Title: Concurrent validity of a continuous glucose monitoring system at rest, during and
following a high-intensity interval training session.
Submission type: Original research
Authors: Clavel P1,2, Tiollier E2, Leduc C1,3, Fabre M1,2, Lacome M1,2, Buchheit M2,4,5,6
1 Performance Department, Paris Saint-Germain FC, Saint-Germain-en-Laye, France
2 French National Institute of Sport (INSEP), Laboratory of Sport, Expertise and Performance
(EA 7370), Paris, France
3 Carnegie Applied Rugby Research (CARR) centre, Institute for Sport, Physical Activity and
Leisure, Carnegie School of Sport, Leeds Beckett University, Leeds, United Kingdom
4 HIITScience, Revelstoke, BC, Canada
5 Institute for Health and Sport, Victoria University, Melbourne, VIC, Australia
6 Kitman Labs, Performance Research Intelligence Initiative, Dublin, Ireland
Paris Saint-Germain FC, Saint-Germain-en-Laye, France
Mobile: +33 6 09 42 78 33
Running head: Glucose monitoring in athletes
Abstract count: 217
Text only word account: 1900
Numbers of tables: 3
Numbers of figures: 3
Conflicts of interest: The authors do not have any conflict of interest.
Purpose: To assess the concurrent validity of a continuous blood glucose monitoring system
(CGM) Post-Breakfast, Pre-exercise, Exercise and Post-exercise, while assessing the impact of
two different breakfasts on the observed level of validity. Methods: Eight non-diabetic
recreational athletes (age: 30.8±9.5 years; height: 173.6±6.6 cm; body mass: 70.3±8.1 kg) took
part in the study. Blood glucose concentration was monitored every 10 min using both a CGM
(FreeStyle Libre, Abbott, France) and finger-prick blood glucose measurements (FreeStyle
Optimum, Abbott, France) over 4 different periods (Post-Breakfast, Pre-Exercise, Exercise and
Post-Exercise). Two different breakfasts (carbohydrates- [CHO] and protein- [PROT] oriented)
over two days (2x2 days in total) were used. Statistical analyses included the Bland-Altman
method, standardized mean bias (expressed in standardized unit), median absolute relative
difference (MARD) and the Clarke Error Grid (EGA). Results: Overall, mean bias was trivial-
to-small at Post-Breakfast (effect size ± 90% confidence limits: -0.12±0.08), Pre-Exercise (-
0.08±0.08) and Post-Exercise (0.25±0.14), while moderate during Exercise (0.66±0.09). Higher
MARD was observed during Exercise (13.6% vs 7 to 9.5% for the other conditions). While
there was no effect of the breakfast type on the MARD results, EGA revealed higher value in
Zone D (i.e. clinically unsafe zone) during Exercise for CHO (10.5%) compared with PROT
(1.6%). Conclusion: The CGM device examined in this study can only be validly used at rest,
after both a CHO and PROT-rich breakfast. Using CGM to monitor blood glucose concentration
during exercise is not recommended. Moreover, the accuracy decreased when carbohydrates
are consumed before exercise.
Regulation of blood glucose has first been widely studied from a health perspective.
Hyperglycemia for example, is believed to be an independent risk factor for the development
of several diseases such as type II diabetes mellitus1 and cardiovascular disease.2 More recently,
the monitoring of blood glucose concentration has also elicited great interest in sport, as
hypoglycaemia influences both physical and cognitive performances.3
In particular, it is known that at the beginning of exercise or after half-time in team sports,
athletes experience transient hypoglycemia, which may affect physical and cognitive
performance.4 Moreover, it has then been shown that a large glycemic variability exists among
individuals in the general population.5 Additionally, similar results have been shown in sub-
elite athletes,6 suggesting that providing more individualized guidelines to regulate blood
glucose would be beneficial for both health and performance goals.
