DIABETES TECHNOLOGY & THERAPEUTICS

Volume 6, Number 3, 2004

Mary Ann Liebert, Inc.

Experience with the Continuous Glucose Monitoring

System in a Medical Intensive Care Unit

PHILIP A. GOLDBERG, M.D.,1 MARK D. SIEGEL, M.D.,2 RAYMOND R. RUSSELL, M.D.,3

ROBERT S. SHERWIN, M.D.,1 JOSHUA I. HALICKMAN,1 DAWN A. COOPER, M.S., R.N.,4
JAMES D. DZIURA, Ph.D.,5 and SILVIO E. INZUCCHI, M.D.1

ABSTRACT

Strict glycemic control improves clinical outcomes in critically ill patients. However, practi-
cal tools for frequent monitoring of blood glucose (BG) levels in the intensive care unit (ICU)
are limited. The Continuous Glucose Monitoring System (CGMS, Medtronic MiniMed,
Northridge, CA) is currently approved for detecting glycemic excursions in outpatients with
diabetes mellitus. The use of this device has never been carefully examined in the inpatient
setting. This preliminary study was designed to investigate the accuracy of the CGMS in crit-
ically ill patients admitted to a medical ICU (MICU). Subjects at risk for hyperglycemia were
recruited from among all patients admitted to our MICU. CGMS sensors were implanted for
up to 72 h. Study subjects wore between one and five consecutive sensors. Four or more stan-
dard capillary BG readings were recorded per 24 h. All paired metersensor (M-S) readings
were used both for CGMS calibration and for data analysis. Twenty-two MICU patients wore
41 CGMS sensors, yielding 546 M-S BG pairs. Overall, the Pearson correlation coefficient (r)
was 0.88, with a mean M-S difference of 3.3 6 26.7 mg/dL (0.6 6 17.4%) and a mean absolute
M-S difference of 19.7 6 18.3 mg/dL (12.8 6 11.9%). Clarke Error Grid analysis categorized
98.7% of the M-S pairs within clinically acceptable zones A and B. The CGMS is promis-
ing for potential use in critically ill patients. If validated in larger studies, the device could
serve as a useful research tool for investigating the role of hyperglycemia (and strict glyce-
mic control) in ICU patients. If further developed as a real-time glucose sensor, CGMS tech-
nology could ultimately prove clinically useful in the ICU, by decreasing nursing workload
and/or by providing alarm signals for impending glycemic excursions.

BACKGROUND infusion protocol (IIP) to maintain strict nor-

moglycemia (80110 mg/dL) reduced mortal-

M ETICULOUS CONTROL of blood glucose (BG)

concentrations reduces morbidity and
mortality in critically ill patients. In Leuven,
Belgium, the use of an intravenous (IV) insulin
ity in a surgical intensive care unit (ICU) by
42%.1 In this landmark study, strict BG control
also lowered the incidence of bloodstream in-
fections, acute renal failure, anemia requiring

Sections of 1 Endocrinology, 2Pulmonary & Critical Care, and 3Cardiovascular Medicine, Department of Internal
Medicine; 4 Department of Nursing; and 5General Clinical Research Center, Yale New Haven Hospital, Yale Univer-
sity School of Medicine, New Haven, Connecticut.
The data presented in this paper were presented (in preliminary form) as a poster at the 63rd Scientific Sessions of
the American Diabetes Association, held in New Orleans, Louisiana, in June 2003.

