You are on page 1of 5


Volume 21, Number 1, 2019

ª Mary Ann Liebert, Inc.
DOI: 10.1089/dia.2018.0292


Use of Continuous Glucose Monitoring Trends

to Facilitate Exercise in Children with Type 1 Diabetes
Marie-Anne Burckhardt, MD,1–3,* Tarini Chetty, MD,1,2,* Grant J. Smith, BSc,1 Peter Adolfsson, MD,4,5
Martin de Bock, PhD,1,2 Timothy W. Jones, MD,1–3 and Elizabeth A. Davis, PhD1–3
Downloaded by Ghana Hinari Band 1 from at 01/20/19. For personal use only.

Diabetes care during exercise frequently requires interruptions to activity and adds extra challenges particularly for
young individuals with type 1 diabetes (T1D). This study investigated the use of a carbohydrate (CHO) intake
algorithm based on continuous glucose monitoring (CGM) trends during physical activity. Children with T1D
diagnosed for >1 year, ages 8–12 years, with a glycated hemoglobin of <10% were recruited into a randomized
crossover study. They attended two similar mornings of fun-based physical activity and adhered to either a CHO intake
algorithm based on CGM trends (intervention) or to standard exercise guidelines (consumption of 0.5 g CHO/kg/h
when glucose <8 mmol/L) (control). Outcome measures included events such as exercise interruptions, CHO intake,
and hypoglycemia events and percentage time spent in different sensor glucose ranges. Fourteen children completed
the study. No episodes of significant hypoglycemia (sensor glucose level <3.0 mmol/L) occurred in either arm. Mean
CHO intake was the same in both arms, 0.3 – 0.2 g/kg/h. However, the intervention algorithm resulted in fewer CHO
intake events per day: rate [95% confidence interval] 2.4 [1.6–2.3] versus 0.9 [0.4–1.5], P < 0.001, and exercise
interruptions: 7.2 [5.9–8.8] versus 1.4 [0.8–2.1], P < 0.001, compared with control. There was no evidence of a
difference in percentage time in range (3.9–10 mmol/L) and percentage time spent high between study arms. Both
control and intervention protocols prevented significant hypoglycemia. Using a CHO intake algorithm based on CGM
trends resulted in fewer CHO intake events and fewer interruptions to exercise. Use of this algorithm may reduce the
burden of diabetes management with potential to facilitate activity in young people with T1D.

Keywords: Type 1 diabetes, Continuous glucose monitoring, Exercise, Carbohydrate intake algorithm.

Introduction carbohydrate (CHO) intake.6,7 Such adjustments are gener-

ally required for prolonged (>30 min) moderate/intensity

A chieving and maintaining stable blood glucose levels

around exercise are challenging for individuals with
type 1 diabetes (T1D). In particular, exercise is associated
exercise or hypoglycemia is likely to occur, especially when
insulin levels are above basal levels.8 The amount of CHO
required to prevent hypoglycemia will depend on the amount
with an increased risk of hypoglycemia. This perceived risk of circulating insulin,9 duration and intensity of activity,10,11
can prevent young people from engaging in a physically and blood glucose level at the start of exercise,12 as well as
active lifestyle,1 despite numerous well-established health the nutritional status13 and fitness level14 of the individual.
benefits of exercise.2–5 These multifactorial effects lead to inter- and intraindividual
Strategies to prevent hypoglycemia during or after exer- variability in glycemic response to exercise, which is chal-
cise typically involve adjustment of insulin dosing and/or lenging for the person with T1D.

Children’s Diabetes Centre, Telethon Kids Institute, The University of Western Australia, Perth, Australia.
Department of Endocrinology and Diabetes, Perth Children’s Hospital, Perth, Western Australia.
Division of Paediatrics, within the Medical School, The University of Western Australia, Perth, Australia.
Department of Paediatrics, The hospital of Halland, Kungsbacka, Sweden.
Institute of Clinical Sciences, Sahlgrenska Academy at University of Gothenburg, Gothenburg, Sweden.
*These authors contributed equally to this study.
Parts of this study were presented at the Advanced Technologies and Therapeutics in Diabetes (ATTD) Conference, February 14–17,
2018, Vienna, Austria.