The emergence of new technologies such as continuous glucose monitoring (CGM) devices has
allowed blood glucose concentration dynamics to be captured more frequently and less
invasively than traditional measures such as finger pricks. Indeed, as CGM devices only need
to be placed once (usually on the back of the arm), it can be used for several days without
disturbing sport practices. So far, these devices have been mainly used by diabetic populations
but as the technology becomes more accurate, less invasive, and less expensive, their use has
increased in other populations and especially in healthy individuals. Therefore, the inclusion of
CGM in sport nutritionists’ monitoring tool box could help to optimize nutritional strategies
before and during exercise, and in turn, improve athletes’ performance by preventing
hypoglycemia. However, to date, the validity of these new systems at rest or during exercise
has been only assessed in diabetics patients and showed promising results.7 Evidence regarding
its relevance with an athletic population is still lacking. Moreover, the ability of such devices
to detect potential glucose fluctuations due to different nutritional intakes need to be confirmed.
Therefore, the aim of this study was to assess the concurrent validity of a new CGM device
during different periods, i.e. pre, during and after exercise, while assessing the potential impact
of different nutritional intakes in the observed level of validity.
Eight non-diabetic recreational athletes (5 females, 3 males) (age: 30.8 ± 9.5 years; height:
173.6 ± 6.6 cm; body mass: 70.3 ± 8.1 kg) who regularly participate in running and resistance-
based training (8±2 hours per week) were included in the study. An a priori power analysis was
conducted using the package pwr from R software (Version 4.0.0) for t-tests for non-parametric
data with a significance level alpha of 0.05 a power of 0.8 and add a non-parametric correction
of 15%. Result showed a minimal sample of 310 paired observations for 8 participants were
necessary. Alcohol intake was prohibited during the study period. Regarding female
participants, we ensured they were all within the same menstrual phase during the study period.
Participants provided informed consent prior to starting the study. Ethics approval was granted
before any data collection wwas undertaken and the recommendations of the Declaration of
Helsinki were respected.
A concurrent validity design was employed to assess the validity of a CGM system against
finger prick measures which was considered as the reference method. Over 2 consecutive
weeks, participants took part in 4 nonconsecutive standardized days. Each standardized day was
broken-down into 4 distinct periods: 1) Post-Breakfast which corresponded to the first hour
after the end of the Breakfast 2) Pre-Exercise which corresponded to the first hour following
the Post-Breakfast, 3) Exercise, which started 2 hours after the end of the breakfast and lasted
from the beginning of the warm up to the end of the workout and 4) Post-exercise, which started
immediately at the end of the workout, and up to 30 min later. A detailed outline of the
standardized day structure is provided in Figure 1. Nutritional intake during breakfast was
manipulated in order to provide either a high carbohydrate (CHO) or protein (PROT) breakfast,
to induce different levels of resting ad pre-exercise glycemia. Each typical breakfast was
repeated twice. Over those standardized days, blood glucose was measured continuously with
a CGM, while finger prick measures were taken every 10 minutes and. Day 1 was used for each
participant to familiarize with the CGM and ensure calibration (as per manufacturer
recommendations) before the experimentation could start. Between day 2 and 13, participants
undertook at their convenience the 4 standardized days. They were also instructed to have at
least one full day of recovery between each experimental day.
**Insert Figure 1**
Continuous glucose monitoring. Each participant was provided with a CGM system (FreeStyle
Libre, Abbott, France) over the full duration of the study. Each participant inserted a sensor
(FreeStyle Libre, Abbott, France) in their non-dominant upper arm (i.e. back the triceps
brachialis) one day before the beginning of the study. Glucose concentration was recorded in
the interstitial fluid every minute.