339
340
blood transfusions, and the need for prolonged
ventilatory support. Strict glycemic control ap-
pears to have clinical benefits in other intensive
care settings as well. Following myocardial
infarction, for example, the use of an IV in-
sulinglucose infusion (followed by an outpa-
tient multidose subcutaneous insulin regimen)
improved long-term prognosis in patients with
diabetes mellitus (DM).2,3 In patients having
open heart surgery, the use of a perioperative
IIP reduced the incidence of deep sternal
wound infections in patients with diabetes to
that of the population without diabetes.4,5
Based on these and other clinical studies,
there are growing efforts worldwide to better
understand the role of hyperglycemia in the
ICU, and to improve glycemic control in criti-
cally ill patients. Unfortunately, there are many
practical barriers to achieving these goals. First,
strict BG control requires extensive nursing ef-
forts, including frequent bedside glucose mon-
itoring and the implementation of complex
IIPs. Second, the preferred method of glucose
testing is a matter of debate; the use of stan-
dard capillary glucose monitoring (finger-
sticks) may be inaccurate in hypotensive pa-
tients,6 while frequent sampling of arterial or
venous blood may lead to excess phlebotomy.
Finally, a prevalent fear of hypoglycemia
among hospital staff serves as a potent barrier
to the widespread implementation of strict IIPs.
At the present time, there are no validated
mechanisms for continuously monitoring BG
levels in hospitalized patients. As a result, there
exist no alarms that could alert the nursing staff
to impending hypo- or hyperglycemia in ICU
patients.
The Continuous Glucose Monitoring System
(CGMS, Medtronic MiniMed, Northridge, CA)
is currently approved by the U.S. Food and
Drug Administration (FDA) as a retrospec-
tive Holter-style glucose monitor.7 In clinical
practice, it is used primarily for guiding insulin
dose adjustments in outpatients with diabetes.
The CGMS uses a glucose oxidase-based plat-
inum electrode to measure an electrical current
that relates to interstitial fluid glucose concen-
tration. This current is transmitted to a pager-
sized monitor, which analyzes incoming data
every 10 s and stores a smoothed, filtered glu-
cose average every 5 min, i.e., 288 readings per
GOLDBERG ET AL.
day. CGMS readings must be calibrated at least
four times daily against standard capillary BG
measurements, which are used to retrospec-
tively calibrate the sensor readings at the time
of data analysis. This FDA-approved approach
uses modified linear regression to determine
and correct for several calibration factors. In
outpatients with diabetes, the CGMS provides
reasonable correlation with circulating BG lev-
els (see below). Implanted sensor sites gener-
ally maintain their reliability for up to 72 h.810
Despite extensive studies in outpatients with
DM, use of the CGMS has never been carefully
examined in the inpatient setting. Theoreti-
cally, disruptive clinical issues such as subcu-
taneous edema, hypotension, and the use of va-
sopressor therapy could impact upon the
accuracy and reliability of the device.
In the ICU setting, a reliable, accurate contin-
uous glucose monitor would provide a valuable
research tool for studying the impact of BG lev-
els on critically ill patients. Additionally, if the
CGMS could be further developed as a real-
time glucose monitor, it could ease nursing
workload in the ICU by greatly reducing the
number of manual BG tests required to main-
tain strict BG control. The development of an
alarm system to detect impending glycemic
excursions would also be very useful clinically.
This preliminary study was performed to in-
vestigate the efficacy and safety of the CGMS
in critically ill patients admitted to a medical
ICU (MICU).
SUBJECTS AND METHODS
Clinical setting
This study took place in the 14-bed MICU of
the Yale New Haven Hospital, a 944-bed ter-
tiary care referral center located in New Haven,
CT. Nearly 40% of the Yale MICU patients are
admitted for primary respiratory disease.
Other common diagnoses (together accounting
for another one-third of admissions) include
gastrointestinal hemorrhage, sepsis/hypoten-
sion, acute renal failure, hyperglycemic crises,
and primary cardiovascular events. In the
MICU, the patient-to-nurse ratio is either 1:1 or
2:1. The annual in-MICU mortality rate is 14%.
CGMS EXPERIENCE IN AN MICU 341

