Clinical Care/Education/Nutrition

O R I G I N A L A R T I C L E

Limitations of Conventional Methods of

Self-Monitoring of Blood Glucose
Lessons learned from 3 days of continuous glucose sensing in pediatric
patients with type 1 diabetes
ELIZABETH BOLAND, MSN APRN, CDE1 CYNTHIA A. BRANDT, MD, MPH1 glycemia than adults in the intensive
TERESA MONSOD, MD1 SANJAY FERNANDO, SC, MS1 treatment group (4), and very young chil-
MARIA DELUCIA, MS1 WILLIAM V. TAMBORLANE, MD1,2 dren with type 1 diabetes have been
shown to be at high risk for asymptomatic
nocturnal hypoglycemia (5).
Intensive treatment of type 1 diabetes
was made possible by the introduction of
OBJECTIVE — Children with type 1 diabetes are usually asked to perform self-monitoring of new methods of monitoring glycemic
blood glucose (SMBG) before meals and at bedtime, and it is assumed that if results are in target control and new strategies of insulin de-
range, along with HbA1c measurements, then overall glycemic control is adequate. However, the
brief glimpses in the 24-h glucose profile provided by SMBG may miss marked glycemic excur-
livery that were introduced in the late
sions. The MiniMed Continuous Glucose Monitoring System (CGMS) has provided a new 1970s and 1980s. Self-monitoring of
method to obtain continuous glucose profiles and opportunities to examine limitations of con- blood glucose (SMBG) with multiple daily
ventional monitoring. injections or insulin pump therapy (con-
tinuous subcutaneous insulin infusion
RESEARCH DESIGN AND METHODS — A total of 56 children with type 1 diabetes [CSII]) offered the possibility of control-
(age 2–18 years) wore the CGMS for 3 days. Patients entered four fingerstick blood samples into ling postprandial hyperglycemia and re-
the monitor for calibration and kept records of food intake, exercise, and hypoglycemic symp- ducing the risks of severe hypoglycemia.
toms. Data were downloaded, and glycemic patterns were identified. However, most youths with type 1 diabe-
tes only measure premeal blood glucose
RESULTS — Despite satisfactory HbA1c levels (7.7 ⫾ 1.4%) and premeal glucose levels near
the target range, the CGMS revealed profound postprandial hyperglycemia. Almost 90% of the
levels during the day and rarely measure
peak postprandial glucose levels after every meal were ⬎180 mg/dl (above target), and almost glucose levels during the night, the time
50% were ⬎300 mg/dl. Additionally, the CGMS revealed frequent and prolonged asymptomatic of greatest vulnerability to hypoglycemia
hypoglycemia (glucose ⬍60 mg/dl) in almost 70% of the children. (5). Thus, marked glycemic excursions
are undoubtedly missed by the brief
CONCLUSIONS — Despite excellent HbA1c levels and target preprandial glucose levels, glimpses into the 24-h glucose profiles
children often experience nocturnal hypoglycemia and postprandial hyperglycemia that are not provided by SMBG. Consequently, the re-
evident with routine monitoring. Repeated use of the CGMS may provide a means to optimize cent development of methods for home
basal and bolus insulin replacement in patients with type 1 diabetes. continuous monitoring of extracellular
glucose has the potential to be the most
Diabetes Care 24:1858 –1862, 2001
important advance in the management of
youth with type 1 diabetes in the past 20
years.

