition or glucose; it is endog- ming the time into the software that diately. You can’t adjust for that. enous production of glucose in the liver. about 40% of the time, the Accu-Cheks Bruce W. Bode, MD Kalman Holdy, MD are every two hours. This reduces 40% Steroids certainly are a big question of the fixed cost of the disposables from with hyperglycemia that we encoun- What are the administrative the point-of-care device because of less ter in most of our patients in the ICUs. aspects of an IIT program? frequent testing. That also saves the five Generally the steroids are given as bolus minutes times the 40%. We find that on Right now with our computerized sys- doses in the ICU. There’s a lag between average, somewhere between 14 and 16 tem, every nurse is trained in a 15-minute the time the steroids are administered, checks a day will keep a patient in con- segment upon enrollment in the hospi- and a different lag in terms of insulin trol. tal. All our glucoses are done by medical coverage. assistants or medical techs. Except in the The decision-making takes some- Kalman Holdy, MD ICU on a one-to-one ratio, the nurse still where between 10 and 15 seconds. The We start with a somewhat higher is very much involved in it. There’s really nurse walks up to a computer and walks target blood glucose range of 80-150 mg/ almost no time involved except to set away with an answer. It’s a few keystrokes dL. By starting with a higher target range up the device, which takes about three and they get used to doing it. The fewer and mixing intermittent medications, minutes. So the true nursing time is very keystrokes they use, the faster they get. such as antibiotics, in normal saline, and W. Patrick Burgess, MD, PhD

82 Executive Summary Conference Report Roundtable

The treatment of hypoglycemia is I believe a fair number of low glu- We also have to take into account extremely labor-intense—with repeat- cose levels are not actually hypoglyce- the duration of the hypoglycemia. If you ing glucoses every 15-20 minutes and mic events, because when the glucom- measure glycemia on a continuous basis, then implementing whatever interven- eter reading is checked against the lab see it drop off to below 40, and correct it tion is called for per your protocol. results, a high proportion of the glucom- immediately, there would probably not R. Daniel Pollom, MD eters are under-reading. But the majority be a problem. But if you don’t monitor of the glucometer readings are not being glycemia continuously or frequently and We have a shortage of nurses. We’re checked. your patient stays below 40 for more using 600 registry, or travel nurses in Simon Finfer, MB, BS, FRCP, FRCA, FJFICM than 30 minutes, then you have a prob- our facilities and retraining isn’t going to lem. help us. We instituted a policy so that if Hospital meters err on the side of Christophe De Block, MD there is a low blood glucose reading and being lower than laboratory values as a the nurse does not validate it and ask protective mechanism. That’s why, when We are currently doing a study in whether the patient really looks hypo- you send your data to the lab, very few of older type 2 diabetics with pacemakers, glycemic, another nurse has to perform the readings were truly less than 40; they and taking them from 150 mg/dL down a double-check. were trending towards 60 but trending to 50 at a certain rate, then taking them Jacqueline Thompson, MAS, RN, CDE toward the low range. back up to 150 and down to 50 at a faster Bruce W. Bode, MD rate, and people develop pretty signifi- What is the risk of cant CNS symptoms between 60 and 65. The hypoglycemia risk does seem hypoglycemia with IIT? When you get them down to 50, they higher in that setting, if you’re not giving often stop talking and they’re “spacey.” The hypoglycemic rate using a com- some basal glucose. We leave them at 50 for about half an puter, EndoTool™ or Glucomander®, to Judith Jacobi, PharmD, FCCM, FCCP, BCPS hour, and we’re looking to see if the EKG calculate the infusion rate is less than The gold standard test of pituitary changes. 1%, not only in the institutions where function is an insulin tolerance test. And they were promoted and designed but The real issue is permanent injury, and the goal sugar is below 40 mg/dL. It’s also in other institutions. In the NICE- that’s fairly rare. a diagnostic test to produce a pituitary SUGAR trial, the hypoglycemia rate is 5%. Jeffrey I. Joseph, DO response, and the goal is to get the Why would there be a difference in the blood sugar transiently below 40. There’s rates of hypoglycemia? a large body of experience over 30 years Chris Hogness, MD, MPH with that particular test showing no adverse effects from transient hypogly- cemia. Robert Osburne, MD Executive Summary Conference Report 83 Roundtable

