DIABETES TECHNOLOGY & THERAPEUTICS

Volume 20, Number 1, 2018

ª Mary Ann Liebert, Inc.
DOI: 10.1089/dia.2017.0287

ORIGINAL ARTICLE

Flash Glucose Measurements in Children

with Type 1 Diabetes in Real-Life Settings:
To Trust or Not to Trust?
Agnieszka Szadkowska, MD, PhD,1 Andrzej Gawrecki, MD, PhD,2
Arkadiusz Michalak,1 Dorota Zozulińska-Zió1kiewicz, MD, PhD,2
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.

Wojciech Fendler, MD, PhD,1,3 and Wojciech M1ynarski, MD, PhD1

Abstract
Background and Aims: To evaluate the clinical accuracy of a flash glucose monitoring device FreeStyle Libre
(FSL) among children with type 1 diabetes in real-world settings during a summer camp.
Materials and Methods: During a summer camp, children with type 1 diabetes (n = 79, aged 8–18 years) were
provided with FSLs for 12 days. On days 3, 7, and 11 of the study, they underwent supervised glucose testing at
8 timepoints. Glycemia was estimated by using FSL and measured with a personal glucometer within a period
of 2 min. The glucose trend arrows were recorded.
Results: The study was completed by 78 children (median: age 12.8 years, diabetes duration 5.8 years, HbA1c
58.5 mmol/mol). Mean absolute relative difference (MARD) between the FSL and the glucometer was
13.5% – 12.9%. FSL was the most accurate in stable glycemic conditions: MARD 11.4% – 10.4%, less accu-
rate when glycemia was falling >2 mg/(dL$min) [0.111 mmol/(L$min)—MARD 22.6% – 18.6%; P < 0.001 vs.
stable conditions] and when the device could not determine the glucose trends (16.5% – 16.3%, P = 0.01 vs.
stable conditions). The FSL demonstrated lower accuracy during the day than the night [MARD 14.9% – 14%
vs. 11.2% – 10.6%, P < 0.0001]. Out of 1655 data pairs of glucometer and FSL, using the Surveillance Error
Grid methodology we determined that 80.36% of FSL readings were associated with no clinical risk, 18.73%
with slight risk and only one high-risk measurement was detected.
Conclusion: FSL is accurate in children, but its accuracy depends on the glucose trend. Results flagged by the
rapid fall flag and ‘‘trend undetermined’’ should be verified by blood glucose measurements.

Keywords: Type 1 diabetes mellitus, Blood glucose self-monitoring, Children, Flash glucose monitoring,
FreeStyle Libre, Accuracy.

Introduction of diabetic ketoacidosis.5 The guidelines of the American

Diabetes Association have recently increased the recommended

T he key aim of therapy in type 1 diabetes mellitus

(T1DM) is to safeguard patients from acute and chronic
complications and to improve health-related quality of life.
number of BG measurements, especially in patients with a
short duration of diabetes or long life expectancy6 who should
perform SMBG tests 6 to 12 times daily. Such a number is,
The DCCT and EDIC trials proved that lowering HbA1c however, difficult to achieve, especially for children and young
protects from vascular complications and cardiovascular people, for whom the invasiveness of SMBG is a significant
disease.1–3 barrier for self-monitoring.7
Maintaining a high frequency of self-monitoring of blood To improve this situation, other systems have been intro-
glucose (SMBG) is the cornerstone for meeting HbA1c goals duced to replace traditional capillary BG measurement. Con-
and preventing serious hypoglycemia.4 An increased num- tinuous glucose monitoring (CGM) offers efficient glycemic
ber of tests is also associated with significantly lower rates control with a reduced need for manual tests and associated
Departments of 1Pediatrics, Oncology, Hematology and Diabetology, and 3Biostatistics and Translational Medicine, Medical University
of Lodz, Lodz, Poland.
2
Department of Internal Medicine and Diabetology, Poznan University of Medical Science, Poznan, Poland.