Subject recruitment tered 6090 min after sensor insertion. For all
study subjects, MICU nurses obtained a mini-
Subjects for this study were recruited from
mum of four fingersticks every 24 h; all mea-
among all patients admitted to the MICU. Pa-
sured capillary BG values were entered into the
tients expected to stay for ,24 h, e.g., admit-
CGMS system.
ted for observation following a procedure,
Persistent sensor malfunction was defined
were excluded from enrollment, as were
by the appearance of an error message on the
boarder patients, e.g., overflow postopera-
CGMS that could not be remedied within 1 h.
tive patients, not primarily cared for by the
In all cases, such malfunction occurred because
MICU physician staff. Patients were selected
of an unacceptably low sensor current, sug-
for participation based on the presence of hy-
gesting loss of proper contact between the elec-
perglycemia (BG $ 200 mg/dL), a clinical his-
trode and the interstitial fluid.
tory of DM, or the use of one or more clinical
interventions known to be associated with hy-
Data analysis
perglycemia in the MICU: corticosteroids, va-
sopressors, and/or enteral/parenteral nutri- Following data acquisition, CGMS devices
tion.11 were downloaded using the MiniMed Com-
Station. Sensor BG values were calculated us-
Data collection methods ing proprietary software (MiniMed Graphs
Version 1.7 and MiniMed Solutions Version
Our study protocol was approved by the
3.0). Concurrent glucose meter and sensor BG
Yale University School of Medicine Human In-
values were compared by calculating both the
vestigation Committee. Informed consent was
mean difference (MD 5 meter BG 2 sensor
obtained from all study subjects and/or their
BG) and mean absolute difference (MAD 5
families. All data were collected prospectively,
uMDu) between BG values, expressed both in
from the active hospital chart and MICU nurs-
mg/dL and as a percentage. A Pearson corre-
ing records. Admission clinical variables in-
lation coefficient (r) was calculated for the en-
cluded age, sex, race, height, weight, history of
tire set of metersensor (M-S) pairs. Finally,
DM, and an Acute Physiology And Chronic
paired M-S values were plotted on a Clarke Er-
Health Evaluation II (APACHE II) score.12
ror Grid,13 which has been accepted by the FDA
APACHE II provides a validated severity of ill-
as a valid supplement to the review of contin-
ness score based on 12 readily accessible clini-
uous glucose measurements. 14 Except where
cal and laboratory parameters (heart rate,
noted, all clinical data are expressed as mean 6
hematocrit, etc.). Reasons for ICU admission
standard deviation (SD), or as a percentage.
and the use of relevant clinical interventions
(nutrition therapy, corticosteroids, vasopres-
sors) were also collected. All study subjects
were followed until discharge from the MICU. RESULTS
Capillary (fingerstick) BG levels were mea-
sured using a standard hospital glucose meter Twenty-four patients and/or their families
(Surestep Flexx, Lifescan, Johnson & Johnson, were approached for study enrollment. Two of
Milpitas, CA). All glucose meters were cali- the 24 declined study participation. Baseline
brated on a daily basis. CGMS sensors were in- characteristics and relevant clinical interven-
serted using the Sen-Serter device (Medtronic tions for the 22 enrolled study subjects are
MiniMed). Sensors were placed in the lateral shown in Table 1. Thirteen (59%) patients had
abdomen, or in the upper, outer portion of the known DM; five of these 13 were insulin-de-
buttocks. Transparent tape was then used to se- pendent. The median BG level upon ICU ad-
cure the sensors to the skin. MICU nurses mission was 175 mg/dL (interquartile range,
and/or one of the investigators inspected the 134279 mg/dL).
site at least twice daily for signs of local irrita- In total, 41 CGMS sensors were worn (one to
tion, infection, or bleeding. To allow time for five sensors per patient), yielding 546 paired
calibration, the first recorded BG value was en- M-S values for analysis. These M-S BG pairs are
342 GOLDBERG ET AL.

TABLE 1. BASELINE CHARACTERISTICS AND of absolute BG levels on measured M-S differ-