It would be a real shame if people keep someone between 80 to 110 mg/ issue. If you’re squeezing the blood, that went away from this conference thinking dL, you need to know that your measure- does alter the measurement. However, that there was any credible evidence at ment is accurate. Checking glucometers I’m finding much more of an issue from all that one blood glucose of less than against laboratory reference values is sampling from a radial arterial line or a 40 was harmful. Because I don’t believe very important, because glucometers are venous line. Everybody does it a little dif- there is any credible evidence. subject to a lot of error. ferently. One, there’s a lot of phlebotomy Simon Finfer, MB, BS, FRCP, FRCA, FJFICM Heather Pidcoke, MD involved, and two, if there’s even a little bit of residual in the stopcock or varia- We had an issue with the glucometers What is the best way to tion in how you’re doing it, this can result at our hospital as well; we were getting measure blood glucose in a low or high dilutional error. some erroneous readings. And when we concentrations for IIT? Jeffrey I. Joseph, DO investigated it further, the blame was I think the quality of the technology placed on nursing, as if their technique Regarding the continuous glucose at the bedside for measuring glucose were inappropriate. But the truth was monitoring, I’m sure when we have is a really important point. At our unit, that if an adequate blood sample wasn’t results real-time and on a computer wenoticed that we had a 20% to 30% dif- applied, if it didn’t wick completely and screen, it will help. But another huge ference between our glucometer results cover the yellow area completely, the advantage is that we can pull so much and laboratory values. meter was capable of giving an errone- data out of these devices — areas under

The glucometers were telling us that ous reading by 54%, high or low. the curve, percentage of time in certain the patient’s values were fine, when in Jacqueline Thompson, MAS, RN, CDE glycemic ranges and indexes of variabil- fact the patient was hypoglycemic. We ity. I’ve heard a lot of criticism of the did a quality analysis and realized that, Christophe De Block, MD meters, but we often will do a sample atleast in our population, the major rea- in duplicate or triplicate with the same When I tested our glucometers, I lined son for the error was a high prevalence of meter and have a fairly small standard up five and did them all at the same time. anemia in our patients. deviation. The bigger issue is getting a They’re very consistent with each other. We tested four different glucometers, quality sample. The size of the drop on The problem is they're precise but not and all of them had an approximately the fingerstick is an issue, I agree, but we accurate. When I would send the sample 20% error rate. The manufacturer tells us, actually had the smaller standard devia- off to the lab, the glucometer results all “Well, 15% is okay.” That was true back tion from the fingersticks. showed the same amount of difference in the era of trying to keep diabetics from the lab value. If a patient is in shock and has low < 200mg/dL, but when you’re trying to Heather Pidcoke, MD blood pressure and cold hands, that’s an

84 Executive Summary Conference Report Appendix

Table. Major Published Randomized Controlled Trials with Insulin Therapy in Critically Ill Patients From: Pittas AG. Meta-analysis of Randomized Trials of Tight Glycemic Control (on pg 21)

Population Insulin-based Glucose Goal N Outcomes Intervention (mg/dL) Malmberg, 1995 Diabetes Intensive Acute and 126 – 180 306 No difference between groups in: (DIGAMI) and Acute Intensive Long-term Acute (in-hospital) mortality Myocardial Therapy Infarction vs. None 314 Reduction with Intensive Therapy in: Routine Therapy Long-term (outpatient) mortality, 26% Van den Berghe, 2001 Adult Intensive IV Therapy 80 – 110 765 In-hospital reductions with Intensive Surgical ICU vs. Therapy in: (63% cardiac Routine Therapy 180 – 200 783 Mortality, 34% surgery Blood-borne infections, 46% patients) Acute renal failure, 41% Transfusions, 50% Critical illness polyneuropathy, 44%