1
 2 SZADKOWSKA ET AL.
inconveniences. Recent studies have demonstrated that the
use of CGM systems reduces glycemic variability and time
spent in hypoglycemia of subjects with diabetes, and HbA1c,
in well-controlled individuals and those using insulin
pumps.8–12 A health-economic analysis by Roze et al. found
that sensor-augmented pump therapy represents good value
for money in terms of improving quality-adjusted life ex-
pectancy and delaying diabetes complications13; however,
the high cost of CGM still hinders its widespread use.
Recently, a new flash glucose monitoring (FGM) system
(Abbott Diabetes Care, Alameda, CA) was introduced into
the market along with a dedicated device: FreeStyle Libre
(FSL). It offers a factory-calibrated sensor that can be worn for
14 days without manual calibration by using blood testing, a
reader combined with a standard glucometer, and dedicated
software. The user receives a current glucose measurement
along with historic results from the preceding 8 h by scanning
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.
the sensor with the FreeStyle reader. Recordings are taken in
15-min intervals. The glucose trends are displayed as arrows
on the reader along with alerts. As the device requires active
scanning by the user, it does not report alerts in real time.
FSL was proposed as an alternative for those who cannot
afford CGM, find CGM too complicated, or complain of alarm
fatigue. Children and adolescents could clearly benefit from
using the FGM system.14 However, initial observations sug-
gested that its users may forego SMBG measurements and rely
solely on FGM. This raises the question—Is FGM accurate
enough to replace standard glucometers, or should it be just a
complementary addition to SMBG? So far, only one study
tested the accuracy of FSL in a pediatric population, and the
data obtained by FGM were hidden from participants.14
This study examines the introduction of FSL to children
with T1D in a real-world setting during a summer camp. During
the study, the children could observe current glucose levels, trend
arrows, and the 8-h glucose concentration history on a reader
screen. The main aim of this study was to establish the accuracy
of FSL FGM system in reference to the glucometer and pinpoint
the factors that may affect said accuracy. The secondary objec-
tive was to assess the children’s attitude toward the device and its
possible association with measurement accuracy.
Materials and Methods
This was a prospective single-arm study conducted during an
annual summer camp for children and adolescents with T1D.
The study protocol was approved by the Ethical Committee of
the Medical University of Lodz (No. RNN/223/16/KE) and
registered in the German Clinical Trials Register (DRKS-ID:
DRKS00011751). Informed written consent was obtained from
parents, and so was informed assent from the children.
Children from 12 diabetes care centers from Poland took part
in the camp, which took place in the period from July 24 to
August 5, 2016. They were provided with full medical care
and asked whether they would like to use FSL. Out of 80 ap-
proached children, 79 accepted the opportunity to participate in
the study. One child did not complete the study. FSL readers
and sensors were provided by Abbott Laboratories Poland LLC.
On the second day of the camp, the children were introduced
to the FGM system and the sensors were attached to the back
of their upper arms, according to the manufacturer’s instruc-
tions. In the case of the sensor dislodging or any technical
issues, the sensors were replaced in each child up to two times.
The children had basic training in using FSL, limited to
device handling and explanation of trends and alerts. On the
seventh day, the authors collected feedback from children
and created initial clinical guidelines for the interpretation of
glucose trend arrows, which were subsequently taught to the
children.
On the 3rd, 7th, and 11th days of FSL use, the children
underwent supervised glucose testing at eight predefined time
points each day: before and 2 h after breakfast, before and 2 h
after their midday meal, before and 2 h after dinner, at mid-
night, and at 3 a.m. At each timepoint, glycemia was mea-
sured twice within 2 min: once in the interstitial fluid by
scanning the FGM sensor, and again in the capillary blood by
using a Contour Plus One glucometer (Ascensia Diabetes Care
Poland, Warszawa, Poland), compliant with ISO 15197:2013
accuracy criteria.15 The glucose trend arrows of the FGM were
recorded. The sensors were removed on the 12th (last) day of
the camp, and the skin was carefully examined by physicians.
After sensor removal, data from the readers were downloaded
and analyzed. Each reader was also reviewed manually.
Medical history was collected, and physical examination
was performed by pediatricians. At the end of the camp, body
height and weight were measured and capillary blood sam-
ples were taken for HbA1c assessment (D-10 Hemoglobin
A1c Program [Bio-Rad Laboratories, Hercules, CA, Bio-
Rad, Marnes-la-Coquette, France]). Body mass index (BMI)
was recalculated into standard deviation Z-scores based on
local BMI charts.16
On the last day, a custom-designed questionnaire was ad-
ministered to assess the degree of satisfaction with FGM use.
The children rated their experience and satisfaction with FSL
on a five-point scale from strongly agree to strongly disagree.
The questionnaire included 13 closed questions regarding
ease of using the device, pain when applying the sensor to the
arm, the comfort of wearing the sensor, possible inconve-
nience, whether the sensor got in the way of daily activities,
and the occurrence of pain or itching at the site of insertion.
The detailed questionnaire is included in the Supplemen-
tary Data (Supplementary Data are available online at http://
online.liebertpub.com/doi/suppl/10.1089/dia.2017.0287). The
users were also asked about their subjective opinion regard-
ing the decrease in the number of hypo- and hyperglycemia
episodes during the use of FGM. In the final open question,
the children were asked to describe the advantages and dis-
advantages of the device.
Statistical analysis