T
he findings of the Diabetes Control trol will optimize growth and normal pu- The Continuous Glucose Monitoring
and Complications Trial (DCCT) bertal development (3), as well as System (CGMS) developed by MiniMed is
and the U.K. Prospective Diabetes decrease the risks of microvascular com- the first system for continuous glucose
Study (UKPDS) have demonstrated that plications. However, near-normal glucose monitoring approved by the U.S. Food
the goals of treatment of diabetes should control is more difficult to achieve in pe- and Drug Administration. The MiniMed
be to achieve glycemic control as close to diatric versus adult patients with type 1 CGMS uses a glucose oxidase– based sen-
normal and as possible (1,2). In youth diabetes. DCCT adolescents had a higher sor to measure extracellular fluid glucose
with type 1 diabetes, strict diabetes con- HbA1c and a greater risk of severe hypo- in subcutaneous tissue, which is cali-
brated against corresponding blood glu-
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
cose levels. It is approved for use as a
From the 1Yale University School of Medicine, Yale University, New Haven; and the 2Yale Children’s Clinical Holter-type monitor. The device does not
Research Center, Yale University, New Haven, Connecticut.
Address correspondence and reprint requests to Elizabeth Boland, MSN, APRN, CDE, Yale University give real-time glucose values to the wear-
PPRU, 2 Church St. South, Suite 111, New Haven, CT 06519. E-mail: [email protected]. er; data can only be downloaded by clini-
Received for publication 9 March 2001 and accepted in revised form 26 July 2001. cians after the fact. It remains to be seen
E.B. and W.V.T. have received honoraria from and are paid consultants of MiniMed. whether repeated use of the CGMS will
Abbreviations: CGMS, Continuous Glucose Monitoring System; CSII, continuous subcutaneous insulin
infusion; DCCT, Diabetes Control and Complications Trial; SMBG, self-monitoring of blood glucose.
have a favorable impact on overall diabe-
A table elsewhere in this issue shows conventional and Système International (SI) units and conversion tes control. However, because the wearer
factors for many substances. is masked to the sensor data, first-time use

1858 DIABETES CARE, VOLUME 24, NUMBER 11, NOVEMBER 2001

 Boland and Associates

Table 1—Clinical characteristics of sample Procedures fluid, which generates an electrical cur-
All subjects were seen in the outpatient rent. The current is carried by a cable to a
Age (years) 11.6 ⫾ 4.6 Yale Children’s Clinical Research Center, pager-size monitor that analyzes the data
Male (%) 44.6 usually in the late afternoon after school. every 10 s and reports average values ev-
White (%) 96.4 Demographic and clinical data were col- ery 5 min, giving a total of 288 readings
Duration of diabetes (years) 5⫾3 lected using a standardized data collec- per day. Sensor readings are calibrated by
Treatment modality (n) tion form. HbA1c levels were measured, the monitor against capillary blood glu-
One pump 42 and the sensor was inserted by the same cose measurements obtained with con-
Two injections daily 12 investigator (advanced practice nurse) for ventional SMBG meters. Patients were
Three injections daily 1 all subjects. Patients/families were in- asked to perform at least four premeal/
Four injections daily 1 structed on the use of the CGMS, and they snack SMBG tests. Each sensor is used
Insulin dose (units 䡠 kg–1 0.9 ⫾ 0.3 were asked to enter a minimum of four continuously for up to 72 h. Glucose val-
䡠 day–1) SMBG samples into the monitor for cali- ues outside the range of 40 – 400 mg/dl
HbA1c (%) 7.7 ⫾ 1.4 bration and to keep detailed written are reported as ⱕ40 or ⱖ400 mg/dl. A
