 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
STATE-OF-THE-ART REVIEW ARTICLE |
a Departments of Surgery
b Medicine
c Anesthesia
d Cardiovascular Surgery
e Cardiology, Children's Hospital Boston, Harvard Medical School, Boston, Massachusetts
ABSTRACT
OBJECTIVES. Given the demonstrated benefit of euglycemia in critically ill patients as well as the risk for hypoglycemia during insulin infusion in children, we sought to validate a subcutaneous sensor for real-time continuous glucose monitoring in pediatric patients during and after cardiac surgery.
METHODS. Children up to 36 months of age who were undergoing cardiac bypass surgery were recruited. After anesthetic induction, a continuous glucose-monitoring system sensor (CGMS, Medtronic Minimed, Northridge, CA) was inserted subcutaneously. Sensors remained in place for up to 72 hours. Arterial blood glucose was measured intermittently in the central laboratory (Bayer Rapidlab 860, Tarrytown, NY). Sensor data, after prospective calibration with 6-hourly laboratory values using the proprietary Medtronic Minimed Guardian RT algorithm, were compared with all laboratory glucose values. Statistical analysis was performed to test whether sensor performance was affected by body temperature, inotrope dose, or body-wall edema.
RESULTS. Twenty patients were enrolled in the study for a total of 40 study days and 246 paired sensor and laboratory glucose values. Consensus error grid analysis demonstrated that 72.0% of sensor value comparisons were within zone A (no effect on clinical action), and 27.6% of comparisons were within zone B (altered clinical action of little or no effect on outcome), with a mean absolute relative deviation of 17.6% for all comparisons. One comparison (0.4%) was in zone C (altered clinical action likely to affect outcome). No significant correlations were found between sensor performance and body temperature, inotrope dose, or body-wall edema. All patients tolerated the sensors well without bleeding or tissue reaction.
CONCLUSIONS. Guardian RT real-time subcutaneous blood glucose measurement is safe and potentially useful for continuous glucose monitoring in critically ill children. Subcutaneous sensors performed well in the setting of hypothermia, inotrope use, and edema. These sensors facilitate identifying and following the effects of interventions to control blood glucose.
Key Words: glucose monitoring • cardiac surgery • hyperglycemia
Abbreviations: CGMS—continuous glucose-monitoring system • CICU—cardiac ICU • MARD—mean absolute relative difference
The benefits of glucose control in the perioperative period are becoming increasingly clear for both diabetic and nondiabetic patients. Clinical trials in the adult surgical population have demonstrated 35% to 53% reductions in mortality and even more significant reductions in morbidity.1, 2 Although similar studies have not been published in the pediatric patient population, critically ill infants and children may also benefit from glycemic control while in the ICU and during the perioperative period.3
One of the major obstacles encountered in designing a pediatric intervention trial for tight glycemic control is the risk of hypoglycemia from insulin administration. Infants and young children often have reduced glycogen reserve, especially in the setting of physiologic stress,4 and insulin therapy in this setting may place them at risk for developing hypoglycemia. Adult studies have demonstrated a significant increase in the incidence of hypoglycemia among patients randomly assigned to tight glucose control ranging from 34% to 39%.2, 5 In critically ill children, hypoglycemia can be difficult to detect clinically and, if not treated immediately, can lead to brain damage or death, particularly in the young, developing child.6
Given the potential risks of establishing strict glycemic control in pediatric patients, the use of a continuous glucose monitor is the only viable option to ensure maximal safety. The US Food and Drug Administration approved the continuous glucose-monitoring system (CGMS, Medtronic Minimed, Northridge, CA) in 1999 for clinical use. The CGMS relies on the interstitial fluid for measuring blood glucose. This device is the first such system to be available worldwide and has been used in multiple published studies, including use in adult patients undergoing cardiac surgery.7 However, it is a retrospective monitor, because data are collected over a 72-hour period and then correlated with values obtained from the patient's hand-held glucose meter, which is considered the gold standard. In 2005, a new real-time glucose monitor became available: Guardian RT (Medtronic Minimed). Although it uses the same hardware as the CGMS, the data from the sensor are interpreted in real time, and glucose concentration is displayed live every 5 minutes.
It is unknown whether glucose concentrations in the interstitial fluid remain in equilibrium with blood concentrations during conditions that could compromise subcutaneous tissue perfusion. Infants undergoing congenital cardiac surgery are ideal for study in this regard. They are subject to changes in cardiac output, body temperature, and edema related to both cardiopulmonary bypass and the induced systemic inflammatory response. In addition, these infants have altered myocardial physiology related to the underlying disease and surgical procedure itself.