FSL readings were compared with glucometer readings as
the reference method. FSL readings beyond the sensor value
range (<40 mg/dL reported as LO (low) and >500 mg/dL
reported as HI (high)) were included in the analysis as 40 and
500 mg/dL, respectively. These extreme readings, however,
were excluded from the analysis of glycemia trends, as such
recordings had no defined trend mark by default. The accu-
racy was assessed by calculating the mean absolute relative
difference ( MARD) for each patient by using the fol-
lowing equation: MARD = meanj[(glucometer-FreeStyle
Libre)/glucometer) · 100%]j.
The Pearson’s test was used to determine the correlation of
the two methods, and the Bland–Altman regression was used
to assess bias (mean difference was compared with 0 by using
 ACCURACY OF FREESTYLE LIBRE IN CHILDREN
the t-test). The clinical accuracy of the readings was tested by
constructing a Clarke Error Grid, a Consensus Error Grid, and
a Surveillance Error Grid for paired FGM and glucometer
measurements.17–19 To determine the impact of clinical vari-
ables, t-tests were used for two-group comparisons, one-way
analysis of variance was used for multiple groups (with the
post hoc Tukey’s test performed for significant results), and the
Spearman’s rank correlation was used for continuous variables
that were not distributed normally.
The questionnaire assessing satisfaction with FSL was
assessed by contingency tables. The patients were divided
into subgroups depending on the answers given in the ques-
tionnaire, and their FGM accuracies were compared between
the groups by using the Kruskal–Wallis test.
Results
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.
A total of 79 children agreed to use FSL for the duration of
the summer camp; however, one girl returned home before
study completion and was excluded from the analysis. The
analyzed group, therefore, numbered 78 children (43% boys,
88.6% treated with CSII) with a median age of 12.8 years
(interquartile range [IQR]: 11.6–14.7 years) and a median
T1D duration of 5.8 years (IQR: 3.8–8.5 years). Their me-
dian HbA1c at inclusion was 7.5% (IQR: 7.0%–8.0%) or
58.5 mmol/mol (53–63.9 mmol/mol). The median standard-
ized BMI in the group was 0.29 (IQR: -0.3 to 0.89), four
children were overweight (BMI between 90th and 95th per-
centile standardized for gender and age), and two children
were obese (>95th percentile).
During the summer camp, children wore the sensors for
median time of 10.5 days (IQR 10 to complete 11 days) and
10 sensors were replaced in eight children due to detachment.
The analysis included 1655 paired sensor/glucometer
readings (median 23 pairs/patient, IQR: 21–24). Due to the
need for sensor replacement, it was not possible to collect all
measurements scheduled for the 3rd, 7th, and 11th day of
FGM. In total, 668 readings were collected from the first
timepoint, 539 from the second, and 448 from the third. The
glycemic variability in the participants over the camp dura-
tion is presented in Table 1.
FSL readings showed a strong positive correlation with glu-
cose meter measurements (r = 0.95, P < 0.0001). However, the
Bland–Altman regression (Fig. 1) showed that the mean dif-
ference (7.35 mg/dL, 0.4 mmol/L) was significantly different
from the expected value of 0 (P < 0.0001), suggesting that FSL
was biased toward an overestimation of glycemia. In line, raw
Table 1. Parameters of Glycemic Variability
in the Studied Children During Flash Glucose
Monitoring Use
Mean glycemia 167 mg/dL, 9.27 mmol/L
[mg/dL, mmol/L]
% of time in range 37.7
(70–140 mg/dL,
3.9–7.8 mmol/L)
Time above range [%] 55.4
Time in hypoglycemia [%] 8.2
Episodes of hypoglycemia 1.35 – 0.93
per day of use [N]
Average duration of 77 – 27 min
hypoglycemia episode
3
FIG. 1. FreeStyle Libre measurements compared to BG
with Bland–Altman regression plot. BG, blood glucose.
differences related to glucometer reference measurements pro-
duced a mean relative difference of -4.42% with a high standard
deviation (17.87%). The MARD between FSL and the gluc-
ometer was 13.5% – 12.9%. The MARD for each patient ranged
from 10.4% to 24.7%. The distribution of MARDs by indi-
viduals is presented in Figure 2. The accuracy of FGM did not
differ between days of sensor use (P = 0.11) and was not
associated with BMI Z-score (R = -0.11, P = 0.31) or HbA1c
(R = 0.07, P = 0.55). However, a weak positive correlation
was found between age and MARD (R = 0.25, P = 0.024), and
the sensors worn by boys displayed a lower accuracy than
those worn by girls—median MARD 14.8% (IQR: 12.4%–
17.3%) versus 12% (9.9%–13.6%), P = 0.0001.
In addition, the accuracy of FSL strongly depended on
the current trend of glucose concentration (P < 0.0001)
(Fig. 3). FSL measurements were the least accurate (MARD
22.6% – 18.6%, mean relative difference -17.23% – 23.73%)
when glycemia was falling at a rate exceeding 2 mg/(dL$min)
[0.111 mmol/(L$min)] compared with other trends (P < 0.0001).
The greatest accuracy (MARD 11.4% – 10.4%, mean relative
difference -2.32% – 15.3%) was achieved in stable condi-
tions, when glycemia changes did not exceed 1 mg/(dL$min)
[0.056 mmol/(L$min)] in either direction. Moreover, when the
reading device could not determine the glucose trends and
displayed no glucose trend arrow, the accuracy of FSL (MARD
16.5% – 16.3%, mean relative difference -6.72% – 22.28%)
was significantly inferior to measurements performed in stable
glycemic conditions (P = 0.001).
We detected a single extreme point in our data (absolute
relative difference of 125%, no glucose trend arrow) that may
have caused overestimation of the disparity between ‘‘unable
to determine tendency’’ and ‘‘stable glycemic conditions.’’
Therefore, we removed this outlier and repeated the ANOVA
analysis, which yielded significant results similar to the
original ones (data not shown).
Further, daytime measurements taken by FSL were found
to be less accurate than those taken at nighttime (i.e., mid-
night, 3 a.m. and after waking up) [MARD 14.9% – 14% vs.
11.2% – 10.6%, P < 0.0001].
Out of 1655 data pairs, 98.43% satisfied the clinical ac-
curacy criteria (class A or B) of the Clarke Error Grid (Fig. 4a
 4 SZADKOWSKA ET AL.
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.