records of insulin administration, food in- representative 24-h sensor tracing is
Data are means ⫾ SD, unless otherwise indicated.
take, exercise, and hypoglycemia symp- shown in Fig. 1.
toms. Patients also entered event markers
of the system provides a unique opportu- into the monitor for these events. Partici- HbA1c measurements
nity to examine how well standard SMBG pants were encouraged to call the investi- HbA1c was measured using the DCA 2000
reflects 24-h glucose excursions in youth gator with any questions. After 3 days, the (Bayer, Tarrytown, NY) instrument (non-
with diabetes. patient returned with the system, and diabetic range 4.3– 6.3). The interassay
data were downloaded using the Mini- coefficient of variation for our DCA 2000
RESEARCH DESIGN AND Med Solutions Software version 2.0b instrument was 3.6% at a normal HbA1c
METHODS — Patients were drawn (Northridge, CA). Insertion sites were in- level (5.3%) and 2.7% at a moderately el-
from the Yale Children’s Diabetes Clinic, spected for evidence of inflammation or evated level (9.2%).
which cares for ⬎600 youth with type 1 infection, and families were questioned
diabetes. This practice has a general goal regarding problems with the use of the Data analysis
of using therapy to attempt to achieve glu- system. Demographic data were entered into the
cose control as close to normal as possi- Yale Trial DB database and checked for
ble, with HbA 1c levels ⬍8% in all The CGMS system accuracy. Data from the sensors were im-
patients. Patients were eligible for partic- The CGMS system has been described in ported into this database (Oracle 7.3 ta-
ipation in this study if they were ⬍18 detail elsewhere (6). Briefly, the sensor is bles and Microsoft Access tables) for
years old, had no other health problem a glucose oxidase– based platinum elec- further data visualization and analysis.
except for treated thyroid disease, and trode that is inserted through an insertion The analyses were performed with SPSS
had been treated with insulin for at least 1 needle into the subcutaneous tissue of the System (version 10). Descriptive statistics
year. All patients meeting selection crite- anterior abdominal wall or other appro- were used to describe the sample. Corre-
ria were asked by an investigator to par- priate site using a spring-loaded device lations (Pearson’s) were used to compare
ticipate in this study during a routine (the Senserter). Glucose oxidase catalyzes sensor readings with SMBG results. The
diabetes clinic visit. The first 56 patients the oxidation of glucose in the interstitial frequency of hypoglycemia was described
(age 2–18 years) invited to enroll in the
study all agreed to participate, and they
are included in this analysis. Most were
using CSII rather than injection therapy,
and all patients were using lispro as their
quick-acting insulin. The parents and pa-
tients (where appropriate) gave written,
informed consent for inclusion in the
study, which was approved by the Yale
University School of Medicine Human In-
vestigations Committee. Clinical data on
entry into the study are shown in Table 1.
In general, the patients were well con-
trolled, with an HbA1c level (mean ⫾ SD)
of 7.7 ⫾ 1.4%. Patients with shorter dia-
betes duration and those who were
younger were more likely to have lower
HbA1c levels (r ⫽ 0.51, P ⬍ 0.005; r ⫽
0.31, P ⬍ 0.05; respectively). However, Figure 1—An example of a representative 24-h glucose sensor profile obtained from one of our
sex and treatment modality were not re- patients, a child aged 11 years and 9 months with type 1 diabetes of 3 years duration. Z, Meter
lated to HbA1c levels. SMBG levels used to calibrate the sensor.