This study was designed to evaluate the real-time glucose sensor under varying physiologic conditions in a cohort of neonates and infants undergoing pediatric cardiac surgery with the hypothesis that the glucose level measured by the Guardian RT will be accurate when compared with the gold standard of direct blood glucose measurement.
METHODS
Patient Enrollment
The study was approved by the Children's Hospital Boston institutional review board, and written informed consent was obtained from the parent or guardian of each patient. We prospectively enrolled a convenience sample of children <36 months of age who weighed at least 2 kg and were referred for elective cardiac surgery.
CGMS
The Guardian RT is a CGMS composed of 3 parts: a subcutaneous glucose sensor, a battery-powered monitor, and a Com-Station for downloading stored data to a computer. The sensor is a platinum microelectrode coated with glucose oxidase and covered with a semipermeable membrane. Ambient glucose from interstitial fluid interacts with the enzyme to produce an electric current proportional to glucose concentration. The sensor is inserted through a small needle into the subcutaneous tissue (most often in the abdomen or lower extremity) and then attached to the small monitor by a flexible cord (Fig 1). The monitor receives an electric signal from the sensor every 10 seconds and stores a mean glucose measurement every 5 minutes. The data can be downloaded to a computer for review at any time by using the CGMS Com-Station and software.
|
Calibration
The live glucose readings were calculated by using the Guardian RT proprietary prospective interpretive algorithm developed by Medtronic Minimed. This algorithm required calibration using arterial values at least every 12 hours, and for this study, calibration values were entered every 6 hours. After calibration, all sensor values were compared with time-matched arterial samples. Arterial values used as part of the calibration algorithm were compared with the sensor value 5 minutes before the time of arterial measurement. Real-time glucose concentrations were not available to the bedside clinical team and were not used to alter clinical management.
Sensor Performance Under Physiologic Stress
The following parameters were used to evaluate the performance of the Guardian RT under different physiologic conditions:
Statistical Analysis
The mean absolute relative difference (MARD) as well as the Pearson correlation coefficient was calculated for all matched Guardian RT and reference arterial values, excluding points used for initial calibration. In addition, all matched points were plotted on a Clarke error grid.10 Clarke error analysis was designed specifically to evaluate the accuracy of capillary blood glucose testing systems for ambulatory diabetics. The graph is divided into 5 zones. Comparison points within zone A represent CGMS and reference values that differ from each other by no more than 20%. Zone B includes comparison points that differ by >20% but do not result in an alteration in treatment. Comparison points in zone C would result in an overcorrection of an acceptable glucose value, and those in zone D would not result in correction when, in truth, treatment should be administered. Comparisons in zone E would prompt inverse treatment. All comparison points were also plotted on a consensus error grid, which is similar to the Clarke error grid but is based on the consensus of 100 endocrinologists.11 The 5 defined risk categories include A (no effect on clinical action), B (altered clinical action with little or no effect on outcome), C (altered clinical action likely to affect outcome), D (altered clinical action that could have significant medical risk), and E (altered clinical action that could have dangerous consequences). Sensor performance was correlated with the temperature, inotrope score, and index of body-wall edema using the Spearman correlation coefficient with significance at P < .05. MARDs under different conditions were compared by using the Wilcoxon rank-sum test with significance at P < .05.
RESULTS
A total of 20 patients were enrolled, and all completed the study (Table 1). All patients underwent cardiac surgery, and 16 required cardiopulmonary bypass. The Guardian RT was in use for an average of 48 ± 19 hours for each patient (Fig 2). Although each sensor can be used for up to 72 hours, the most common indication for premature removal in our study was transfer of the patient from the CICU to the cardiac ward.
|
|
Overall Performance
Throughout the study a total of 246 blood glucose values were compared between the sensors and our arterial gold-standard values. The number of comparison points obtained for each sensor in the study ranged from 1 to 26, with an average of 12 comparisons per sensor. The overall MARD between sensor glucose values and arterial samples was 17.6%; the Pearson's correlation coefficient was 0.787 (P < .001). However, because some of the sensors had very few comparison points, we also analyzed the pooled data for all sensors tested. When comparison points were plotted on the traditional Clarke error grid, 66.3% of values were in zone A, 32.5% in zone B, 1.2% in zone D, and 0% in zones C or E (Fig 3). When using the consensus error grid, 72.0% of points were found in zone A, 27.6% in zone B, and 0.4% in zone C (Fig 4).