Malmberg, 2005 Type 2 Intensive Acute and 126 – 180 (Acute) 474 No differences between groups in: (DIGAMI 2) Diabetes Intensive Long-term 90 – 180 (Long-term) Acute (in-hospital) mortality and Acute Therapy Long-term (outpatient) mortality Myocardial vs. 126 – 180 (Acute) 473 Infarction Intensive Acute and None (Long-term) Routine Long-term Therapy None (Acute) 306 None (Long-term) vs. Routine Acute and Long-term Therapy Mehta, 2005 Acute Glucose-Insulin- None 10,110 No differences between groups in: (CREATE-ECLA) Myocardial Potassium (GIK) IV 30-day mortality Infarction infusion x 24 hours Cardiac arrest vs. None 10,091 Cardiogenic shock Routine Therapy Reinfarction Heart failure

Van den Berghe, 2006 Adult Intensive IV Therapy 80 – 110 595 No differences between groups in: Medical ICU vs. Acute (in-hospital) mortality Routine Therapy < 180 605 In-hospital reduction with Intensive Therapy in: Patients staying in the ICU for 3 or more days In-hospital improvements with intensive therapy in: Acquired kidney disease Weaning from mechanical ventilation Hospital discharge

Executive Summary Conference Report 85 The CareFusion Center for Safety and Clinical Excellence Invited Conference Intensive Insulin Therapy for Tight Glycemic Control June 7-8, 2007