DIABETES CARE, VOLUME 24, NUMBER 11, NOVEMBER 2001 1859

CGMS and pediatrics

Nighttime (11:00 P.M. to 6:00 A.M.)

glycemic control
The CGMS was particularly useful in de-
tecting asymptomatic nocturnal hypogly-
cemia. As shown in Fig. 3, 67.8% of the
subjects had a recorded nadir glucose
level ⬍60 mg/dl during at least 1 night of
sensing, and 32.1% had levels ⱕ40 mg/
dl. Glucose levels ⬍60 mg/dl were ob-
served on all 3 nights in 12 children. The
duration of nocturnal hypoglycemia was
also prolonged. When glucose levels from
all nocturnal profiles were analyzed, sen-
sor glucose levels were between 41 and 59
mg/dl for a mean of 52 min/night (median
32, range 0 –280 min), and they were
ⱕ40 mg/dl for 65 min/night (35, 0 –178).
Only one of the patients was awakened
from sleep because of hypoglycemic
symptoms. When the group as a whole
was examined, no differences in the fre-
quency of hypoglycemia were found to be
related to factors such as age, sex, HbA1c,
and treatment modality (CSII versus in-
jections). There were likewise no signifi-
Figure 2—Percentage of peak postmeal glucose levels over the target level of 180 mg/dl. 䡺, ⬎300
mg/dl; f, 214 –300 mg/dl; u, 181–240 mg/dl.
cant correlations between the duration of
nocturnal hypoglycemia (time at ⬍60,
41–59, and ⱖ0 mg/dl) and patient age,
using descriptive statistics. Two levels of entire 3-day period was remarkably sim- duration of diabetes, and HbA1c levels.
biochemical hypoglycemia were used for ilar (150 ⫾ 43 mg/dl). The overall mean Similarly, treatment modality and sex did
these analyses, values from 41 to 59 mg/dl glucose levels obtained by each method in not significantly affect nocturnal hypogly-
and those ⱕ40 mg/dl. Readings between each subject correlated closely with one cemia duration. However, the 12 young-
6:00 A.M. and 11:00 P.M. were considered another (r ⫽ 0.78). sters with nocturnal hypoglycemia on all
daytime values, and those between 11:00 3 nights had lower HbA1c levels (6.9 ⫾
P.M. and 6:00 A.M. were nighttime values. 0.8%) than the other patients, who had at
The peak postprandial glucose level was Daytime (6:00 A.M. to 11:00 P.M.) least 1 night without hypoglycemia
the highest sensor value during the 3 h glycemic control (7.9 ⫾ 1.4%, P ⬍ 0.01); other clinical
after each main meal. In agreement with meter glucose read- factors did not differ between these two
ings, which were used to calibrate the sen- groups.
RESULTS — In 2 of the 56 children, sor, premeal sensor values tended to be
the first sensor had to be replaced during lower before breakfast than before other
the initial visit because it failed to meet meals (126 ⫾ 43, 134 ⫾ 83, and 145 ⫾ CONCLUSIONS — The present study
performance parameters. During home 92 mg/dl for breakfast, lunch, and dinner, demonstrates that standard premeal and
use, 50 of the 56 patients successfully bedtime SBGM provides a surprisingly ac-
respectively). Despite glucose levels in or
used the CGMS for ⬎60 h, which in- curate reflection of overall glycemic con-
near our target range before meals, the
cluded three overnight profiles. In the trol in youth. During the 3-day period of
sensor demonstrated that peak postpran-
other six children, the sensor was either study, the mean of the 12–15 m blood
dislodged or disconnected prematurely. dial values were markedly elevated (293 ⫾ glucose levels were remarkably similar to
The system was well tolerated by all sub- 84, 291 ⫾ 81, and 280 ⫾ 80 mg/dl for the mean of ⬎800 sensor glucose levels
jects, and there was no evidence of infec- postbreakfast, postlunch, and postdin- obtained during the day and night over
tion or inflammation at the insertion site. ner). As shown in Fig. 2, after meals the same time period. In addition, both
⬃90% of the peak postprandial glucose mean meter and sensor levels were con-
Overall glycemic control levels exceeded our postprandial target of sistent with HbA1c values. Normograms
The overall mean of the 10 –15 SMBG lev- ⬍180 mg/dl, with almost 50% of these indicate that our patients’ mean HbA1c of
els obtained by the subjects before meals readings being ⬎300 mg/dl. There were 7.7% is equivalent to a 3-month average
during the 3 days of sensor use was 155 ⫾ also 0.9 events 䡠 patient–1 䡠 day–1 in which of 150 –160 mg/dl (7). It is noteworthy
38 mg/dl. In comparison, the average of sensor glucose levels fell below 60 mg/dl that only 3 days of monitoring with either
⬎800 sensor readings (which included between breakfast and bedtime. Only 29% method related so well to HbA1c levels.
the 10 –15 points used for calibration) ob- of these events were accompanied by hy- Whereas SMBG may provide a sur-
tained during the day and night over the poglycemia symptoms. prisingly robust estimate of overall glu-

1860 DIABETES CARE, VOLUME 24, NUMBER 11, NOVEMBER 2001

 Boland and Associates

were using insulin pumps that had the

ability to vary insulin basal rates fre-
quently. However, the minimal nocturnal
glucose data that is available with SMBG
to patients and clinicians limits their abil-
ity to fully exploit this capability of CSII.
Extensive studies have been per-
formed in adults with type 1 diabetes, val-
idating sensor accuracy in the 40 – 400
mg/dl range (9,10). Nevertheless, a note
of caution needs to be sounded regarding
interpretation of low glucose levels re-
ported by the sensor at night in our pa-
tients. The relative error of all methods of
glucose monitoring increases to some ex-
tent at the lower end of the detection
range. In addition, it has not been estab-
lished in type 1 diabetic children whether
performance of the sensor is affected by
pressure, temperature, or physiological
changes that might occur while the pa-
tient is asleep. It should also be noted that
the sensor measures the extracellular fluid
glucose levels, but the monitor is cali-
brated against SMBG levels. Recent stud-
ies (11) have used microdialysis to
measure glucose concentrations directly
in extracellular fluid of muscle and adi-
Figure 3—Percentage of patients with nadir night sensor glucose level in hypoglycemic range pose tissue during insulin-induced hypo-
(either 41– 60 mg/dl or ⱕ40 mg/dl) for 1, 2, or all 3 nights of CGMS use. f, 41– 60 mg/dl; u, glycemia. In comparison to values at
ⱕ40 mg/dl. euglycemia, the concentration gradient
between plasma and extracellular fluid
cose control, results of the CGMS show potential usefulness of monitoring post- glucose increased when plasma glucose
that conventional glucose testing misses prandial as well as preprandial glucose levels were lowered (9). If the same were
the marked day-to-day excursions in levels in youth with type 1 diabetes. The true in subcutaneous tissue, the net result
plasma glucose from high to low values sensor also detected many more hypogly- would be to artificially lower sensor esti-
that characterize type 1 diabetes in chil- cemic events during the day than were mates of blood glucose concentrations.
dren. To limit the number of fingersticks, appreciated clinically. Further studies are needed to validate the
most children are asked to not test blood In a recent study, Porter et al. (5) re- accuracy of sensor glucose measurements
glucose levels immediately after meals, ported that asymptomatic, nocturnal hy- during hypoglycemia. Although there can
but rather to test before meals. It is often poglycemia was common in young be little doubt that our patients frequently
assumed that control of preprandial glu- children with type 1 diabetes. On a single experienced low glucose levels during the
cose levels is indicative of overall good night of sampling blood glucose levels at night, a question remains as to how low
daytime glucose control. Quite a different only 11:00 P.M. and 2:00 A.M., the lowest the sugar levels actually were.
picture emerged from our patients’ day- glucose level was, respectively, ⬍64 and The CGMS was well tolerated and
time CGMS profiles. Even though HbA1c ⬍45 mg/dl in 37.8 and 13% of their pa- used appropriately by our patients, and
and premeal glucose levels suggested tients, who were mostly treated with two there was no evidence of infection or in-
good diabetes control, peak postprandial injections per day. Continuous glucose flammation. The physical characteristics
values were almost always greater than monitoring for 3 successive nights pro- of the CGMS are very similar to an insulin
our target value of ⬍180 mg/dl, and al- vides a much more complete picture of pump. Interestingly, our youngsters
most 50% exceeded 300 mg/dl. This the scope of this problem. Even though found that wearing the CGMS was so easy
severe postprandial hyperglycemia was 75% of our patients were using CSII, that it encouraged 8 of the 14 injection
observed even though most patients asymptomatic nocturnal hypoglycemia patients to switch to CSII. These were
received lispro before meals. These obser- was observed in two-thirds of our pa- children or parents who previously had
vations have important clinical impli- tients, regardless of age and sex. In addi- been reluctant to try CSII because it in-
cations, because recent evidence suggests tion to being frequent and significant volved wearing an external device.
that postprandial hyperglycemia plays a (ⱕ40 mg/dl), reductions in glucose levels Results of this study demonstrate that
particularly important role in the devel- during the night were also prolonged. there is more to controlling diabetes than
opment of vascular complications of dia- This degree of nocturnal hypoglycemia just lowering HbA1c levels and premeal
betes (8). These data also illustrate the was observed even though most patients glucose levels toward normal. In our chil-