Finger prick blood glucose. Finger prick (FreeStyle Optium, Abbott, France) measures were
collected following the procedure described by Gomez.8 Each sample was immediately
analysed using the FreeStyle Libre reader (FreeStyle Libre Reader, Abbott, France) (The
validity and reliability of this device has been previously confirmed.9
Breakfast. Two typical breakfasts were employed. The CHO breakfast contained a high
proportion of carbohydrates (CHO) with 1 g.Kg-1 of body mass with a ceiling set at 70g of
carbohydrates per breakfast (e.g. breakfast contained a mix of orange juice, bread and jam).The
macronutrients and energy were as follow: 65±7g of carbohydrates, 9±1g of proteins and 1±0g
of fat for a total of 311±31 Kcal. The protein (PROT) breakfast was isoenergetic compared with
CHO (e.g. breakfast contained a mix of eggs, ham and cheese). The macronutrients and energy
were as follow: 1±0g of carbohydrates, 30±0g of proteins and 23±0g of fat for a total of 311±31
Standardized exercise. Participants completed the 30-15 Intermittent Fitness Test (30-15IFT) as
described by Buchheit et al.10 prior the beginning of the study. The speed (km·hr-1) achieved
by each participant during the last successfully completed stage of the test was recorded (VIFT)
in order to prescribe exercise intensity. The standardized exercise started with a 10-min low-
intensity run (30 to 40% of VIFT) and was followed by a high-intensity intermittent training
exercise performed outdoor. The trials consisted of six reps of 3-min running intervals
interspersed with 2 min of passive recovery. Reps 1 and 2 were performed at 75% VIFT, reps 3
and 4 at 80% VIFT and reps 5 and 6 at 85% VIFT. The session was ended with a 10-min walk.
Data processing. Each time point within a specific period was averaged as described above to
perform the concurrent validity analysis for each method (CGM and finger prick) and per
specific period (Figure 1). Each standardized day was analyzed first without (overall) and then
as a function of breakfast type (CHO and PROT).
Bland-Altmann method for repeated measures and standardized mean bias were first applied to
assess the agreement between CGM and finger prick measures at each specific period.11 The
following thresholds were applied to rate the magnitude of the bias as follow: >0.2 (small), >0.6
(moderate), >1.2 (large) and >2 (very large).12
Additionally, analysis of the median average relative difference (MARD)13 and the Clarke Error
Grid Analysis (EGA)14 were conducted. Regarding MARD, further comparisons between the
different periods were performed using Wilcoxon test and/or Kruskal-Wallis tests. Level of
statistical significance was set at P<0.05. Results were further analyzed while calculating
standardized differences, i.e. Wilcoxon effect sizes. The thresholds to rate the magnitude of
the effects were the same than those used for mean bias. Regarding EGA, results were divided
into 5 zones (A, B, C, D, E). Each zone denotes a degree of clinical implications of blood
glucose concentration measures. Zones A and B were considered clinically acceptable while
zone C, D and E (erroneous treatment) were deemed possibly unsafe.14
The Bland-Altman analysis for the 4 periods is presented in Figure 2 and reported as mean bias
(standard error). Irrespectively of the breakfast content, mean biases were trivial-to-small for
Post-Breakfast (-2.99 [17.75] mg/dL), Pre-Exercise (-1.67 [10.95] mg/dL), Post-Exercise (4.18
[17.88] mg/dL) and moderate during Exercise (12.25 [13.86] mg/dL). Regarding CHO
breakfast, mean biases were trivial-to-small for Post-Breakfast (-1.43 [25.98] mg/dL), Pre-
Exercise (-4.29 [11.66] mg/dL), Post-Exercise (3.32 [18.18] mg/dL) and moderate during
Exercise (14.06 [13.81] mg/dL). For PROT Breakfast, trivial mean bias was observed for Pre-
Exercise (0.91 [8.98] mg/dL), Post-Breakfast (-4.51 [8.31] mg/dL) and Post-Exercise (5.13
[15.98] mg/dL), while moderate mean biases were observed for Exercise (10.47 [13.19]
**Insert Figure 2**
**Insert Figure 3**
The results of the MARD analysis between the different periods are presented in Table 1 and
**Insert Table 1 and 2**
Results regarding EGA are presented in Table 3. Irrespectively of the breakfast content, Post-
Breakfast, Pre-Exercise, and Post-Exercise periods fell into Zone A (accurate) and B (benign
errors) (100%). However, during Exercise, 94% of the values fell into A (70.4%) and B
(23.6%), and 6% in Zone D (failure to treat errors). For CHO breakfast, 10.5% of data fell into
Zone D for Exercise, while the other periods fell into Zone A and B. Similarly, for PROT
breakfast, 1.6% fell into Zone D during the Exercise period.