FIG. 2. A bar graph of MARD values of individual patients. MARD, mean absolute relative difference.

and Table 2) and 99.1% satisfied that of the Consensus Error
Grid (Fig. 4b and Table 2). The Surveillance Error Grid
(Fig. 4c and Table 3) labeled 1330 measurements (80.36%) as
associated with no risk and 310 (18.73%) as associated with
slight risk. The distribution of records in different classes of the
Consensus Error Grid was similar among the patients (Sup-
plementary Figure S1).
No serious adverse effects were found to be associated
with FSL sensor use during the study. In two boys, edema and
rash were observed after sensor removal. During the camp,

FIG. 3. Accuracy of FreeStyle Libre in different glycemic conditions. The glycemic trends were determined by in-built
software in the FreeStyle Libre reading device.
 ACCURACY OF FREESTYLE LIBRE IN CHILDREN 5
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.

FIG. 4. Error Grid Analysis of FreeStyle Libre results plotted against BG concentrations with: (a) Clarke Error Grid.
(b) Consensus Error Grid. (c) Surveillance Error Grid (graph prepared through the courtesy of Prof. Boris P. Kovatchev,
Prof. Marc D. Breton, Dr. David C. Klonoff, and Mr. Christian Wakeman). The codes and definitions for each class are
summed up in Tables 2 and 3.

none of the children reported feeling pain or itching. How-
ever, in the retrospective questionnaire, 4 reported pain on
sensor insertion, 13 reported itching, and 2 reported moderate
pain during sensor use.
Sixty-eight (87.2%) children expressed a willingness to
continue FGM use (see Supplementary Figure S2). In the
open question regarding the advantages and disadvantages of
the device, most of the participants regarded the discretion
and painlessness of measurements as key advantages, and its
inaccuracy, as perceived by the children, as a disadvantage.
Interestingly, sensor accuracy was not clearly associated
with the accuracy of FSL, as perceived by children (P = 0.72),