Continuous glucose monitoring (CGM) offers the oppor- Several outcome measures were explored: the number of
tunity to improve glucose levels during exercise by allowing exercise interruption events (such as stopping to perform self-
patients to respond not only to sensor glucose levels (SGLs) monitored blood glucose [SMBG] tests or ingest CHO), the
but also to directional arrows that indicate rates of change in frequency of CHO intake, as well as the CHO intake in grams
glycemia in real time. and the number of hypoglycemic events defined as SGL
Advancement in CGM technology means that devices are <3.9 mmol/L and <3.0 mmol/L for >20 min. Percentage time
increasingly accurate15,16 and user friendly as reflected by in- spent in different sensor glucose ranges during physical ac-
creased patient use. However, further evidence-based guidance tivity (09:00 to 12:30 h on the sports morning) was calcu-
is needed to help children and their families use CGM effec- lated: percentage time in range (TIR), 3.9–10.0 mmol/L, and
tively to facilitate exercise. in hyperglycemia >10 mmol/L.
An observational study in 25 adolescents using a CHO in- The number of hypoglycemic events and the number of
take algorithm in response to CGM trend arrows showed low participants experiencing events are presented for each arm.
rates of hypoglycemia, but did not have a control arm to the Medians and interquartile ranges are presented for percentage
study.17 The aim of our pilot study was to investigate whether time spent in different SGL ranges for each arm; paired sign
the use of a CHO intake algorithm based on rate of change of tests were used to compare medians across arms. Event rates
real-time CGM data could impact on hypoglycemia events or with exact Poisson 95% confidence intervals (CIs) were
exercise interruptions and to produce estimates to inform the calculated for CHO intake events, exercise interruption, and
design and power calculation for further research. mild hypoglycemic events (<3.9 mmol/L for >20 min) in
Downloaded by Ghana Hinari Band 1 from at 01/20/19. For personal use only.

each arm; mixed-effects Poisson models were used to com-

pare rates across arms.
All statistical analyses were performed using Stata/SE for
Children diagnosed with T1D for >1 year, ages from 8 to Windows, version 13.0 (StataCorp LP, College Station, TX).
12 years, were recruited into the study. The inclusion criteria All tests are two sided, and p values <0.05 were considered
were any insulin regimen and the latest glycated hemoglobin statistically significant.
(HbA1c) below 10.0% (86 mmol/mol).
The protocol was approved by the institutional ethics
committee. Results
The study used a two-condition, randomized crossover
design. Randomization to the sequence of intervention and The 14 children (5 boys, 9 girls) who completed the study
control was computer generated ( had a mean (–standard deviation [SD]) age of 10.5 (–1.4)
Participants came together for a camp-style, sports morn- years, mean diabetes duration of 4.8 (–2.7) years, and mean
ing of fun-based activity organized by a professional sporting HbA1c of 7.7 (–0.6) %. The majority of the participants were
organization, on two occasions, 2 weeks apart. Each sports using an insulin pump (n = 9, 64%). Mean total daily insulin
morning was sports/play based with a mix of aerobic and dose was 0.72 (–0.16) U/(kg$d) and the children performed a
skills-based activities consisting of two 90-min exercise mean of 4.4 (–2.0) hours of exercise per week. None of the
blocks with a 30-min break in between. study participants had self-reported impaired hypoglycemia
Participants were fitted with the Dexcom G5 Mobile awareness. One participant had a history of severe hypogly-
system (Dexcom, Inc., San Diego, CA) at least 24 h before cemia (hypoglycemic seizure) 4 months before the first sports
the sports mornings. The Dexcom G5 Mobile CGM system day morning. The hypoglycemia event rate (95% CI) in the
allows transmission of SGLs via Bluetooth to a mobile device 24 h before the sports mornings (SGL <3.9 mmmol/L) was
that generates alerts. This information can be shared via comparable in both groups: 1.36 events per person/day (0.79–
‘‘cloud’’ with up to five individuals, who are then able to 2.18) in the control and 1.62 (0.99–2.47) events per per-
remotely monitor the CGM reading of the user in real time son/day in the intervention group.
along with the possibility to use individualized alerts. On the Table 1 shows the mean SGL (–SD) at the start of exercise,
sports morning, participants consumed a standardized the median percentage time spent in different SGL ranges,
breakfast and were advised to reduce the breakfast bolus by and the rates of mild hypoglycemic events, exercise inter-
10%. Insulin pump users reduced their basal insulin delivery ruption events, and CHO intake events in the control and
by 20% and patients on intermediate insulin reduced their intervention arm.
intermediate acting insulin by 10%. During the physical ac- No hypoglycemic events with SGL <3.0 mmol/L for
tivity on the sports morning, participants adhered to either a ‡20 min occurred in either arm. Mild hypoglycemic events
CHO intake algorithm based on CGM trends (intervention) occurred on two occasions during the control arm and on five
modified from Riddell and Milliken,17 shown in Figure 1, or occasions (three participants with one episode and one par-
to a control protocol where blood glucose levels (BGL) were ticipant with two episodes) during the intervention arm.
checked every 30 min and CHO was given according to standard The distribution for percentage TIR was negatively
exercise guidelines (0.5 g CHO/kg/h given as 0.25 g CHO/kg skewed and the distribution for percentage time high was
every 30 min when the capillary glucose was <8 mmol/L).7,18 positively skewed; both had values at both upper (100%) and
CHO was administered in the form of glucose tablets for both lower bounds (0%). Median percentage TIR and median time
groups. spent high were comparable across arms and there were no
During the intervention arm, participants were able to see statistically significant differences (Table 1).
their SGL in real time and study personnel were following the Mean CHO intake during exercise was the same in both
participants remotely. In the control arm, participants and arms, 0.3 – 0.2 g/kg/h. However, compared with the control
study personnel were blinded to the CGM. arm, the intervention algorithm was associated with fewer