**Insert Table 3**
The aim of this study was 1) to investigate the concurrent validity of a new CGM device in
recreational athletes at Post-Breakfast, Pre-exercise, Exercise and Post-exercise, and 2) to
assess the potential impact of either a CHO-rich or protein-rich breakfast on the observed level
of validity. The main results highlighted that, while the validity of CGM was acceptable at rest
(i.e. Post-Breakfast, Pre-Exercise and Post-Exercise), it was lower during Exercise and
especially after the CHO breakfast.
The first results demonstrated trivial-to-small mean bias during all the non-exercise periods,
irrespectively of nutritional intake. Moreover, all results from EGA fell into the “clinically safe
zone” (A and B), albeit during Exercise. These results are similar to those shown previously in
non-athletic diabetic populations.15 Indeed, the present results suggest that assessing glucose
dynamics at rest is feasible with this CGM device. This could open the door to a better
individualization of nutritional strategies.5
Yet, we observed a higher bias during Exercise compared with the other periods, confirming
previous studies in a non-athletic diabetic population.16 Reasons that may contribute to the
reduced validity of the CGM device in this context include microcirculation perturbations as a
as a result of movements around or within the insertion area, increases in body temperature and
rapid fluxes in glucose levels during exercise.17 Regarding the likely physiological time lag of
glucose transport between blood and interstitial fluid compartments (see Figure 3, finger pricks
measures changed faster Post-Breakfast than that of the CGM device), it should be noted that
it might not have accounted for the observed difference in accuracy as the pattern is not only
delayed but it varies with time and conditions. Indeed, while a clear hypoglycemia was observed
with finger prick measures immediately at the start of exercise (which was the expected
physiological response), the CGM showed an increased blood glucose response (Figure 2).
Nonetheless, this discrepancy indicates that the CGM device was unable to detect a potential
hypoglycemia observed at the onset of exercise, and could therefore not be used to assess
strategies aiming at preventing this phenomenon in practice. It is worth mentioning that a trend
for a better agreement was observed toward the end of the exercise periods (Figure 2). If the
duration of the exercise also affects the accuracy of CGM, it means that while the device may
not be suitable for sport including short and intermittent exercise durations, its use could
perhaps be considered during longer event such as cycling, trail or triathlon. This potential
better accuracy toward longer exercise duration highlights the need to conduct further research
involving 1) longer exercise duration, 2) nutritional intake during long endurance race 3)
various exercise modalities and 4) different intensities.
To examine the potential effect of the absolute levels of glycemia on the validity of the CGM
device, different breakfasts were proposed (CHO and PRO). Similar MARD and EGA results
were observed, suggesting that the CGM validity was not affected by the breakfast content
during non-exercise periods (i.e. Post-Breakfast, Pre-Exercise, Post-Exercise). Specific pre-
competition nutritional strategies can have a positive influence on both the acute running
performance among rugby league players18 or endurance athletes,19 and the chronic training
adaptations to training.20 Consequently, the use of this CGM device could be considered by
practitioners willing to monitor glycemic responses before and after competition or training, to
ensure the efficacy of the nutritional strategies employed.
However, during the Exercise period, the CGM accuracy was modulated by breakfast content.