Table 2. Summary of Clarke Error Grid and Consensus Error Grid Analysis
Clarke Error Grid Consensus Error Grid
Coding Definition N (%) Definition N (%)
Class A B Within 20% of reference value 1329 (80.3) Clinically accurate, no effect 1460 (88.22)
on clinical action
Class B  Outside of 20% range of 300 (18.13) Possibly causing an altered 180 (10.88)
reference value but not leading clinical action but with little
to inappropriate treatment or no effect on clinical outcome
Class C > Leading to unnecessary treatment 9 (0.54) Likely to affect the clinical outcome 14 (0.84)
Class D A Indicating a potentially dangerous 16 (0.97) Associated with significant risk 0 (0)
failure to detect hypoglycemia
or hyperglycemia
Class E
* Would confuse treatment of
hypoglycemia for hyperglycemia
1 (0.06) Having possible dangerous
consequences
1 (0.06)
 6 SZADKOWSKA ET AL.
Table 3. Summary of Surveillance Error
Grid Analysis
Degree of risk Color coding N (%)
None Deep green 1330 (80.36)
Slight, underestimated Light green 265 (13.01)
Slight, overestimated Yellow 45 (2.72)
Moderate, underestimated Light orange 14 (0.95)
Moderate, overestimated Dark orange 0 (0)
Great, underestimated Light red 1 (0.06)
Great, overestimated Dark red 0 (0)
Extreme Brown 0 (0)
their trust in the sensor readings (P = 0.33), or their willing-
ness to continue using FSL (P = 0.52).
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.
Discussion
The study examines the utility of FSL in children and ad-
olescents with T1D during a summer camp. FSL readings were
found to correlate closely with BG measurements. So far, only
three studies have analyzed the accuracy of FSL in real-life
conditions. In adults, Bailey et al. observed MARDs as high as
12% in reference to a glucometer and 12.1% in reference to
venous glycemia measured with YSI.20 Similarly, Ji et al. re-
ported MARDs of 10% in reference to BG measurements and
of 10.7% in reference to venous measurements.21 Edge et al.
reported a slightly higher MARD in children (13.9%), which
corresponds well with our findings.14
We were also the first to evaluate FGM readings system-
atically with the Bland–Altman analysis, in which FSL
demonstrated a fixed tendency to overestimate glucometer-
measured glycemia. This bias was small (mean 7.35 mg/dL,
0.4 mmol/L) and negligible in most clinical situations, but on
very rare occasions it may lead to the wrong clinical deci-
sions. For this reason, all collected data pairs were assessed
against published error grid scores17,18: the Clark, Consensus,
and Surveillance Error grids. The most recently developed
grid is the Surveillance Error Grid, which was developed in
2014. It was constructed by using feedback from medical
professionals who represent a modern approach to diabetes
care, based on DCCT study results and incorporation of
technological advances such as insulin pumps into therapy.
Our study demonstrated that most FSL readings were lo-
calized in zone A or B of the error grids for more than 98% of
cases, which means that they would not impact clinical de-
cisions in a severe manner. Other studies have reported
similar values regarding grid-assessed clinical accuracy for
FGM versus glucometer in adult20,21 and pediatric popula-
tions.14 A single incidence of significant disparity was found
between an FSL reading (<40 mg/dL, <2.2 mmol/L, reported
as LO) and BG measurement (215 mg/dL, 11.9 mmol/L),
which could be associated with potential clinically important
consequences (class E by both Clark and Consensus Error
Grid). This record was associated with sensor failure.
FGM accuracy did not correlate with the patients’ BMI or
HbA1c, which was in line with results reported by Bailey and
by Edge.20 However, in our study, MARD values were
influenced by sex and age: The accuracy was a little worse in
the boys. This may be associated with the higher physical
activity demonstrated by the boys during the summer camp;
however, as no objective measure of physical activity was
included in our study group, this is impossible to confirm.
FSL accuracy was also found to be associated with the