|
|
Description of Clarke Zone D Comparisons
Sensor 12
This sensor overread an arterial sample of 63 mg/dL as 126 mg/dL, an MARD of 100.1%. This zone D comparison occurred 1 hour after infusion of intravenous KCl carried in 10% dextrose. The sensor glucose measurement peaked at 225 mg/dL near the time of KCl administration and was dropping at a rate of 1.7 mg/dL per min until the comparison, at which time the sensor self-deactivated because of the significant measurement discrepancy. The MARDs of the nearest comparisons before and after the D point were 0.4% and 5.5%, respectively.
Sensor 13
This sensor underread an arterial sample of 275 mg/dL as 156 mg/dL, an MARD of 43.2%. This zone D comparison did not correspond in time to any notable clinical events. The MARDs of the nearest comparisons before and after the D point were 19.0% and 28.9%, respectively.
Sensor 14
This sensor underread an arterial sample of 264 mg/dL as 111 mg/dL, an MARD of 57.9%. This was a zone D comparison on the Clarke error grid and a zone C comparison on the consensus error grid. This did not correspond in time to any notable clinical events. The nearest arterial glucose measurement before this comparison was obtained in the operating room and was used for an initial calibration; the nearest comparison afterward had an MARD of 9.9%.
Sensor Performance in the Operating Room
On the basis of paired glucose values obtained from the Guardian RT algorithm and arterial samples, we found that there was no significant difference in sensor performance in the operating room compared with postoperative performance. The MARD between sensor glucose values and arterial samples was 17.9% outside of the operating room (N = 203) compared with 16.6% during surgery (N = 43) (P = .37).
During 10 episodes in the operating room, however, the glucose-monitoring system produced an error and sounded an alarm because of electrical signals received from the sensor that were outside the range of expected performance. The cause of these malfunctions was not confirmed but correlated in time to use of electrocautery by the operating surgeon. As a result, there were 10 of 20 patients in whom continuous glucose monitoring was not fully successful intraoperatively. In all of these instances, monitoring in the ICU postoperatively was successful.
Correlation During Physiologic Stress
Sensors were tested in conditions of variable temperature, inotrope use, and body-wall edema. We found that within a range of 23.1 to 37.5°C there was no significant correlation between temperature and sensor performance (r = 0.55; P = .39). None of the patients were febrile during the study. The MARD for sensors during intraoperative hypothermia (N = 13) was 16.0% compared with 17.7% for sensors during normothermia (N = 233; P = .40). There was no significant correlation between sensor performance and inotrope score (r = –0.97; P = .13). The MARD for sensors in patients receiving inotropes (N = 169) was 15.9% compared with 21.4% in patients who were not receiving inotropic support (N = 77; P = .12). Finally, there was no significant correlation between sensor performance and the radiologic index of edema (r = 0.17; P = .40). The MARD for sensors in patients with edema (N = 25) was 17.4% compared with 18.9% in patients without edema (N = 30; P = .45).
Hyperglycemia in the Perioperative Period
A persistent elevation in blood glucose concentrations was documented postoperatively in our study patients compared with preoperative concentrations. Blood glucose peaked on the day of surgery and gradually declined over the next 48 hours (Fig 5). Average values calculated for all enrolled patients based on arterial blood measurements were 84 ± 24 mg/dL preoperatively, 157 ± 63 mg/dL on the day of surgery, 137 ± 53 mg/dL on postoperative day 1, and 114 ± 21 mg/dL on postoperative day 2.
|
DISCUSSION
We found that the Guardian RT real-time subcutaneous glucose monitor provided clinically reliable measurement when compared with blood glucose concentrations and that sensor performance remained reliable under conditions of hypothermia, inotrope use, and body-wall edema. Although the benefit of maintaining euglycemia in critically ill adults is compelling, the significant risk for hypoglycemia and the need for frequent blood sampling has been an argument against applying this strategy to neonates, infants, and young children. The performance of the Guardian RT in this study indicates for the first time that there is now a method for real-time, continuous glucose measurement that can be used in pediatric patients to enable tight glucose control during critical illness.