Indeed, a 10 times higher value in Zone D of the EGA (i.e. clinically unsafe) was observed post
CHO (10.5%) compared with post PROT (1.6%) breakfast. In our study, zone D corresponds
to the situation where finger prick measures indicate an hypoglycemic state whereas CGM
measures are within the normal range14 suggesting that CGM failed to detect the hypoglycemia
occurring during exercise after the CHO-rich breakfast. It is well known there is a rapid drop
of blood glucose concentration at the onset of exercise, due to an increased glucose uptake by
exercising muscles.21 This physiological mechanism could explain why the sensor lacks
sensitivity to rapid changes in glucose concentration, as observed in the present study. As it
stands, if practitioners want to monitor blood glucose during high-intensity intermittent
exercise, they need to consider other devices than CGM (e.g. finger prick).
- The present CGM system provided valid measures at rest. Therefore, the use of such a
system may allow for a better individualization of nutritional strategies before or after
- The level of validity was lower during high-intensity intermittent training and was in
addition influenced by the type of breakfast consumed (i.e. high carbohydrates or high
protein). Consequently, practitioners should avoid using this device during intermittent
Daily monitoring of blood glucose is of importance in athletes given the likely impact of
glycemia on performance and the individualized nutritional recommendations that can be made
with CGM. Our results highlighted that the CGM device examined in the present study
presented only trivial-to-small bias when compared with a traditional fingerpick device at rest,
suggesting that it could be used confidently during this specific period. The CGM device is not
valid enough to monitor glucose during intermittent exercise. Further analyses should however
evaluate the validity of this device over longer exercise duration.
1. Cavalot F, Pagliarino A, Valle M, et al. Postprandial blood glucose predicts
cardiovascular events and all-cause mortality in type 2 diabetes in a 14-year follow-up:
Lessons from the San Luigi Gonzaga diabetes study. Diabetes Care.
2. Gallwitz B. Implications of postprandial glucose and weight control in people with type
2 diabetes: understanding and implementing the International Diabetes Federation
guidelines. Diabetes Care. 2009;32 Suppl 2(suppl 2):S322-S325. doi:10.2337/dc09-
3. Harper LD, Briggs MA, McNamee G, et al. Physiological and performance effects of
carbohydrate gels consumed prior to the extra-time period of prolonged simulated
soccer match-play. J Sci Med Sport. 2016;19(6):509-514.
4. Kingsley M, Penas-Ruiz C, Terry C, Russell M. Effects of carbohydrate-hydration
strategies on glucose metabolism, sprint performance and hydration during a soccer
match simulation in recreational players. J Sci Med Sport. 2014;17(2):239-243.
5. Hall H, Perelman D, Breschi A, et al. Glucotypes reveal new patterns of glucose
dysregulation. PLoS Biol. 2018;16(7):1-23. doi:10.1371/journal.pbio.2005143
6. Thomas F, Pretty CG, Desaive T, Chase JG. Blood Glucose Levels of Subelite Athletes
during 6 Days of Free Living. J Diabetes Sci Technol. 2016;10(6):1335-1343.
7. Greene J, Louis J, Korostynska O, Mason A. State-of-the-art methods for skeletal
muscle glycogen analysis in athletes-the need for novel non-invasive techniques.
Biosensors. 2017;7(1):1-16. doi:10.3390/bios7010011
8. Gómez AM, Umpierrez GE, Muñoz OM, et al. Continuous Glucose Monitoring Versus
Capillary Point-of-Care Testing for Inpatient Glycemic Control in Type 2 Diabetes
Patients Hospitalized in the General Ward and Treated with a Basal Bolus Insulin
Regimen. J Diabetes Sci Technol. 2016;10(2):325-329.
9. Rodrigo EP, Deib-Morgan K, Diego OG de, García-Velasco P, Sgaramella GA,
González IG. Accuracy and reliability between glucose meters: A study under normal
clinical practice conditions. Semer - Med Fam. 2017;43(1):20-27.