current glycemia trend: When glucose was falling rapidly, the
FSL reading differed by up to 20% from BG measurements.
This corresponds with results of Bonora et al., who report a
significant increase in the MARD between FGM and CGM
during hypoglycemic conditions.22 Moreover, when glycemia
was falling by more than 2 mg/(dL$min) [0.111 mmol/(L$min)],
the mean difference reached -20 mg/dL (-1.1 mmol/L), which
caused the FGM unit to overestimate the true current glycemia
value. This discrepancy may have been caused by the lag time
known to exist between glucose concentrations in the inter-
stitial fluid reaching those in the blood.20 This phenomenon
was first described by Aussedat et al.23 in rats and has become
one of the focal points in theorycrafting of artificial pancreas.
It complicates the interpretation of glucose concentrations
measured in the interstitial fluid as the delay must be accounted
for in closed-loop algorithms and by users of FGM and CGM
devices who want to make therapeutic decisions.
Importantly, the lag is unfortunately no constant difference
but its magnitude, direction, and dynamics may change, de-
pending on glucose uptake, utilization, and elimination in
peripheral tissues.24 In humans, physiological lag between
venous and interstitial glycemia was measured to be around 6
to 10 min in adults without25 and with type 1 diabetes.26 So
far, no study has investigated the physiology of the glycemic
lag in children. The tendency of FGM to overestimate falling
glycemia fits well with the ‘‘push and pull’’ hypothesis
formulated by Aussedat et al.23 and their observation that
during insulin-driven hypoglycemia the interstitial glucose
concentrations fall slower than intravenous ones.
Moreover, FSL demonstrated significantly lower accuracy
(compared with stable conditions), when the reading device
could not determine the glycemic trend. This and the rapid
decline scenario are, therefore, indicators for the patient to
verify the result with capillary glucose measurement by using
a glucometer. These results are in contrast to those reported
by Edge et al.,14 who report a steady relative difference of
6%–7% between FSL readings and BG under extreme glu-
cose trends (rapidly falling and rapidly rising) but negligible
differences between the two methods (around 1%) for other
trends. However, their study examined children using FSL in
home-based conditions, whereas this study evaluates children
during a summer camp: Increased physical activity, irregular
eating patterns, and other factors may have contributed to the
differences between the reported results and the ones shown
in our study. These conditions might make both studies in-
comparable in this regard.
In the studies regarding CGM, the main barriers to con-
tinue using them were unrealistic expectations of users and
lack of proper education.27 In our study, the children were
asked to assess the usefulness of FSL. In general, children
graded the 2-week-long experience with FSL positively.
Most of the children expressed willingness to continue using
this device. It is possible that the good reception of FGM may
result from its minimal invasiveness, when compared with
standard BG measurements, and added value in the form of
easily interpretable glucose trend arrows.
As noted in other studies, no serious adverse effects were
observed while using FSL.14,20 Skin allergic reactions were
only found in fewer than 3% of patients.
 ACCURACY OF FREESTYLE LIBRE IN CHILDREN
The limitation of our study is its short (12 days) duration.
Studies with longer observation times are needed to evaluate
whether FSL could provide lasting improvement in glycemic
outcomes. Moreover, our results are limited by the high rate
of sensor removal. More effective means are needed to
safely fix the sensor in place for studies in children.
A minor limitation was also comparing FSL readings only
with glucometer measurements. However, use of the labora-
tory reference method during the camp would demand draw-
ing venous blood in a nonsterile environment and would create