Directional arrows or :
< 6.1- 6.9 mmol/L Take 8g CHO

Directional arrows :
Take 24g CHO
Sensor Glucose
< 7mmol/L < 5-6 mmol/L
during exercise and
Directional arrows :
Take 20g CHO

< 5 mmol/L Take 24g CHO

Downloaded by Ghana Hinari Band 1 from at 01/20/19. For personal use only.

FIG. 1. CHO intake algorithm, modified from Riddell and Milliken.17 CHO, carbohydrate.

CHO intake events and fewer exercise interruption events, as The CHO intake algorithm used in this pilot study was
illustrated in Table 1. modified from Riddell and Milliken.17 Modifications to
Action according to the algorithm was taken 12 times Riddell’s algorithm encompassed giving larger amounts of
during the intervention arm. Five algorithm actions were CHO at specific thresholds, as illustrated in Figure 1. Despite
followed by mild hypoglycemia in a time frame of 60 min increased CHO dosing, use of the modified algorithm did not
after algorithm action. None of the algorithm actions resulted result in hyperglycemia (>10 mmol/L) within 60 min of CHO
in a glucose level >10 mmol/L in a time frame of 60 min after ingestion. We found that the total amount of CHO ingested
algorithm action. during exercise was similar in both groups. However, par-
ticipants using the CHO intake algorithm ingested larger
amounts of CHO less frequently than the control group. It
remains to be determined if this pattern of CHO consumption
This pilot study shows that both use of a CHO intake al- would have been feasible and effective in the control group.
gorithm based on CGM trends (intervention) and adhering to The strength of this study is that it uses a randomized
standard exercise CHO intake guidelines (control) (0.5 g crossover design to minimize interindividual confounding
CHO/kg/h) have the potential to minimize hypoglycemia factors that may influence glycemia during exercise, such as
during a physical activity in children with T1D, as demon- fitness levels and nutritional status. In addition, exercise type
strated by no episodes of significant hypoglycemia (SGL and duration, insulin adjustment, and consumption of breakfast
<3.0 mmol/L for ‡20 min) in both study arms. Moreover, the and snacks were kept constant between study days. However,
rate of mild hypoglycemic events (SGL <3.9 mmol/L for the study has several limitations. First, given that the risk of
‡20 min) was low in both arms. Similarly, Riddell and Mil- hypoglycemia during exercise is known to be higher with
liken17 found that CGM use coupled with a CHO intake al- sustained moderate/intensity exercise compared with inter-
gorithm largely maintained euglycemia during exercise. mittent high-intensity exercise,10 it is important to note that our
Furthermore, this study demonstrates that use of CGM sports mornings comprised mixed activities, and consequently,
with a CHO intake algorithm is associated with a reduction in this may have attenuated the risk of hypoglycemia. However,
the number of interruptions to exercise for ‘‘events’’ such as as type and duration of activity were standardized, the risk of
performing SMBG or stopping to ingest CHO intake, with a hypoglycemia was constant across both groups. Furthermore,
similar time spent in defined glucose ranges compared with the program of activity performed was chosen to be relevant to
using standard exercise CHO intake guidelines. These find- the types of fun-based activity children in this age group may
ings are relevant to active young people with T1D who may partake in, such as a school sports event. Second, it has to be
find it difficult or undesirable to stop and ‘‘test’’ during ex- noted that the mean SGL at the start of the exercise was in the
ercise. Moreover, performing SMBG may not be practical hyperglycemic range, which also lowered the risk of hypo-
during specific activities such as cycling, swimming, or ski- glycemia during exercise. It may have also been useful to have
ing. In particular, in very cold temperatures, blood glucose performed this study without reducing basal insulin, to simu-
meters may not work. late spontaneous exercise that is common in children—and
It should be noted that this study was a pilot study to generate may have been a stronger challenge to both the control and
estimates of potential outcome measures and assesses the fea- intervention algorithm.
sibility of the study design to inform a larger trial and was not In summary, both CHO intake based on standard exercise
powered to detect differences between the two arms. We gained CHO intake guidelines (control) and use of a CHO intake
insight into the distributions of potential outcome measures and algorithm based on CGM trends (intervention) have the po-
their variability that will inform the design of future trials. tential to minimize hypoglycemia, and largely maintained