RELEVANT CLINICAL INTERVENTIONS ences.
EMPLOYED FOR THE 22 STUDY SUBJECTS
Notably, our study did not test the accuracy
IIP patients of the CGMS during hypoglycemic conditions.
Characteristic (n 5 22) Of the 546 M-S pairs obtained, only four had
Age (years) 61 6 15
BG levels ,60 mg/dL using either method of
Male (%) 50% BG measurement. In three of these four in-
Race (%) stances, the CGMS value was lower than the
Caucasian 55% corresponding meter BG level. In one case, a
African American 36%
Hispanic 5% meter BG level of 40 mg/dL was assigned a
Other 5% value of 89 mg/dL by the CGMS.
Body mass index (kg/m2) 28.8 6 11.0 In this study, the presence of DM did not sig-
APACHE II score 20.0 6 6.80
Primary diagnosis (%) nificantly affect the accuracy of the CGMS. The
Respiratory failure 50% MD% for M-S pairs from patients with diabetes
Hyperglycemic crisis 18% was 0.5 6 19.1% (vs. 0.8 6 15.2% for patients
Sepsis/hypotension 14%
Primary cardiovascular events 9%
without diabetes, P 5 0.86), while the MAD%
Other 9% was 13.6 6 13.5% (vs. 11.8 6 9.5%, P 5 0.07).
Clinical interventions (%) The use of vasopressors also did not affect sen-
Corticosteroid therapy 59% sor accuracy. The MD% for M-S pairs from pa-
Vasopressor therapy 32%
Enteral nutrition 68% tients on vasopressor therapy was 0.0 6 16.9%
Parenteral nutrition 9% (vs. 1.1 6 17.9% for patients not on vasopres-
sors, P 5 0.47), while the MAD% was 11.8 6
All data are expressed as mean 6 SD, or as a percent-
age. 12.0% (vs. 13.5 6 11.7%, P 5 0.10). Because we
enrolled only 22 subjects, this study lacked the
power to assess the impact of individual clini-
cal diagnoses on CGMS performance.
plotted in Figure 1. For all 546 pairs, the MD Results of the Clarke Error Grid analysis are
was 3.3 6 26.7 mg/dL (0.6 6 17.4%), and the shown in Figure 3. Of the 546 M-S pairs, 78.4%
MAD was 19.7 6 18.3 mg/dL (12.8 6 11.9%). fell within Zone A, indicating agreement
The overall r was 0.88. Overall, 63.4% of M-S within 20% (or agreement on hypoglycemia),
pairs fell within 20 mg/dL of one another, 20.3% fell within Zone B, indicating a M-S dif-
while 87.8% fell within 40 mg/dL. Aside from ference of more than 20% that would not likely
cases of persistent sensor malfunction prior to result in treatment error, and only 1.3% fell
72 h (seven of the 41 sensors), only 16 of 562 within Zones CE, indicating that these pairs
(2.9%) meter values were not assigned a paired would likely have lead to a treatment error. In
sensor reading. In all of these cases, a tem- summary, then, 98.7% of the 546 M-S pairs
porarily low sensor current was responsible for were judged as clinically acceptable using
the lack of data. Clarke Error Grid criteria.
To determine the effects of absolute BG lev- There were no serious adverse events re-
els on M-S differences, we separately analyzed ported during use of the CGMS. In two patients
the M-S pairs by absolute BG value. For this (both receiving anticoagulant), mild bleeding
analysis, shown in Table 2, mean M-S BG occurred during sensor insertion. In one sub-
levels were employed in order to minimize ject, the bleeding stopped quickly with gentle
measurement bias. As expected, MAD rose pressure. In the second, an alternate sensor site
slowly as absolute BG levels increased, from was needed to achieve timely hemostasis. In-
15.4 mg/dL when BG was ,100 mg/dL to 28.7 terestingly, these two events did not appear to
when BG was $ 250 mg/dL. However, MAD% affect subsequent CGMS measurements. Other
decreased with higher absolute BG levels, from than the single sensor insertion described
20.0% when BG was ,100 mg/dL to 9.8% when above, no CGMS sensors had to be removed
BG was $ 250 mg/dL. A BlandAltman plot,15 because of local pain, irritation, or bleeding,
shown in Figure 2, visually displays the impact and there were no infectious complications.
CGMS EXPERIENCE IN AN MICU 343

FIG. 1. Scatter plot of the 546 M-S BG pairs. BG values from the CCGMS are shown on the y-axis. BG values from
the referent glucose meter are shown on the x-axis.

Seven of the 41 sensors (17%) exhibited per-
sistent sensor malfunction prior to 72 h. In
one subject (a thin 83-year-old man with dia-
betes admitted for an exacerbation of chronic
obstructive pulmonary disease, who re-
mained normotensive throughout his ex-
tended ICU stay), three of three implanted
sensors failed prior to 72 h; as a result, this
one subject accounted for nearly one-half of
our studys persistent sensor malfunctions.
Following detailed review of this subjects
CGMS data, it was clear that his sensor cur-
rents were continuously low during all three
of his sensor placements.

TABLE 2. PERFORMANCE OF THE CGMS BY ABSOLUTE BG LEVEL

Mean BG (mg/dL) MD MAD

Mean M-S BG value
(mg/dL) Number of M-S pairs Meter Sensor mg/dL % mg/dL %

,100 43 82.3 82.4 20.1 23.4 15.4 20.0

100149 219 127.5 122.7 24.8 22.1 17.3 13.5
150199 185 173.3 171.4 1.9 20.2 19.7 11.3
200249 66 221.8 216.1 5.7 1.4 25.7 11.4
$250 33 295.6 294.8 0.8 20.9 28.7 9.8

To minimize measurement bias, mean M-S BG values (shown in the leftmost column) were used to categorize the
BG data. Note that while MADs (in mg/dL) climbed with rising BG levels, MADs by percent became smaller at higher
BG concentrations.
344 GOLDBERG ET AL.