logistic problems with storage and transport of samples.
Similarly, using the Hemocue system for reference was not
possible in this study for a few reasons. First, although the
Hemocue microcuvettes can be stored in room temperature,
the measurements were often taken in the outdoor conditions
in temperatures exceeding 25C; the microcuvettes are highly
moisture sensitive and could be easily damaged. Further,
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.
the sampling size needed for Hemocue is around 4 lL of blood.
We used the glucometer that required a considerably
smaller volume of only 0.6 lL to prevent children’s pain
and discomfort and to ensure good compliance. Finally, the
number of participants in our study would demand many
Hemocue analyzers to be used at once to carry out the testing.
Finally, the comparison of FGM with reference laboratory
methods has been already done by the producer and other
researchers14,20,21—the added value of our study is that it
provides an opportunity to analyze the accuracy from a user’s
point of view. The reference-related error was minimized by
using the Contour Plus One glucometer (Bayer HealthCare
LLC, Diabetes Care, Whippany, NJ) as a comparator: It has
high accuracy with an MARD of below 10% in comparison
with reference methods.15 On the other hand, such a protocol
provided great insight into FGM accuracy from the users’
perspective, as most clinical decisions are based on SMBG
performed with a glucometer.
Conclusions
Results showed that the FSL FGM device offers good
accuracy in comparison to capillary BG in children and ad-
olescents with T1DM. However, the performance of FGM is
strongly affected by the glycemia change trend at the time of
measurement. Therapeutic decisions should not be solely
based on measurements flagged by a rapid change flag on the
FGM: Such measurements should first be verified by using a
BG measurement.
Acknowledgments
The authors would like to thank Prof. Boris P. Kovatchev,
Prof. Marc D. Breton, Dr. David C. Klonoff, and Mr. Christian
Wakeman, who provided the tool for calculating the Sur-
veillance Grid Score risks, and the medical staff for col-
lecting data during the summer camp. The authors would also
like to thank the organizers of the camp, charity foundation
‘‘Diabeciaki’’ for the opportunity to carry out the study and
for providing FSL sensors for children.
Author Disclosure Statement
All authors took part in the study. None of the authors have
any competing financial interests with the content of the
study. The FSL sensors were provided by the producer to the
7
camp organizers (charity foundation ‘‘Diabeciaki’’), but Ab-
bott Laboratories Poland had no involvement in the study
design and course. A.S. reports being a speaker for Abbott
Laboratories Poland and Bayer. A.G. reports being a speaker
for Abbott Laboratories Poland and Bayer. A.M., D.Z.-Z.,
W.F., and W.M. report no conflict of interest.
References
1. Writing Group for the DCCT/EDIC Research Group TJ,
Orchard TJ, Nathan DM, et al.: Association between 7
years of intensive treatment of type 1 diabetes and long-
term mortality. JAMA 2015;313:45–53.
2. de Boer IH, Rue TC, Cleary PA, et al.: Long-term renal
outcomes of patients with type 1 diabetes mellitus and
microalbuminuria: an analysis of the Diabetes Control and
Complications Trial/Epidemiology of Diabetes Interventions
and Complications cohort. Arch Intern Med 2011;171:
412–420.
3. Nathan DM, Cleary PA, Backlund J-YC, et al.: Intensive
diabetes treatment and cardiovascular disease in patients
with type 1 diabetes. N Engl J Med 2005;353:2643–2653.
4. Miller KM, Beck RW, Bergenstal RM, et al.: Evidence of a
strong association between frequency of self-monitoring of
blood glucose and hemoglobin A1c levels in T1D exchange
clinic registry participants. Diabetes Care 2013;36:2009–
2014.
5. Ziegler R, Heidtmann B, Hilgard D, et al.: Frequency of
SMBG correlates with HbA1c and acute complications in
children and adolescents with type 1 diabetes. Pediatr Dia-
betes 2011;12:11–17.
6. American Diabetes Association: Standards of Medical Care