derived from paired Sign tests. Events are expressed as rate of events per person per exercise morning with exact Poisson 95% CI; IRR with 95% CI comparing intervention to control and P-values are
SGL at exercise start is expressed as mean (mmol/L) and SD; the P-value is derived from a paired t-test. Percentage time spent in different SGL ranges is expressed as medians and IQR; P-values are
euglycemia during exercise in young children with T1D. In


addition, use of a CHO intake algorithm resulted in fewer

CHO intake events and fewer interruptions to exercise. Use
of this algorithm may reduce the burden of diabetes man-
agement during exercise with the potential to facilitate ac-

[0.55 to 15.78]
tivity in young people with T1D.

[0.22 to 0.65]
[0.10 to 0.37]
[-3.6 to 2.6]
95% CI

95% CI

95% CI
We thank the patients and families who participated in the
study, and Sports Challenge Australia who provided the ex-
ercise program during the sports mornings. This study was
Table 1. Percentage Time Spent in Different Sensor Glucose Level Ranges and Event Rates

supported by a seeding grant from the Children’s Diabetes

Mean difference

CHO, carbohydrate; CI, confidence interval; IQR, interquartile ranges; IRR, incidence rate ratios; SD, standard deviation; SGL, sensor glucose level.
Centre in Perth, a JDRF/NHMRC funded Centre of Research
Excellence. M.A.B. was funded by a research fellowship of




the Ettore and Valeria Rossi Foundation, Switzerland, and

supported by an International Postgraduate Research Fel-
lowship from the University of Western Australia. Dexcom
provided the sensors and hardware through an unrestricted
Downloaded by Ghana Hinari Band 1 from at 01/20/19. For personal use only.

[0.12 to 0.83]

[33.3 to 86.0]
[2.3 to 66.7]

[0.4 to 1.4]
[0.8 to 2.1]

Authors’ Contributions

95% CI

95% CI


M.A.B. and T.C. were responsible for the study design,

data collection, data analysis, and article preparation. G.S.
was responsible for data analysis and reviewed and edited the
article. P.A. contributed to the study design and reviewed and



edited the article. M.D.B. contributed to the study design,




data collection, and reviewed and edited the article. T.W.J.

and E.A.D. contributed to the study design and reviewed and
edited the article. All authors approved the final version of
this article.
[0.02 to 0.52]

[48.7 to 85.7]
[11.9 to 42.9]

[1.6 to 3.3]
[5.9 to 8.8]

95% CI

95% CI

Author Disclosure Statement


No competing financial interests exist.






1. Brazeau AS, Rabasa-Lhoret R, Strychar I, Mircescu H:

Barriers to physical activity among patients with type 1
diabetes. Diabetes Care 2008;31:2108–2109.
2. MacMillan F, Kirk A, Mutrie N, et al.: A systematic review
Mild hypoglycemic events (SGL <3.9 mmol/L for <20 min)

of physical activity and sedentary behavior intervention

studies in youth with type 1 diabetes: study characteristics,
intervention design, and efficacy. Pediatr Diabetes 2014;15:
derived from the mixed models. *P < 0.05 significant.