10. Buchheit M. The 30-15 Intermittent Fitness Test: Accuracy for Individualizing Interval
Training of Young Intermittent Sport Players. J Strength Cond Res. 2008;22(2):365-
11. Altman D, Bland J. sensitivity.pdf. BMJ. 1994;308:1552.
12. Hopkins WG. Measures of Reliability in Sports Medicine and Science. Sport Med.
13. Reiterer F, Polterauer P, Schoemaker M, et al. Significance and Reliability of MARD
for the Accuracy of CGM Systems. J Diabetes Sci Technol. 2017;11(1):59-67.
14. Clarke WL, Cox D, Gonder-Frederick LA, Carter W, Pohl SL. Evaluating clinical
accuracy of systems for self-monitoring of blood glucose. Diabetes Care.
15. Freckmann G, Pleus S, Link M, Zschornack E, Kloẗzer HM, Haug C. Performance
evaluation of three continuous glucose monitoring systems: Comparison of six sensors
per subject in parallel. J Diabetes Sci Technol. 2013;7(4):842-853.
16. Biagi L, Bertachi A, Quirós C, et al. Accuracy of continuous glucose monitoring
before, during, and after aerobic and anaerobic exercise in patients with type 1 diabetes
mellitus. Biosensors. 2018;8(1):1-8. doi:10.3390/bios8010022
17. Kumareswaran K, Elleri D, Allen JM, et al. Accuracy of continuous glucose
monitoring during exercise in type 1 diabetes pregnancy. Diabetes Technol Ther.
18. Bradley WJ, Morehen JC, Haigh J, et al. Muscle glycogen utilisation during Rugby
match play: Effects of pre-game carbohydrate. J Sci Med Sport. 2016;19(12):1033-
19. Rothschild JA, Kilding AE, Plews DJ. What should i eat before exercise? Pre-exercise
nutrition and the response to endurance exercise: Current prospective and future
directions. Nutrients. 2020;12(11):1-23. doi:10.3390/nu12113473
20. Jeukendrup AE. Periodized Nutrition for Athletes. Sport Med. 2017;47(s1):51-63.
21. Richter EA, Hargreaves M. Exercise, GLUT4, and skeletal muscle glucose uptake.
Physiol Rev. 2013;93(3):993-1017. doi:10.1152/physrev.00038.2012
Table and figure caption
Figure 1. Schematic representation of the study design.
Figure 2. Bland-Altman analysis between the continuous glucose monitoring device (CGM)
and finger prick measures (FPBG). Dash lines represent the limits of agreements.
Figure 3. Continuous glucose monitoring (CGM) and finger prick measures during each
standardized condition, when ingesting a carbohydrate- (upper) and protein- (lower) oriented
breakfasts, with the 2 days of each breakfast condition pooled for each participant (n = 2 x 8
for each curve). Data are presented as mean (SE).
Table 1. Median Absolute Relative Difference between the continuous glucose monitoring
device (CGM) and finger prick measures. Data are median (interquartile range) and expressed
in percentage. *: significantly different from Post-Breakfast. #: significantly different from Pre-
Exercise. †: significantly different from Exercise. Comparisons between period are presented
as effect size with 90% confidence interval.
Table 2. Comparisons between period are presented as effect size for Wilcoxon test with 90%
Table 3. Clark Error Grid Analysis between the continuous glucose monitoring device (CGM)
and finger prick measures. Zone A represents a clinically accurate measure. Zone B stands for
benign errors. Zone C represents overcorrection errors. Zone D and E represent failure to treat
errors and erroneous treatment errors respectively. For more details see Clarke et al. (1987).
(0.17 to 0.31)
(0.01 to 0.16)
(0.24 to 0.38)
(0.07 to 0.24)
(0.06 to 0.23)
(0.13 to 0.34)
(0.01 to 0.18)
(0.27 to 0.46)
(0.07 to 0.31)
(0.07 to 0.28)
(0.14 to 0.34)
(0.01 to 0.2)
(0.16 to 0.36)
(0.01 to 0.24)
(0.02 to 0.24)
D (Failure to
D (Failure to
D (Failure to