DIABETES CARE, VOLUME 24, NUMBER 11, NOVEMBER 2001 1861

CGMS and pediatrics

dren, these two parameters of glycemic 1999

References 7. Nathan D: Glycosylated hemoglobin:
control were closely related primarily be- 1. Diabetes Control and Complications Trial
cause elevated postprandial levels were what it is and how to use it. Clin Diab
Research Group: The effect of intensive
offset by low nighttime values. It is rea- 1:1–7, 1983
treatment of diabetes on the development
and progression of long-term complica- 8. Ceriello A: The post-prandial state and
sonable to assume, as small pilot studies
tions in insulin-dependent diabetes mel- cardiovascular disease: relevance to dia-
in both adults (12) and children (13) have betes mellitus. Diabete Metab Rev 16:125–
suggested, that the wealth of data pro- litus. N Engl J Med 329:977–986, 1993
2. UK Prospective Diabetes Study (UKPDS) 132, 2000
vided by glucose sensors will provide cli- Group: Intensive blood-glucose control 9. Gross TM, Bode BW, Einhorn D, Kayne
nicians and patients with a means to with sulfonylureas or insulin compared DM, Reed JH, White NH, Mastrototaro JJ:
optimize basal and bolus insulin replace- with conventional treatment and risk of Performance and evaluation of the Mini-
ment with type 1 diabetes. Further stud- complications in patients with type 2 di- Med Continuous Glucose Monitoring
ies in larger groups of youth with type 1 abetes. Lancet 652:837– 853, 1998 System during patient home use. Diabetes
diabetes are needed to determine whether 3. Peters AL, Davidson MB, Schriger DL, Technol Ther 2:49 –56, 2000
Hasselblad V: A clinical approach for the 10. Gross TM, Mastrototaro JJ: Efficacy and
repeated use of the CGMS can lower
diagnosis of diabetes mellitus: an analysis reliability of the Continuous Glucose
HbA1c levels and reduce the risk of hypo- using glycosylated hemoglobin levels. Monitoring System. Diabetes Technol Ther
glycemia. It is even more exciting to con- Meta-analysis Research Group on the Di- 2:S19 –S26, 2000
sider that this technological advance may agnosis of Diabetes Using Glycated He- 11. Kamel A, Norgren S, Persson B, Marcus C:
represent a first step toward the develop- moglobin Levels. JAMA 276:1246 –1252, Insulin induced hypoglycaemia: compar-
ment of a true and practically applicable 1996 ison of glucose and glycerol concentra-
artificial pancreas. 4. Diabetes Control and Complications Trial tions in plasma and microdialysate from
Research Group: Effect of intensive diabe- subcutaneous adipose tissue. Arch Dis
tes treatment on the development of long-
Child 80:42– 45, 1999
term complications in adolescents with
12. Bode BW, Gross TM, Thornton KR, Mas-
insulin-dependent diabetes mellitus: Dia-
Acknowledgments — This study was sup- betes Control and Complications Trial. trototaro JJ: Continuous glucose monitor-
ported by grants from the National Institutes J Pediatr 125:177–188, 1994 ing used to adjust diabetes therapy
of Health (RR-06022), the Juvenile Diabetes 5. Porter PA, Keating B, Byrne G, Jones TW: improves glycosylated hemoglobin: a pi-
Research Foundation, the Pediatric Pharma- Incidence and predictive criteria of noc- lot study. Diabetes Res Clin Pract 46:183–
cology Research Unit Network (HD 37251), turnal hypoglycemia in young children 190, 1999
the Stephan Morse Pediatric Diabetes Research with insulin-dependent diabetes mellitus. 13. Chase HP, Kim LM, Owen SL, MacKenzie
Fund, and MiniMed. J Pediatr 130:339 –341, 1997 TA, Klingensmith GJ, Murtfeldt R, Garg
This research was presented in part at the 6. Mastrototaro JJ: The MiniMed Continu- SK: Continuous subcutaneous glucose
61st annual meeting of the American Diabetes ous Glucose Monitoring System (CGMS). monitoring in children with type 1 diabe-
Association, Philadelphia, PA, 22–26 June 2001. J Pediatr Endocrin Metab 12:751–758, tes. Pediatr 107:222–226, 2001

1862 DIABETES CARE, VOLUME 24, NUMBER 11, NOVEMBER 2001