ATTENDEES Chris Hogness, MD, MPH R. Daniel Pollom, MD Daleen Aragon, PhD, CCRN, FCCM Chair, Multidisciplinary Glycemic Control Committee Community Physicians of Indiana Director, Advanced Practice Nursing and Research Southwest Washington Medical Center Indianapolis, IN Vancouver, WA Orlando Regional Healthcare Richard C. Prielipp, MD, MBA, FCCM Orlando, FL Kalman Holdy, MD, ABPNS JJ Buckley Professor and Chair, Timothy S. Bailey, MD, FACE, FACP, CPI Medical Director of Clinical Nutrition Department of Anesthesiology President, Advanced Metabolic Care and Research Sharp Memorial Hospital University of Minnesota Escondido, CA San Diego, CA Minneapolis, MN Sean Berenholtz, MD, MHS, FCCM Susan Jacob, PharmD Rhonda S. Rea, PharmD Assistant Professor, Department of Anesthesiology Resident in Pharmacy Practice Assistant Professor, Pharmacy and Therapeutics and Critical Care Medicine Loma Linda Veterans Affairs Health Care System University of Pittsburgh, School of Pharmacy Johns Hopkins University Loma Linda, CA Pittsburgh, PA Baltimore, MD Judith Jacobi, PharmD, FCCM, FCCP, BCPS John P. Santell, MS, RPh, FASHP Glenn J. Bingle, MD, PhD, FACP Critical Care Pharmacist Director, Practitioner Programs and Services Vice President, Medical and Academic Affairs Methodist Hospital/ Clarian Health U.S. Pharmacopeia Community Health Network Indianapolis, IN Rockville, MD Indianapolis, IN Jeffrey I Joseph, DO Philip J. Schneider, MS, FASHP Bruce W. Bode, MD, FACE Director, Artificial Pancreas Center Clinical Professor and Director, Atlanta Diabetes Associates Department of Anesthesiology Latiolais Leadership Program Atlanta, GA Jefferson Medical College of Thomas Jefferson College of Pharmacy University The Ohio State University W. Patrick Burgess, MD, PhD Philadelphia, PA Columbus, OH Carolinas Medical Center Charlotte, NC Janet L. Kelly, PharmD Judy Smetzer, RN, BSN Outcome and Cost Management Pharmacist Vice President Elaine Button, RN University of Washington Medical Center Institute for Safe Medication Practices Director of Inpatient Diabetes Services for Healthways Seattle, WA Huntingdon Valley, PA Moses Cone Health System Greensboro, NC Genna Waldman Klein, MD Virginia Smith, RN, NP, CDE Fellow, Pediatric Endocrinology and Diabetes Critical Care Educator Alan J. Conrad, MD Mount Sinai School of Medicine Huntington Hospital Medical Director, Diabetes Health New York, NY Huntington, NY Palomar Pomerado Health Poway, CA John P. Kress, MD Guy W. Soo Hoo, MD, MPH Assistant Professor of Medicine Pulmonary and Critical Care Section Christophe De Block, MD, PhD University of Chicago VA Greater Los Angeles Health Care System Department of Diabetology Chicago, IL Los Angeles, CA Antwerp University Hospital Belgium James Krinsley, MD, FCCM, FCCP (via tape) Ying P. Tabak, PhD Director, Critical Care Director, Biostatistics Philippe Devos, MD, PhD (student) Stamford Hospital MediQual CareFusion Department of General Intensive Care Stamford, CT Marlborough, MA University Hospital of Liege Belgium Carol Manchester, MSN, APRN, BC-ADM, CDE Jacqueline Thompson, MAS, RN, CDE Diabetes Clinical Nurse Specialist Director, Diabetes Service Line Simon Finfer, MB, BS, FRCP, FRCA, FJFICM University of Minnesota Medical Center, Fairview Sharp HealthCare Professor, Faculty of Medicine Minneapolis, MN San Diego, CA University of Sydney, Sydney, Australia Director, Critical Care and Trauma Greg Maynard, MD, MS Nam K. Tran George Institute for International Health Associate Clinical Professor of Medicine and Chief Assistant Director, Point-of-Care Testing Center for UCSD Division of Hospital Medicine Teaching and Research Simon Finney, MBChB, MRCP, FRCA, PhD San Diego, CA University of California-Davis Consultant in Intensive Care and Anaesthesia Davis, CA Royal Brompton Hospital Joseph E. Mazur, PharmD, BCPS, BCNSP London, UK Clinical Pharmacy Manager, Clinical Associate Tim Vanderveen, PharmD, MS Professor, Medical University of South Carolina Vice President, Center for Safety and Tony Furnary, MD (via video conference) Charleston, SC Clinical Excellence Starr-Wood Cardiac Group CareFusion Portland, OR Karla Miller, PharmD, BCPP San Diego, CA Director of Medication Usage and Safety Michael Gottschalk, MD, PhD HCA Quality Department John R. White, Jr., PA, PharmD Chief, Pediatric Endocrinology Nashville, TN Professor, Dept. of Pharmacotherapy, UCSD/Rady Children’s Hospital College of Pharmacy San Diego, CA Robert C. Osburne, MD Washington State University Spokane and Endocrinology Spokane, WA Frances A. Griffin, RRT, MPA Atlanta Medical Center Director, Institute for Healthcare Improvement Atlanta, GA Cambridge, MA Heather Pidcoke, MD CareFusion Karl F. Gumpper, RPh, BCNSP, BCPS, FASHP US Army Institute of Surgical Research 1. Sherri Almeida, DrPH, MSN, RN, CEN, FAEN Director, Section of Pharmacy Informatics Army Burn Center and Technology Fort Sam Houston, TX 2. Charlie Ham, PharmD American Society of Health-System Pharmacists Bethesda, MD Anastassios G. Pittas, MD, MS 3. Stephen R. Lewis, MD Assistant Professor of Medicine Valarie Harmon, RN, BSN, CCRN, ICU/PACU RN Division of Endocrinology, Diabetes and Metabolism 4. Carl Peterson, PharmD Atlanta Medical Center Tufts-New England Medical Center 5. Curt Quap, MS, RPh Atlanta, GA Boston, MA 6. Michael Yang, MD Kris Hedges, MBA Anne Pohlman, RN, MSN, CCRN System Director, DiabetesHealth Critical Care Clinical Research Palomar Pomerado Health University of Chicago San Diego, CA Chicago, IL

SC2601 0311/PDF