To date, there have been few published studies investigating the blood glucose trends in the pediatric population throughout critical illness. An initial study by Srinivasan et al3 retrospectively compared a cohort of survivors and nonsurvivors in the PICU and found that nonsurvivors had higher blood glucose values that persisted for significantly longer than survivors. Faustino and Apkon12 reported the prevalence of hyperglycemia in the PICU to be from 17% to 75% within the first 10 days of admission and identified a threefold increase in mortality over matched euglycemic controls. The increased risk of mortality was confirmed subsequently in a review of pediatric patients with septic shock, which demonstrated that peak glucose concentration is the only independent risk factor associated with mortality.13 In addition, a study by Beardsall et al14 found that neonates weighing <1500 g were hyperglycemic for a median of 36% of the time during the first 5 days of life. Our study has demonstrated that infants undergoing cardiac surgery develop hyperglycemia during the perioperative period. The available data suggest that pediatric patients are at risk of becoming hyperglycemic when undergoing physiologic stress and therefore may be exposed to similar risks as adult patients.
Continuous glucose-monitoring devices have been developed primarily for use by ambulatory diabetic patients to provide a short-term intensive view of glycemic control. However, the device's relatively noninvasive components and ease of use render it an attractive option for glucose monitoring in pediatric inpatients. We have demonstrated that the Guardian RT system can be safely used in young children in the perioperative period and that this prospective system performs at a level similar to previously published studies of retrospective calibration. The MARD between sensor readings and laboratory blood glucose values in this study was 17.6%, whereas other groups have reported a range between 16% and 19%.15–17 Beyond simple comparison of paired values, this system allows clinicians to accurately monitor the trend in blood glucose fluctuations, which may ultimately allow for titration of insulin infusions for tight glycemic control. The frequency of calibration used in this study is easily implemented in the setting of an ICU for patients in any age group.
Our study has also demonstrated that the Guardian RT can be used reliably under certain conditions of physiologic stress. During our study, the subcutaneous sensors were subjected to temperatures as low as 23.1°C, and although cardiac output was not measured directly, a wide range of inotrope scores were calculated during the perioperative period. The sensors continued to function accurately under these circumstances.
Continuous glucose monitoring with the Guardian RT was not without some limitations. Occasionally technical interruption of transmission from the sensor to the monitor occurred during sensor use in the operating room. Data collected during these episodes were not included in the final analysis, and therefore it is possible that there was some selection bias in the values used in the analysis. However, these episodes were associated with an audible system alarm that did not occur at any time during the study outside of the operating room. The cause of the signal errors while in the operating room has not been entirely elucidated. One possible explanation may be interference with the electrocautery used during the operation. In a recent study by Vriesendorp et al,18 the CGMS was tested in the operating room in adult patients. The authors found that technical failure occurred more often during surgery than after surgery (66% vs 18%), and they surmised that the high failure rate may have been a result of interference with electrical equipment in the operating room. Our findings would also be consistent with this. There were no episodes of transmission interruption during our study once patients were out of the operating room. This theory requires additional investigation, particularly as continuous glucose monitoring becomes more widespread.
There were also specific brief instances of divergence of the subcutaneous sensor and measured arterial values, leading to low correlations in 3 subjects (Table 2). In retrospect, we were not able to identify whether the sensor or arterial value more likely reflected the true glucose concentration in the patient's blood. This fact reinforces the importance, therefore, of confirming blood glucose concentration with a laboratory sample before acting on glucose concentrations reported by the sensor. This is a current requirement as part of the Food and Drug Administration approval of Guardian RT for clinical use.
|
ACKNOWLEDGMENTS
This research was supported by a grant from the Good Samaritan Foundation (to Drs Jaksic and Agus). The sensors and monitors were generously donated by Medtronic Minimed.
Data were analyzed by Dr Rebecca Gottlieb and Ms Nandita Patel according to the Guardian RT proprietary prospective interpretive algorithm developed by Medtronic Minimed.
FOOTNOTES
Accepted Apr 28, 2006.
Address correspondence to Michael S.D. Agus, MD, Children's Hospital Boston, 300 Longwood Ave, Main 9 South, Boston, MA 02115. E-mail: michael.agus{at}childrens.harvard.edu
Financial disclosure: Dr Agus is the recipient of unrestricted research grants by Medtronic Minimed to support different investigator-initiated research projects. Monitors and sensors for this project were donated by Medtronic Minimed.
REFERENCES
This article has been cited by other articles:
 |
|
 |
|
  H. E. Skillman and P. E. Wischmeyer Nutrition Therapy in Critically Ill Infants and Children JPEN J Parenter Enteral Nutr, September 1, 2008; 32(5): 520 - 534. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