in Diabetes—2014. Diabetes Care 2014;37(suppl 1):S14–
S80.
7. Wagner J, Malchoff C, Abbott G: Invasiveness as a barrier
to self-monitoring of blood glucose in diabetes. Diabetes
Technol Ther 2005;7:612–619.
8. Juvenile Diabetes Research Foundation Continuous Glu-
cose Monitoring Study Group, Beck RW, Hirsch IB, et al.:
The effect of continuous glucose monitoring in well-controlled
type 1 diabetes. Diabetes Care 2009;32:1378–1383.
9. Danne T, de Valk HW, Kracht T, et al.: Reducing gly-
caemic variability in type 1 diabetes self-management with
a continuous glucose monitoring system based on wired
enzyme technology. Diabetologia 2009;52:1496–1503.
10. Hirsch IB, Abelseth J, Bode BW, et al.: Sensor-augmented
insulin pump therapy: results of the first randomized treat-
to-target study. Diabetes Technol Ther 2008;10:377–383.
11. Battelino T, Phillip M, Bratina N, et al.: Effect of contin-
uous glucose monitoring on hypoglycemia in type 1 dia-
betes. Diabetes Care 2011;34:795–800.
12. Bergenstal RM, Tamborlane WV, Ahmann A, et al.: Ef-
fectiveness of sensor-augmented insulin-pump therapy in
type 1 diabetes. N Engl J Med 2010;363:311–320.
13. Roze S, Saunders R, Brandt A-S, et al.: Health-economic
analysis of real-time continuous glucose monitoring in
people with Type 1 diabetes. Diabet Med 2015;32:618–626.
14. Edge J, Acerini C, Campbell F, et al.: An alternative
sensor-based method for glucose monitoring in children
and young people with diabetes. Arch Dis Child 2017;102:
543–549.
15. Bailey T, Wallace JF, Greene C, et al.: Accuracy and user
performance evaluation of the Contour Next Link 2.4 blood
glucose monitoring system. Clin Chim Acta 2015;448:
139–145.
 8 SZADKOWSKA ET AL.
16. Nawarycz T, Ostrowska-Nawarycz L: [Body mass index in
the school age children and youth from the city of Lodz].
Pol Merkur Lekarski 2007;23:264–270.
17. Klonoff DC, Lias C, Vigersky R, et al.: The surveillance
error grid. J Diabetes Sci Technol 2014;8:658–672.
18. Parkes JL, Slatin SL, Pardo S, Ginsberg BH: A new con-
sensus error grid to evaluate the clinical significance of
inaccuracies in the measurement of blood glucose. Diabetes
Care 2000;23:1143–1148.
19. Clarke WL, Cox D, Gonder-Frederick LA, et al.: Evaluat-
ing clinical accuracy of systems for self-monitoring of blood
glucose. Diabetes Care 1987;10:622–628.
20. Bailey T, Bode BW, Christiansen MP, et al.: The per-
formance and usability of a factory-calibrated flash glu-
cose monitoring system. Diabetes Technol Ther 2015;17:
787–794.
21. Ji L, Guo X, Guo L, et al.: A multicenter evaluation of
the performance and usability of a novel glucose monitor-
Downloaded by Gothenburg University Library from online.liebertpub.com at 12/14/17. For personal use only.
ing system in Chinese adults with diabetes. J Diabetes Sci
Technol 2017;11:290–295.
22. Bonora B, Maran A, Ciciliot S, et al.: Head-to-head com-
parison between flash and continuous glucose monitoring
systems in outpatients with type 1 diabetes. J Endocrinol
Invest 2016;39:1391–1399.
23. Aussedat B, Dupire-Angel M, Gifford R, et al.: Interstitial
glucose concentration and glycemia: implications for con-
tinuous subcutaneous glucose monitoring. Am J Physiol
Endocrinol Metab 2000;278:E716–E728.
24. Kulcu E, Tamada JA, Reach G, et al.: Physiological differ-
ences between interstitial glucose and blood glucose mea-
sured in human subjects. Diabetes Care 2003;26:2405–2409.
25. Basu A, Dube S, Slama M, et al.: Time lag of glucose
from intravascular to interstitial compartment in humans.
Diabetes 2013;62:4083–4087.
26. Basu A, Dube S, Veettil S, et al.: Time lag of glucose from
intravascular to interstitial compartment in type 1 diabetes.
J Diabetes Sci Technol 2015;9:63–68.
27. Kubiak T, Mann CG, Barnard KC, Heinemann L: Psy-
chosocial aspects of continuous glucose monitoring: con-
necting to the Patients’ Experience. J Diabetes Sci Technol
2016;10:859–863.
Address correspondence to:
Agnieszka Szadkowska, MD, PhD
Department of Pediatrics, Oncology,
Hematology and Diabetology
Medical University of Lodz
Sporna 36/50
Lodz 91-738
Poland
E-mail: agnieszka.szadkowska@wp.pl