Continuous Glucose Monitoring—A Study of the Enlite
Sensor During Hypo- and Hyperbaric Conditions
Peter Adolfsson, M.D.,1Hans O¨rnhagen, M.D., Ph.D.,2Bengt M. Eriksson, M.D.,3
Ken Cooper, M.E.,4and Johan Jendle, M.D., Ph.D.5,6
Background: The performance and accuracy of the Enlite?(Medtronic, Inc., Northridge, CA) sensor may be affected by
microbubble formation at the electrode surface during hypo- and hyperbaric conditions. The effects of acute pressure changes
and of prewetting of sensors were investigated.
Materials and Methods: On Day 1, 24 sensors were inserted on the right side of the abdomen and back in one healthy
individual; 12 were prewetted with saline solution, and 12 were inserted dry. On Day 2, this procedure was repeated on
the left side. All sensors were attached to an iPro continuous glucose monitoring (CGM) recorder. Hypobaric and hy-
perbaric tests were conducted in a pressure chamber, with each test lasting 105min. Plasma glucose values were obtained
at 5-min intervals with a HemoCue?(A¨ngelholm, Sweden) model 201 glucose analyzer for comparison with sensor
Results: Ninety percent of the CGM systems operated during the tests. The mean absolute relative difference was lower
during hyperbaric than hypobaric conditions (6.7% vs. 14.9%, P<0.001). Sensor sensitivity was slightly decreased (P<0.05)
during hypobaric but not during hyperbaric conditions. Clarke Error Grid Analysis showed that 100% of the values were
found in the A+B region. No differences were found between prewetted and dry sensors.
Conclusions: The Enlite sensor performed adequately during acute pressure changes and was more accurate during hy-
perbaric than hypobaric conditions. Prewetting the sensors did not improve accuracy. Further studies on type 1 diabetes
subjects are needed under various pressure conditions.
more important during specific situations such as flying and
diving. During these conditions, a major hypoglycemic event
could be catastrophic.
For those who have diabetes, there are restrictions re-
garding aviation as well as scuba diving. In most countries,
those with type 1 diabetes are prohibited from working as a
pilot, but in the United States an individual with the disease
may be considered for a third-class airman medical certificate
by demonstrating good overall diabetes control. In some
countries, recreational diving is prohibited for those with
either type 1 or insulin-treated type 2 diabetes.
requent systematic measurement of glucose is im-
portant for persons with type 1 diabetes and becomes
Self-monitored blood glucose (SMBG) is frequently used as
the most common method of obtaining glucose measure-
ments, but SMBG only provides a value when sampling is
carried out. The accuracy of SMBG readings is important
when used to evaluate the insulin dose or to initiate treatment
of hypoglycemic episodes. Glucose meters with inaccuracies
(measured as the mean absolute relative difference [MARD]
Atmospheric pressure and the partial pressure of oxygen
canaffect theaccuracyofaglucosemeter.Test stripsusingthe
glucose dehydrogenase method are affected less than test
strips using glucose oxygenase, and differences are seen
among meters from different manufacturers.2Altitude also
1Go ¨teborg Pediatric Growth Research Center, Department of Pediatrics, Institute of Clinical Sciences, Sahlgrenska Academy at University
of Gothenburg, Go ¨teborg, Sweden.
2Swedish Sports Diving Federation, Farsta, Sweden.
3Hyperbaric Medicine, Department of Anesthesiology, Karolinska University Hospital, Solna, Sweden.
4Medtronic Diabetes (Sensor R&D), Northridge, California.
5Endocrine and Diabetes Center, Karlstad Hospital, Karlstad, Sweden.
6Faculty of Medicine and Health, O¨rebro University Hospital, O¨rebro, Sweden.
DIABETES TECHNOLOGY & THERAPEUTICS
Volume 14, Number 6, 2012
ª Mary Ann Liebert, Inc.
has an impact on the accuracy of glucose readings,3as do
temperature and relative humidity.4The glucose meters un-
derestimated glucose levels by approximately 1–2% for each
1,000 feet of increased elevation after correction for changes in
temperature and humidity.
New pharmacological agents for the treatment of diabetes
have been developed with the aim of reducing the risk of
hypoglycemia. In parallel to this, continuous glucose moni-
toring (CGM) is also possible. Today, CGM can provide data
either retrospectively or in real time. This technique enables
detection as well as reduction of hypoglycemia.5–7An indi-
vidual assessment of diabetes subjects’ risk of hypoglycemia
is of importance during activities such as flying and recrea-
tional scuba diving.8,9
The number of hypoglycemic events may be decreased
when monitors showing glycemic trends or equipped with
predictive alarms prompt appropriate intervention.10How-
ever, glycemic excursions during hypobaric (flying) or during
hyperbaric (diving) conditions have not been carefully eval-
uated, nor have CGM techniques been assessed for accuracy
during changes in atmospheric pressure. The aim of the
present study was to evaluate the accuracy and reliability of
the Enlite?(Medtronic, Inc., Northridge, CA) subcutaneous
glucose sensor in hypobaric and hyperbaric conditions and to
determine whether prewetted sensors improve the accuracy
of sensor readings.
Research Design and Methods
One healthy individual was included. Study eligibility re-
quired scuba diving certification and a past experience of
pressure chambers. Prior to the test, a health declaration and
laboratory test certified that the individual was in good
health. One healthcare professional accompanied the partici-
pant in the pressure chamber for adverse event monitor-
ing and frequent blood sampling. The ethics committee at
Uppsala University, Uppsala, Sweden, approved the study
protocol (Dnr. 2010/246), and signed informed consent was
obtained prior to the study procedures.
Capillary blood glucose (P-glucose) values were measured
by a healthcare professional using a HemoCue (A¨ngelholm,
Sweden) model 201 glucose analyzer at 5-min intervals.
Forty-eight Enlite glucose sensors were used together with
iPro?(Medtronic) recording devices. A prototype firmware
was used to collect data in the iPro recorder at a faster sam-
pling rate, 1-min intervals, compared with the commercial
version, 5-min intervals. On Day 1, 24 sensors were inserted
on the right side of the body, on the abdomen and back.
Twelve of the sensors were prewetted with a sterile saline
solution prior to insertion, whereas the other 12 sensors were
not. Sensors were calibrated to two P-glucose values prior to
pressure changes and to two P-glucose values after the pres-
sure changes. After the 2-h sensor initialization period and a
stabilization period at 1.0 atm outside the pressure chamber,
the hypobaric test was implemented. On Day 2, the insertion
procedure was repeated with 24 new sensors being inserted
on the left side of the body. After the initialization and sta-
bilization period the hyperbaric test was then conducted.
pressure chamber at the Hyperbaric Medicine Unit at Kar-
olinska University Hospital, Stockholm, Sweden. The ambi-
ent pressure during the experiment was varied according to
On Day 1, after the stabilization period of 30min at
101kPa/758mm Hg (0m, 1 atm, 21kPa of O2), the chamber
10kPa of O2) for a period of 20min and then increased to
75kPa/563mm Hg (2,500m, 0.75 atm, 16kPa of O2) for
10min before return to 101kPa/758mm Hg (0m, 1.0 atm,
at 101kPa/758mm Hg (0m, 1.0 atm, 21 of kPa O2), the
chamber pressure was increased to 404kPa/3,030mm Hg
(30m, 4.0 atm, 84kPa of O2) for a period of 20min, followed
byadecompression stop at130kPa/975mmHg(3m,1.3atm,
27kPa of O2) and return to 101kPa/758mm Hg (0m, 1.0 atm,
21kPa of O2).
The specific pressure levels were chosen tomimic extensive
height (5,500m) and air cabin pressure (2,500m) and the
specific depth to mimic extensive depth (30m) and decom-
pression stop at 3m according to the diving table.11
In order to evaluate the possible effect of microbubbles,
formed at insertion at the sensor tip, half of the sensors were
prewetted with a sterile saline solution prior to insertion.
The Statistical Package for the Social Sciences version 19.0
(SPSS, Chicago, IL) was used for statistical analysis. Differ-
ences in sensor function during the hypobaric and hyperbaric
tests were evaluated using the Wilcoxon signed rank test.
Sensor accuracy was evaluated by comparison with contem-
poraneous blood glucose values as the MARD, defined
as 100·the absolute value of ([sensor glucose – P-glucose]/
P-glucose). The Mann–Whitney test was used for comparison
of dry and prewetted sensor accuracy under hypobaric and
hyperbaric conditions. Clarke Error Grid Analysis plots were
generated to evaluate the distribution of sensor glucose val-
ues. Sensitivity variation is defined as the state of the sensor:
Sensitivity variation=1/(meter P-glucose/paired ISIG). Meter
P-glucose is defined as that measured with the HemoCue
device, and ISIG is the electrical output of the glucose sensor.
Sensitivity variation is a method to determine the stability of
the sensor over time and was evaluated during the different
pressures. The Kruskal–Wallis test was used for compari-
sons of sensitivity versus pressure. A P value of<0.05 was
regarded as significant.
All of the sensors (24 of 24 [100%]) worked during the
hyperbaric test, compared with 19 of 24 (79%) during the
hypobaric test (P=0.025). All cases of failure during the hy-
pobaric test, the first test in thestudy, were determined to be a
result of the prototype firmware used to collect data in the
CGMS recorder at a faster sampling rate than the commercial
version. Once the failure was detected during the hypobaric
528ADOLFSSON ET AL.
test, the units were reprogrammed to correct the firmware
failure and subsequently used during the second test, which
was under hyperbaric conditions.
The mean ISIG values (in nA) from all sensors and the
timing of the calibration during the hypobaric and hyperbaric
test are shown in Figure 2.
During the hypobaric test, sensor accuracy (MARD) was
14.9–9.1% (95% confidence interval, 13.1–16.7) and was sig-
nificantly lower during the hyperbaric test (6.7–7.9% [95%
confidence interval, 5.8–7.5]) (P<0.001). The variation of
sensor sensitivity at different pressure conditions is demon-
strated in Table 1. Figure 3 shows a separate analysis of all
sensors and the MARD of each sensor during both tests,
illustrating that MARD was consistently lower during the
hyperbaric test. When all sensor values were analyzed with
the Clarke Error Grid Analysis, 100% of the sensor values
were found to be in the A+B region.
The sensitivity variation was assessed during the altered
pressures. When a glucose sensor is calibrated, its electrical
output (ISIG) is related to the blood glucose value at that
specific time. A stable relationship between P-glucose and
ISIG signal would verify that altered pressure does not affect
the ISIG signal. During hyperbaric conditions no significant
differences were seen. However, when the sensitivity was
evaluated during hypobaric conditions, a higher value of 1/
sensitivity (meter P-glucose/ISIG) was seen at 51kPa/383mm
Hg (5,500m height) and at 75kPa/563mm Hg (2,500m
height) compared with the corresponding value at 101kPa/
758mm Hg (sea level=0m) (P<0.001). This means that the
ISIG signal was increased in relation to P-glucose during
hypobaric conditions. Table 1 and Figure 4 show sensitivity
variations during both tests.
No significant difference was found between MARD for
the prewetted sensors (11.3–9.5%; 95% confidence interval,
6.4–15.2) and for dry sensors at insertion (9.4–6.8%; 95%
confidence interval, 5.6–11.8).
In this study we demonstrate that use of the Medtronic
Enlite sensor in hypo- and hyperbaric conditions resulted in
satisfying sensor and system performance, indicating that use
of CGM in people with diabetes, in these specific conditions,
may also be justified.
CGM has previously been evaluated in hyperbaric condi-
tions in connection to recreational scuba diving8,9,12but not in
hypobaric conditions. The CGMS?(Medtronic) that was used
in these studies on divers with type 1 diabetes and healthy
controls showed promising results as it was possible to eval-
uate the glucose fluctuation during dive8,12and to detect hy-
poglycemia before, during, or after a dive.8Furthermore, it
was possible to identify those individuals suitable for diving.8
The CGMS was used with good accuracy, having a mean
MARD of 14.4–6%, during a 3-day period and five dives,9
which also is confirmed by an outpatient ambulatory study
MARD values during both hypobaric (14.9%) and hyperbaric
(6.7%) tests, which are below the 15% level suggested as the
Study design of hypo- and hyperbaric tests. Pressure illustrated corresponding to height or depth in meters.
ACCURACY OF CGM DURING PRESSURE CHANGES 529
limit for proper adjustment of insulin dosing.1In the Clarke
Error Grid Analysis, we also found all values in the A+B
region of clinically accurate results. One possible contributing
factor to the relatively low MARD seen in our study could be
that two blood glucose values were used to calibrate the
sensors prior to the pressure changes and another two values
after the tests. Another contributing factor is that the partici-
pant was a healthy individual, with a stable glycemic control.
not experience any problem about the time lag between blood
and interstitial glucose. The described time lag between in-
to increased MARD. Although we show that sensors could be
used during both conditions, significantly more systems
failed during the hypobaric test; we attribute this to a failure
of the prototype firmware used to collect data in the
CGMS recorder. In the hypobaric conditions, MARD was
significantly increased compared with MARD in hyperbaric
conditions. A lower oxygen partial pressure in hypobaric
conditions could be one explanation of the increased MARD
as this has been shown to affect home glucose meters.2This
phenomenon merits further investigation.
The difference recorded can also be attributed to the in-
creased variation in sensor sensitivity, also seen during the
hypobaric test. Sensitivity variation is a method to calculate
conditions of changes in environment (ie oxygen partial
pressure, pressure conditions). The clinical significance of the
decreased sensitivity under hypobaric conditions might be of
minor importance taking into account the values of sensor
sensitivity during our tests. A test to determine this variation
in sensor sensitivity shows that the variation was well within
the sensitivity threshold of 1.5–15 as defined by the real-time
In this study it was hypothesized that microscopic gas
bubbles formed at insertion of the sensor might impair sensor
Table 1. Variation of Sensor Sensitivity at Different Pressure Conditions
Pressure 1/Sensitivity (meter P-glucose/ISIG)
TestPressure (kPa)mm Hgatm Height/depth (m)Paired values (n) MeanMedian MinimumMaximumSD
Data are presented for hypobaric and hyperbaric tests. Sensitivity was examined as 1/(meter P-glucose/paired ISIG), where P-glucose is
capillary blood glucose measured with the HemoCue device and ISIG is the electrical output of the glucose sensor.
hyperbaric tests. Calibration of the sensors was conducted before and after pressure changes. Calibrations and pressure
changes are illustrated to enhance interpretation of the figure. Color graphics available online at www.liebertonline.com/dia
Mean electrical output of the glucose sensor (ISIG) values (in nA) from all sensors during the hypobaric and
530ADOLFSSON ET AL.
accuracy because such bubbles would expand under hyper-
baric conditions, causing a reduced contact area between the
sensor electrode and the surrounding interstitial fluid. The
noted difference in MARD between the hypobaric and hy-
CGM has been shown to reduce the incidence and duration
of hypoglycemia7,15and to reduce the number of severe
hypoglycemic episodes in subgroups of patients5as well as
in individuals with well-regulated diabetes.16In persons
with type 1 diabetes, especially those under intensive insulin
treatment, hypoglycemia is common not only during day-
time but also at night.17Studies show that physical activity is
sensors and their accuracy, measured
as mean absolute relative difference
(MARD) (%), in hypo- and hyperbaric
conditions. The respective condition’s
MARD was lower during the hyperbaric
test (6.7%) compared with the hypobaric
test (14.9%) (P<0.001).
conditions. The box-plots present five statistical measures. The whiskers show minimum, maximum. The outer parts of the
box show the range P25–P75, and the line in the box represents the median. The sensor sensitivity showed a slight decrease
(P<0.05) during hypobaric but not during hyperbaric conditions. ISIG, electrical output of the glucose sensor; P-glucose,
capillary blood glucose measured with the HemoCue device.
Variation of sensor sensitivity within the instrument’s specification 1.5–15, in (A) hypobaric and (B) hyperbaric
ACCURACY OF CGM DURING PRESSURE CHANGES 531
associated to an increased variability in glucose before, dur-
ing, and after physical exercise.18CGM also reveals delayed
nocturnal hypoglycemia after intermittent high-intensity ex-
ercise.19In spite of the differences found during the hypobaric
and hyperbaric tests, the stability and level of accuracy,
shown in our study, is of great importance because one of the
main benefits would be to both identify and reduce the oc-
currence of hypoglycemia in hypo- and hyperbaric condi-
tions. CGM has considerable potential to become a tool to
detect but also prevent hypoglycemia during special cir-
cumstances such as when flying or diving. However, to en-
sure the value of CGM in these conditions, further studies are
needed on type 1 diabetes subjects as well as during large
changes in glucose levels at different pressure conditions.
This work was supported by unrestricted grants from
Medtronic and a grant from the Center of Clinical Research at
the County Council of Va ¨rmland, Sweden. The skillful tech-
nical assistance of Joakim Trogen, M.D., at the Hyperbaric
Medicine Unit at Sahlgrenska University Hospital and Peter
Kronlund at the Hyperbaric Medicine Unit at Karolinska
acknowledged. We also acknowledge the skillful assistance of
Raghavendhar Gautham, M.S., Medtronic Diabetes (Sensor
R&D), Northridge, CA.
Sweden, is gratefully
Author Disclosure Statement
K.C. is an employee of Medtronic. No competing financial
interests exist for the other authors.
1. Karon BS, Boyd JC, Klee GG: Glucose meter performance
criteria for tight glycemic control estimated by simulation
modeling. Clin Chem 2010;56:1091–1097.
2. Tang Z, Louie RF, Lee JH, Lee DM, Miller EE, Kost GJ:
Oxygen effects on glucose meter measurements with glucose
dehydrogenase- and oxidase-based test strips for point-of-
care testing. Crit Care Med 2001;29:1062–1070.
3. Gautier JF, Bigard AX, Douce P, Duvallet A, Cathelineau G:
Influence of simulated altitude on the performance of five
blood glucose meters. Diabetes Care 1996;19:1430–1433.
4. Fink KS, Christensen DB, Ellsworth A: Effect of high altitude
on blood glucose meter performance. Diabetes Technol Ther
5. Ryan EA, Germsheid J: Use of continuous glucose moni-
toring system in the management of severe hypoglycemia.
Diabetes Technol Ther 2009;11:635–639.
6. Juvenile Diabetes Research Foundation Continuous Glucose
Monitoring Study Group: Effectiveness of continuous glu-
cose monitoring in a clinical care environment: evidence
from the Juvenile Diabetes Research Foundation continuous
glucose monitoring (JDRF-CGM) trial. Diabetes Care 2010;
7. Davey RJ, Jones TW, Fournier PA: Effect of short-term use of
a continuous glucose monitoring system with a real-time
glucose display and a low glucose alarm on incidence and
duration of hypoglycemia in a home setting in type 1 dia-
betes mellitus. J Diabetes Sci Technol 2010;4:1457–1464.
8. Adolfsson P, Ornhagen H, Jendle J: The benefits of contin-
uous glucose monitoring and a glucose monitoring schedule
in individuals with type 1 diabetes during recreational div-
ing. J Diabetes Sci Technol 2008;2:778–784.
9. Adolfsson P, Ornhagen H, Jendle J: Accuracy and reliability
of continuous glucose monitoring in individuals with type 1
diabetes during recreational diving. Diabetes Technol Ther
10. Riddell MC, Milliken J: Preventing exercise-induced hypo-
glycemia in type 1 diabetes using real-time continuous glu-
cose monitoring and a new carbohydrate intake algorithm:
an observational field study. Diabetes Technol Ther 2011;13:
11. Arntzen A, Eidsvik S, Risberg J, eds. Norwegian Diving- and
Treatment Tables: Tables and Guidelines for Surface Or-
ientated Diving on Air and Nitrox. Tables and Guidelines
for Treatment of Decompression Illness, 3rded. Bjørndalstræ,
Norway: Barotech AS, 2008.
12. Bonomo M, Cairoli R, Verde G, Morelli L, Moreo A,
Grottaglie MD, Brambilla MC, Meneghini E, Aghemo P,
Corigliano G, Marroni A: Safety of recreational scuba diving
in type 1 diabetic patients: the Deep Monitoring programme.
Diabetes Metab 2009;35:101–107.
13. Mastrototaro J, Shin J, Marcus A, Sulur G: The accuracy and
efficacy of real-time continuous glucose monitoring sensor in
patients with type 1 diabetes. Diabetes Technol Ther 2008;10:
14. Kovatchev BP, Shields D, Breton M: Graphical and numer-
ical evaluation of continuous glucose sensing time lag.
Diabetes Technol Ther 2009;11:139–143.
15. Battelino T, Phillip M, Bratina N, Nimri R, Oskarsson P, Bo-
linder J: Effect of continuous glucose monitoring on hypo-
glycemia in type 1 diabetes. Diabetes Care 2011;34:795–800.
16. Juvenile Diabetes Research Foundation Continuous Glucose
Monitoring Study Group: The effect of continuous glucose
monitoring in well-controlled type 1 diabetes. Diabetes Care
17. Juvenile Diabetes Research Foundation Continuous Glucose
Monitoring Study Group: Prolonged nocturnal hypoglyce-
mia is common during 12 months of continuous glucose
monitoring in children and adults with type 1 diabetes.
Diabetes Care 2010;33:1004–1008.
18. Kapitza C, Hovelmann U, Nosek L, Kurth HJ, Essenpreis M,
Heinemann L: Continuous glucose monitoring during exer-
cise in patients with type 1 diabetes on continuous subcuta-
neous insulin infusion. J Diabetes Sci Technol 2010;4:123–131.
19. Maran A, Pavan P, Bonsembiante B, Brugin E, Ermolao A,
Avogaro A, Zaccaria M: Continuous glucose monitoring
reveals delayed nocturnal hypoglycemia after intermittent
high-intensity exercise in nontrained patients with type 1
diabetes. Diabetes Technol Ther 2010;12:763–768.
Address correspondence to:
Peter Adolfsson, M.D.
The Queen Silvia Children’s Hospital
S-41685 Go ¨teborg, Sweden
532ADOLFSSON ET AL.