FIG. 2. BlandAltman plot15 for the 546 M-S pairs. Mean M-S BG levels are shown on the x-axis. The y-axis displays
M-S BG differences.

DISCUSSION AND CONCLUSIONS betes. 810 In an early company-sponsored

This preliminary study was performed to in-
vestigate the accuracy of the CGMS in critically
ill patients admitted to an MICU. The accuracy
and reliability of the CGMS have been previ-
ously demonstrated in outpatients with dia-
study,8 62 subjects with diabetes utilized the
CGMS for up to 3 weeks, yielding over 300,000
individual sensor readings and more than 9,000
paired M-S values. Daily correlation coeffi-
cients showed decent trend agreement, with a
median daily r 5 0.92. Overall, sensor readings

FIG. 3. Scatter plot of the 562 M-S BG pairs using Clarke Error Grid analysis (see text for explanation). BG values
from the CGMS are shown on the y-axis, while BG values from the referent glucose meter are shown on the x-axis.
Percentages of M-S pairs falling within each zone of the Clarke Error Grid are shown. Using this method, 98.7% of
M-S pairs were clinically acceptable, falling within Zone A or B.
CGMS EXPERIENCE IN AN MICU
were 5.4 6 44.2 mg/dL below standard meter
readings, with a MD% of 20.3% 6 32.4%. A
follow-up study of 135 patients yielded similar
results, with r 5 0.91, MAD% 5 18.0 6 19.8%,
and a Clarke Error Grid analysis showing
96.2% of M-S pairs within clinically acceptable
zones A and B.9 A larger postmarketing study
in 238 patients reported a similar r (0.91) and a
MAD% of 12.6%; in this study, age, race, and
the type/duration of DM were found to have
no impact upon the accuracy of the device.10
Importantly, CGMS measurements appear to
be reliable under conditions of hypo- and hy-
perglycemia, 16,17 and seem unaffected by am-
bient insulin concentrations. 18 The accuracy of
the CGMS may improve slightly when calibra-
tions are performed by an experienced nurse.19
To date, however, this technology has not been
validated in the inpatient setting.
In our study, the accuracy of the CGMS in
the MICU compared favorably with its accu-
racy in outpatients with DM. In 22 MICU pa-
tients, we found an MD of 3.3 6 26.7 mg/dL
(0.6 6 17.4%), a MAD of 19.7 6 18.3 mg/dL
(12.8 6 11.9%), and an overall r of 0.88. As ex-
pected, MAD rose slightly during hyper-
glycemia. However, MAD% fell as BG levels
climbed. In other words, when analyzed by per-
cent difference, the accuracy of the CGMS im-
proved at higher BG levels. Finally, using
Clarke Error Grid analysis, 98.7% of the CGMS
readings in our study were deemed clinically
acceptable; this compares favorably with the
96.2% rate published in outpatient studies.9
Though our study subject numbers were too
small for detailed statistical analyses, poten-
tially disruptive clinical issues such as edema,
hypotension, and vasopressor therapy (all of
which were present in some of our patients)
did not appear to affect the accuracy of the
CGMS. We hypothesize that both the sedentary
nature and the supervised environment of
MICU patients contributed significantly to the
devices accuracy in this setting.
The results of this preliminary study should
be interpreted with the following important
caveats:
1. Accuracy of the CGMS. Though several pub-
lished studies have demonstrated clinical
benefit employing the CGMS in outpatients
345
with diabetes,2022 it should be noted that le-
gitimate concerns have been raised about
the reproducibility of BG measurements ob-
tained from CGMS monitoring.23 In addi-
tion, the accuracy of the CGMS is unavoid-
ably affected by the accuracy of the referent
glucose measurement. In hypotensive pa-
tients, for example, capillary (fingerstick)
BG determinations may underestimate cir-
culating BG levels.6 Despite this finding, fin-
gersticks remain the standard method for
BG measurement in many ICUs. During our
study, we elected to use standard capillary
readings as the BG referent, primarily to best
allow comparisons with previously pub-
lished outpatient studies. However, future
studies comparing CGMS readings to arter-
ial (or venous) blood will be required before
use of the CGMS can be validated in the ICU
setting.
2. Hypoglycemia detection. Though prior studies
suggest unaltered CGMS accuracy during
hypoglycemia,15,16 small BG differences can
have significant clinical relevance when the
levels fall into the hypoglycemic range. This
point has important implications for the po-
tential clinical use of the CGMS as a real-
time glucose monitor. In our study, limited
subject numbers prohibited a meaningful
exploration of the CGMS accuracy during
hypoglycemia. A larger, prospective study
performed in patients on intensive IIPs will
be required to validate use of the CGMS un-
der hypoglycemic conditions.
3. Retrospective data. An important limitation of
the current CGMS is its function as a ret-
rospective glucose monitor. At the present
time, data from the monitor can be accessed
only after the completion of a 72-h data col-
lection period. Before the CGMS can be re-
lied upon clinically, extensive inpatient
studies using real-time CGMS technology
will be required. In addition, for optimal
clinical application, the creation and valida-
tion of an ICU-specific error grid (designed
to test how the CGMS guides an IV insulin
infusion) will likely be required.
In conclusion, this preliminary study sug-
gests that the CGMS is promising for potential
use in ICU patients. The device appears to pro-
346
vide similar accuracy in critically ill MICU pa-
tients as compared with outpatients with DM.
We believe that the CGMS could serve as a use-
ful research tool for investigating the role of hy-
perglycemia (and strict glycemic control) in crit-
ically ill patients. If further developed as a
real-time glucose sensor, CGMS technology
may also prove clinically useful in the ICU, by
decreasing nursing workload and/or by pro-
viding alarm signals for impending glycemic ex-
cursions. Such technology would address the
most important barriers to the successful imple-
mentation of intensive insulin infusions, and
would likely facilitate the use of such protocols
in critically ill patients. In turn, improved glyce-
mic control in the ICU could lead to reduced
morbidity and mortality in critically ill patients.
ACKNOWLEDGMENTS
CGMS monitors, sensors, software, and an-
cillary equipment for this study were provided
by Medtronic MiniMed Corp., Northridge, CA.
Additional unrestricted study funds were pro-
vided by Eli Lilly and Co., Indianapolis, IN. Dr.
Goldberg was supported by Juvenile Diabetes
Research Foundation fellowship grant JDRF-3-
2003-95. Study support was also derived from
Yales General Clincial Research Center, Na-
tional Institutes of Health center grant MO1
RR-00125.
REFERENCES
1. Van den Berghe G, Wouters P, Weekers F, Verwaest
C, Bruyninckx F, Schetz M, Vlasselaers D, Ferdinande
P, Lauwers P, Bouillon R: Intensive insulin therapy in
critically ill patients. N Engl J Med 2001;345:1359
1367.
2. Malmberg K, Ryden L, Efendic S, Herlitz J, Nicol P,
Waldenstrom A, Wedel H, Welin L, DIGAMI Study
Group: Randomized trial of insulin-glucose infusion
followed by subcutaneous insulin treatment in dia-
betic patients with acute myocardial infarction
(DIGAMI study): effects on mortality at 1 year. J Am
Coll Cardiol 1995;26:5765.
3. Malmberg K, DIGAMI Study Group: Prospective ran-
domised study of intensive insulin treatment on long-
term survival after acute myocardial infarction in pa-
tients with diabetes mellitus. Br Med J 1997;314:1512
1515.
4. Zerr KJ, Furnary AP, Grunkemeier GL, Bookin S, Kan-
here V, Starr A: Glucose control lowers the risk of
GOLDBERG ET AL.
wound infection in diabetics after open heart opera-
tions. Ann Thor Surg 1997;63:356361.
5. Furnary AP, Zerr KJ, Grunkemeier GL, Starr A: Con-
tinuous intravenous insulin infusion reduces the in-
cidence of deep sternal wound infection in diabetic
patients after cardiac surgical procedures. Ann Thor
Surg 1999;67:352362.
6. Atkin SH, Dasmahapatra A, Jaker M, Chorost MI,
Reddy S: Fingerstick glucose determination in shock.
Ann Intern Med 1991;114:10201024.
7. Mastrototaro JJ: The MiniMed continuous glucose
monitoring system. Diabetes Technol Ther 2000;2
(Suppl 1):S13S18.
8. Gross TM, Mastrototaro JJ: Efficacy and reliability of
the continuous glucose monitoring system. Diabetes
Technol Ther 2000;2(Suppl 1):S19S26.
9. Gross TM, Bode BW, Einhorn D, Kayne DM, Reed JH,
White NH, Mastrototaro JJ: Performance evaluation
of the MiniMed coninuous glucose monitoring sys-
tem during patient home use. Diabetes Technol Ther
2000;2:4956.
10. Gross TM, Ter Veer A: Continuous glucose monitor-
ing in previously unstudied population subgroups.
Diabetes Technol Ther 2000;2(Suppl 1):S27S34.
11. Inzucchi S, Goldberg P, Dziura J, Siegel M, Lee M,
Halickman J, Sherwin R: Risk factors for poor glyce-
mic control in a medical intensive care unit (MICU)
[abstract]. Diabetes 2003;52(Suppl 1):A96.
12. Knaus WA, Draper EA, Wagner DP, Zimmerman JE:
APACHE II: a severity of disease classification sys-
tem. Crit Care Med 1985;13:818829.
13. Clarke WL, Cox D, Gonder-Frederick LA, Carter W,
Pils SL: Evaluating clinical efficacy of systems for self-
monitoring of blood glucose. Diabetes Care 1987;10:
622628.
14. Food and Drug Administration: Review Criteria As-
sessment of Portable Glucose Monitoring In Vitro Di-
agnostic Devices Using Glucose Oxidase, Dehydro-
genase, or Hexokinase Methodology. Draft Guidance
Document. Washington, DC: Food and Drug Admin-
istration, 1997.
15. Bland JM, Altman DG: Statistical methods for assess-
ing agreement between two methods of clinical mea-
surement. Lancet 1986;i:307310.
16. Steil GM: Accurate determination of plasma glucose
during hyper- and hypoglycemia with a subcuta-
neous glucose sensor [abstract]. Diabetes 2000;49
(Suppl 1):510.
17. Cheyne EH, Cavan DA, Kerr D: Performance of a con-
tinuous glucose monitoring system during controlled
hypoglycemia in healthy volunteers. Diabetes Tech-
nol Ther 2002;4:607613.
18. Rebrin K, Steil GM, Van Antwerp WP, Mastrototaro
JJ: Subcutaneous glucose predicts plasma glucose in-
dependent of insulin: implications for continuous
monitoring. Am J Physiol 1999;277:E561E571.
19. Djakoure-Platanoff C, Radermercker R, Reach G,
Slama G, Selam JL: Accuracy of the continuous glu-
cose monitoring system in inpatient and outpatient
conditions. Diabetes Metab 2003;29:159162.
CGMS EXPERIENCE IN AN MICU 347

20. Kaufman F, Gibson L, Halvorson M, Carpenter S, 23. Metzger M, Leibowitz G, Wainstein J, Glaser B, Raz
Fisher LK, Pitukcheewanont P: A pilot study of the I: Reproducibility of glucose measurements using the
Continuous Glucose Monitoring System: clinical de- glucose sensor. Diabetes Care 2002;25:11851191.
cisions and glycemic control after its use in pediatric
type 1 diabetic subjects. Diabetes Care 2001;24:
20302034. Address reprint requests to:
21. Ludvigsson J, Hanas R: Continuous subcutaneous
glucose monitoring improved metabolic control in pe-
Silvio E. Inzucchi, M.D.
diatric patients with type 1 diabetes: a controlled Section of Endocrinology
crossover study. Pediatrics 2003;111:933938. Yale University School of Medicine
22. Bode B, Lane W, Levetan C, Mestman J, Peters- 333 Cedar Street, LLCI 101
Harmel A, Tanenberg R, Tobian J, Gross T, Mastro- P.O. Box 208020
totaro J: Therapy adjustments based on CGMS data
New Haven, CT 06520-8020
lower HbA1c with less hypoglycemia than blood glu-
cose meter data alone [abstract]. Diabetes 2003;52
(Suppl 1):A90. E-mail: silvio.inzucchi@yale.edu
This article has been cited by:

1. Masami Hoshino, Yoshikura Haraguchi, Iwanori Mizushima, Motohiro Sakai. 2009. Recent progress in
mechanical artificial pancreas. Journal of Artificial Organs 12:3, 141-149. [CrossRef]
2. Ulrike Holzinger, Joanna Warszawska, Reinhard Kitzberger, Harald Herkner, Philipp G. H. Metnitz, Christian
Madl. 2009. Impact of shock requiring norepinephrine on the accuracy and reliability of subcutaneous continuous
glucose monitoring. Intensive Care Medicine . [CrossRef]
3. R. P. Radermecker, A. Sultan, C. Piot, A. S. Remy, A. Avignon, E. Renard. 2009. Continuous glucose monitoring
as a tool to identify hyperglycaemia in non-diabetic patients with acute coronary syndromes. Diabetic Medicine
26:2, 167-170. [CrossRef]
4. Susan J. Logtenberg , Nanne Kleefstra , Ferdinand T. Snellen , Klaas H. Groenier , Robbert J. Slingerland , Arno
P. Nierich , Henk J. Bilo . 2009. Pre- and Postoperative Accuracy and Safety of a Real-Time Continuous Glucose
Monitoring System in Cardiac Surgical Patients: A Randomized Pilot StudyPre- and Postoperative Accuracy
and Safety of a Real-Time Continuous Glucose Monitoring System in Cardiac Surgical Patients: A Randomized
Pilot Study. Diabetes Technology Therapeutics 11:1, 31-37. [Abstract] [PDF] [PDF Plus]
5. Holley F. Allen, Alyssa Rake, Marybeth Roy, Dennis Brenner, Christine A. McKiernan. 2008. Prospective
detection of hyperglycemia in critically ill children using continuous glucose monitoring*. Pediatric Critical Care
Medicine 9:2, 153-158. [CrossRef]
6. Venkata Radhakrishna Kondepati, H. Michael Heise. 2007. Recent progress in analytical instrumentation for
glycemic control in diabetic and critically ill patients. Analytical and Bioanalytical Chemistry 388:3, 545-563.
[CrossRef]
7. Kupper A. Wintergerst , Dorothee Deiss , Bruce Buckingham , Martin Cantwell , Saraswati Kache , Swati
Agarwal , Darrell M. Wilson , Garry Steil . 2007. Glucose Control in Pediatric Intensive Care Unit Patients
Using an InsulinGlucose AlgorithmGlucose Control in Pediatric Intensive Care Unit Patients Using an
InsulinGlucose Algorithm. Diabetes Technology Therapeutics 9:3, 211-222. [Abstract] [PDF] [PDF Plus]
8. Michael Balkin, Charles Mascioli, Virginia Smith, Hicham Alnachawati, Sandeep Mehrishi, Ghulam Saydain,
Helen Slone, Joseph Alessandrini, Lawrence Brown. 2007. Achieving durable glucose control in the intensive
care unit without hypoglycaemia : a new practical IV insulin protocol. Diabetes/Metabolism Research and Reviews
23:1, 49-55. [CrossRef]
9. Kathie L. Hermayer. 2006. Strategies for Controlling Glucose in the Intensive Care Unit. Clinical Pulmonary
Medicine 13:6, 332-347. [CrossRef]
10. J. Pftzner, T. Forst, R. Butzer, S. Forst, M. M. Weber, A. H. Pftzner, A. Pftzner. 2006. Performance of
the continuous glucose monitoring system (CGMS) during development of ketosis in patients on insulin pump
therapy. Diabetic Medicine 23:10, 1124-1129. [CrossRef]
11. David C. Klonoff. 2006. Continuous Glucose Monitoring Technology Delivers Detailed Diabetes Data. Point of
Care: The Journal of Near-Patient Testing & Technology 5:3, 105-111. [CrossRef]
12. R. Hovorka. 2006. Continuous glucose monitoring and closed-loop systems. Diabetic Medicine 23:1, 1-12.
[CrossRef]
13. Associate Professor J. Geoffrey Chase , Geoffrey M. Shaw , Jessica Lin , Carmen V. Doran , Chris Hann
, Thomas Lotz , Graeme C. Wake , Bob Broughton . 2005. Targeted Glycemic Reduction in Critical Care
Using Closed-Loop ControlTargeted Glycemic Reduction in Critical Care Using Closed-Loop Control. Diabetes
Technology Therapeutics 7:2, 274-282. [Abstract] [PDF] [PDF Plus]