3. Salem MA, Aboelasrar MA, Elbarbary NS, et al.: Is exercise
a therapeutic tool for improvement of cardiovascular risk
factors in adolescents with type 1 diabetes mellitus? A ran-
domised controlled trial. Diabetol Metab Syndr 2010;2:47.
4. Seeger JP, Thijssen DH, Noordam K, et al.: Exercise
training improves physical fitness and vascular function in
SGL at exercise start (mmol/L)

children with type 1 diabetes. Diabetes Obes Metab 2011;

% time SGL 3.9–10 mmol/L

Exercise interruption events

% time SGL >10 mmol/L

5. Dunton GF, Huh J, Leventhal AM, et al.: Momentary as-

sessment of affect, physical feeling states, and physical
activity in children. Health Psychol 2014;33:255–263.
CHO intake events

6. Robertson K, Adolfsson P, Scheiner G, et al.: Exercise in

children and adolescents with diabetes. Pediatr Diabetes
2009;10(Suppl. 12):154–168.

7. Riddell MC, Gallen IW, Smart CE, et al.: Exercise man-

agement in type 1 diabetes: a consensus statement. Lancet
Diabetes Endocrinol 2017;5:377–390.

8. Rabasa-Lhoret R, Bourque J, Ducros F, Chiasson JL: 14. Al Khalifah RA, Suppere C, Haidar A, et al.: Association of
Guidelines for premeal insulin dose reduction for post- aerobic fitness level with exercise-induced hypoglycaemia
prandial exercise of different intensities and durations in in type 1 diabetes. Diabet Med 2016;33:1686–1690.
type 1 diabetic subjects treated intensively with a basal- 15. Yardley JE, Sigal RJ, Kenny GP, et al.: Point accuracy of
bolus insulin regimen (ultralente-lispro). Diabetes Care interstitial continuous glucose monitoring during exercise
2001;24:625–630. in type 1 diabetes. Diabetes Technol Ther 2013;15:46–49.
9. Francescato MP, Geat M, Fusi S, et al.: Carbohydrate re- 16. Bally L, Zueger T, Pasi N, et al.: Accuracy of continuous
quirement and insulin concentration during moderate ex- glucose monitoring during differing exercise conditions.
ercise in type 1 diabetic patients. Metabolism 2004;53: Diabetes Res Clin Pract 2016;112:1–5.
1126–1130. 17. Riddell MC, Milliken J: Preventing exercise-induced hy-
10. Shetty V, Fournier P, Davey R, et al.: Effect of exercise poglycemia in type 1 diabetes using real-time continuous
intensity and blood glucose level on glucose requirements glucose monitoring and a new carbohydrate intake algo-
to maintain stable glycaemia during exercise in individuals rithm: an observational field study. Diabetes Technol Ther
with type 1 diabetes. Int J Pediatr Endocrinol 2015;2015 2011;13:819–825.
(Suppl. 1):O39. 18. Robertson K, Riddell MC, Guinhouya BC, et al.: ISPAD
11. Bally L, Zueger T, Buehler T, et al.: Metabolic and hor- Clinical Practice Consensus Guidelines 2014. Exercise in
monal response to intermittent high-intensity and continu- children and adolescents with diabetes. Pediatr Diabetes
ous moderate intensity exercise in individuals with type 1 2014;15(Suppl. 20):203–223.
diabetes: a randomised crossover study. Diabetologia 2016;
Downloaded by Ghana Hinari Band 1 from at 01/20/19. For personal use only.

59:776–784. Address correspondence to:

12. Garcia-Garcia F, Kumareswaran K, Hovorka R, Hernando Dr Marie-Anne Burckhardt, MD
ME: Quantifying the acute changes in glucose with exer- Department of Endocrinology and Diabetes
cise in type 1 diabetes: a systematic review and meta- Perth Children’s Hospital
analysis. Sports Med 2015;45:587–599. 15 Hospital Avenue
13. Burke LM, Hawley JA, Angus DJ, et al.: Adaptations to Nedlands, Perth 6009
short-term high-fat diet persist during exercise despite high Western Australia
carbohydrate availability. Med Sci Sports Exerc 2002;34:
83–91. E-mail: