1. Background
Regular physical activity and exercise enhance glycaemic control in people with type 1 diabetes (T1D) and reduce the risk of diabetes‐related complications [1]. Exercise improves metabolic risk factors and muscle insulin sensitivity, leading to reduced insulin requirements [2], partly through increased skeletal muscle glucose uptake via insulin‐independent pathways involving GLUT1 [3]. Despite these benefits, fear of hypoglycaemia remains a major barrier to regular physical activity, as exercise‐induced and delayed hypoglycaemic events are common and difficult to predict [4].
In real‐world settings, physical activity and exercise are often estimated using step counts as a proxy for activity levels [5, 6]. However, step counts provide limited information on metabolic demand, and the relationship between energy expenditure and glucose dynamics in T1D remains insufficiently characterized, particularly over the long term.
Large‐scale datasets enable a more detailed assessment of this relationship. While prior studies such as the T1DEXI study [6] linked step count with glycaemic outcomes, energy expenditure reflects overall metabolic demand more comprehensively than step count alone [7]. When combined with CGM‐derived metrics, it allows direct investigation of the association between physical activity and glucose levels [8].
When combined with CGM‐derived metrics, it allows direct investigation of the relationship between physical activity and glucose levels. Therefore, the aim of this analysis was to investigate the long‐term impact of physical activity and exercise, quantified as weekly energy expenditure (kcal/week), on CGM‐derived metrics in a physically active cohort with T1D over a 6‐month period.
2. Methods
This study is a retrospective analysis of anonymised data collected between 2018 and 2024, voluntarily donated for scientific purposes by individuals with T1D using the Diabetes Cockpit application (Syntactiq, Austria). The application integrates health information from Apple Health [9], Nightscout [10] and/or applications for continuous glucose monitoring (CGM). This information includes step count, exercise minutes, calories burned, as well as CGM‐derived metrics such as mean glucose with standard deviation, time in range (TIR; 70–180 mg/dL), carbohydrate intake, and both basal and bolus insulin doses. The Diabetes Cockpit application does not display glucose values directly but summarizes them in various metrics and overviews. The sources for these metrics are the direct applications of the sensor systems and additional health metrics, offering clear visualizations and personalized insights to support optimizing diabetes management [11].
Eligible participants were adults (≥ 18 years) with T1D who used the Diabetes Cockpit application (Syntactiq, Austria) and provided consent for scientific use of their data.
The data was organized into non‐overlapping six‐month blocks. Only complete calendar months from the same user (“user‐months”) were included. A user‐month was eligible if it contained at least 21 days with ≥ 200 glucose readings per day and 21 days with recorded energy expenditure; overlap between these days was not required. This ensured at least 70% data coverage. The blocks were generated using a sliding window (i.e., every calendar month between 2018 and 2024 was checked as a possible starting month), whereby a block was only valid if all 6 consecutive months met the coverage criteria. For each user, the block with the highest data coverage was selected. In total, a maximum of 413 non‐overlapping blocks would have been possible for all users. For each participant, a “best single block” with the highest overall coverage was identified.
For detailed activity analyses, a stricter duplicate‐filtering procedure was applied, as some activities had been recorded more than once (e.g., when two applications were used simultaneously). After removing duplicates, workouts were categorized as strength, cardio, or other (e.g., yoga).
Energy expenditure was extracted from participants' movement data and defined as the estimated energy used above the resting metabolic rate. Participants were stratified into quartiles of weekly energy expenditure (< 2206, 2206–3056, 3057–4165, > 4165 kcal/week) to compare activity levels and explore dose–response relationships. To mitigate the impact of potential measurement errors or data gaps, all energy expenditure data were included in the analysis, as no standardized thresholds were predefined in our inclusion criteria.
The study protocol and all associated regulatory aspects were approved by the Ethics Committee of the University of Bayreuth, Germany (approval number: 25‐036). The study was registered in the German Register of Clinical Trials (DRKS00039550).
The primary endpoint was the comparison of TIR across four groups defined by weekly energy expenditure (kcal/week) over 6 months. Secondary endpoints included additional CGM‐derived metrics: time in tight range (TITR; 70–140 mg/dL), time below range (TBR1: 54–70 mg/dL; TBR2: < 54 mg/dL), time above range (TAR1: 180–250 mg/dL; TAR2: > 250 mg/dL), coefficient of variation (CV), mean glucose, and total weekly energy expenditure. Correlations were assessed using Spearman's rank correlation. Group comparisons were performed using one‐way ANOVA or Kruskal–Wallis tests with appropriate post hoc analyses. Exploratory linear regression analyses were additionally performed to quantify changes in TBR1 and TBR 2 per 1000 kcal/week increase in energy expenditure. Analyses were conducted in GraphPad Prism (v8.0.2); p < 0.05 was considered significant.
3. Results
The baseline characteristics, insulin administration devices, and countries of residence of the participants are presented in Table 1. General physical activity and 13 476 exercise sessions were recorded. The application of the duplicate filter resulted in 11 355 tracked activities, consisting of 1273 strength–, 8857 cardio‐sessions, and 1225 sessions assigned to the other category. However, these results should be interpreted with caution, as several participants did not track their activity, leading to incomplete data. Among the 179 participants, activity tracking was observed for 35 individuals in the strength and other categories and for 15 individuals in the cardio category.
TABLE 1.
Baseline characteristics.
| Characteristic | Q1 | Q2 | Q3 | Q4 | All |
|---|---|---|---|---|---|
| Total, n (%) | 45 (25.14) | 45 (25.14) | 45 (25.14) | 44 (24.58) | 179 (100) |
| Age, years, median (IQR) | 46 (31–59) | 41 (34–54) | 43 (36–51) | 42 (37–53) | 42 (35–53) |
| Female, n (%) | 22 (12.29) | 17 (9.50) | 18 (10.06) | 6 (3.35) | 63 (35.20) |
| Exercise sessions, n | 70 | 341 | 1952 | 8992 | 11 355 |
| Country of origin | |||||
| Europe, n (%) | 16 (35.56) | 19 (21.35) | 23 (25.84) | 31 (70.45) | 89 (49.7) |
| North America, n (%) | 27 (60.0) | 22 (48.89) | 18 (40.0) | 13 (29.55) | 80 (44.7) |
| Asia, n (%) | 1 (2.22) | 1 (2.22) | 3 (6.67) | 0 (0.0) | 5 (2.8) |
| Africa, n (%) | 1 (2.22) | 2 (4.44) | 0 (0.0) | 0 (0.0) | 3 (1.7) |
| Oceania, n (%) | 0 (0.0) | 1 (2.22) | 1 (2.22) | 0 (0.0) | 2 (1.1) |
| Insulin application | |||||
| AID/CSII, n (%) | 16 (35.56) | 24 (53.33) | 25 (55.56) | 22 (50.0) | 87 (49) |
| MDI, n (%) | 7 (15.56) | 5 (11.11) | 8 (17.78) | 7 (15.91) | 27 (15) |
| No specification, n (%) | 22 (48.89) | 16 (35.56) | 12 (26.67) | 15 (34.09) | 65 (36) |
Note: Baseline characteristics of the participants by quartiles. Values are presented as n (% within quartile) or as medians with interquartile ranges (IQR). Percentages refer to column totals unless stated otherwise. Exercise sessions (all) are presented after duplicate filter.
Abbreviations: AID, automated insulin delivery; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; PA, physical activity.
CGM‐derived metrics of the cohort investigated were (median [IQR]): TIR (70–180 mg/dL) 77.9% [66.7; 87.2] with a TITR (70–140 mg/dL) of 53.9% [39.2; 64.6], TBR1 (54–70 mg/dL) of 1.5% [0.6; 3.1] and a TBR2 of 0.2% [0.0; 0.6]. TAR1 (180–250 mg/dL) was 15.9% [8.4; 23.3] and TAR2 (> 250 mg/dL) at 2.4% [0.4; 6.4]. The mean glucose level was 141 mg/dL [128; 159] with a standard deviation of 45 mg/dL [36; 55] and CV of 31% [28; 36] (see Table S1). The overall median [IQR] energy expenditure was 3056 kcal/week [2205; 4165]. Defining the four groups based on the regular energy expenditure of the participants yielded the following median [IQR] values: Q1: 1589 kcal/week [887; 2011], Q2: 2592 kcal/week [2400; 2846], Q3: 3432 kcal/week [3166; 3700] and Q4: 5044 kcal/week [4561; 5498] (see Figure S1).
When comparing CGM‐derived metrics across the four physical activity groups, significant differences were found only for TBR2 (< 54 mg/dL) (p = 0.03). Post hoc analyses showed that TBR2 was significantly higher in the highest energy expenditure group (Q4) compared with the lowest energy expenditure group (Q1), with median values of 0.43% (IQR 0.11–0.78) versus 0.12% (IQR 0.00–0.37), respectively (p = 0.03). No significant differences were observed for the remaining CGM‐derived metrics (p > 0.05). Correlation analyses between overall kcal/week and CGM‐derived metrics showed significant correlations for TBR2 (r = 0.23, 95% CI: 0.08–0.37; p = 0.002), TBR1 (54–70 mg/dL) (r = 0.20, 95% CI: 0.05–0.34; p = 0.008) and CV (r = 0.19, 95% CI: 0.04–0.33; p = 0.009; see Figure 1). Exploratory linear regression analyses showed that each increase of 1000 kcal/week in energy expenditure was associated with a non‐significant 0.067 percentage‐point increase in TBR1 (95% confidence interval (CI) −0.006 to 0.14, p = 0.073), while no association was observed for TBR2 (slope −0.027 percentage points, 95% CI −0.21 to 0.16, p = 0.771). For CV, the regression analyses demonstrated a small but statistically significant association between weekly energy expenditure and glycemic variability. The regression coefficient was 0.67 (95% CI 0.09 to 1.30, p = 0.024), indicating an increase of 0.67 percentage points in CV per additional 1000 kcal/week. However, the explained variance was low (R 2 = 0.029).
FIGURE 1.
Correlation of the glucose metrics and energy expenditure per week. (A) Time in range (70–180 mg/dL), (B) time in tight range (70–140 mg/dL), (C) time below range 1 (54–70 mg/dL), (D) time below range 2 (< 54 mg/dL), (E) time above range 1 (180–250 mg/dL), (F) time above range 2 (> 250 mg/dL), (G) median glucose, (H) coefficient of variation. Symbols represent median. Correlation was conducted using Spearman's rank correlation; significant differences are indicated in the figure.
4. Discussion
In this study, it was found that within the observed cohort with T1D, a physically active lifestyle was associated with an increased risk of clinically relevant hypoglycemia. The Q4 group with the highest energy expenditure showed more time spent below range 2 (< 54 mg/dL) compared to the other groups. There are several possible reasons why hypoglycemia can occur more frequently during physical activity and exercise. Moderate‐intensity exercise typically stabilizes blood glucose in T1D via enhanced insulin‐independent uptake [12 ], whereas high‐intensity activity may first raise glucose (catecholamine release) before a subsequent drop [13]. It would have been important to find out whether the individuals in the study population were engaged in everyday and known or in unknown sporting activities, as this would have made an acute effect of exercise on insulin sensitivity more likely.
The correlation between overall kcal/week and the CGM‐derived metrics for TBR1 and TBR2 supports the assumption that there is a small but consistent association between higher energy expenditure and an increased risk of hypoglycemia.
The exploratory regression analyses further support the findings of the correlation analyses, suggesting that higher weekly energy expenditure was not associated with clinically meaningful increases in hypoglycemia risk. While a small positive trend was observed for TBR1, the association did not reach statistical significance, and no relationship was found for TBR2. A statistically significant association was observed between energy expenditure and glycemic variability; however, the effect size was small, with an increase of less than 1 percentage point in CV per additional 1000 kcal/week and a very low explained variance (R 2 = 0.029).
More frequent or more intense exercise physiologically elevates the likelihood of hypoglycemia, a finding that is well aligned with the existing literature [14], particularly when carbohydrate intake and insulin dose adjustments are insufficient. The observed effect of glucose variability, as reflected by the CV, was also significant over the 6‐month period. A statistically significant association was observed between energy expenditure and glycemic variability; however, the effect size was small, with an increase of less than 1 percentage point in CV per additional 1000 kcal/week and a very low explained variance (R 2 = 0.029). Overall, these findings suggest that higher physical activity levels may only have a minor impact on glycemic variability and hypoglycemia exposure in this cohort. Higher levels of physical activity were associated with greater glucose variability, indicating more short‐term fluctuations despite overall good diabetes management. This aligns with previous findings suggesting that increased CV is linked to a higher risk of hypoglycemia [15]. Although the absolute hypoglycemia values were statistically significant between groups, with TBR1 at 1.5% and TBR2 at 0.2% in total, they remain moderate from a clinical perspective. Nonetheless, it is important to emphasize that TBR2 events, glucose levels below 54 mg/dL, are considered safety‐critical due to the potential for neurological impairment [16] and should be avoided whenever possible.
As this is a retrospective observation conducted in a real‐world setting, the use of the app provides valuable insight into the participants' daily conditions, conferring a high degree of ecological validity. The cohort under consideration and the observational period provide sufficient statistical power to detect significant differences in CGM‐derived metrics. Despite these strengths, the study has several limitations. The accuracy of the recorded movement data cannot be fully ascertained. Previous findings indicate that smartphone‐based applications exhibit lower accuracy compared to wrist‐worn smartwatches [17]. Similarly, with regard to data accuracy, it was not specified which CGM devices were used by the participants. The accuracy of different sensor systems varies to some extent, which may have influenced the results [18]. The retrospective design precludes causal inferences, as numerous potential confounders, such as insulin dosing, dietary intake, timing of exercise and exercise‐related adaptations, were not controlled for. In addition, self‐selection bias is likely, as many application users may be technologically proficient and possess greater knowledge of diabetes management. Finally, the type and intensity of physical activity and exercise were not systematically documented, which is relevant because different exercise modalities can interfere in various ways with the glucose (e.g., endurance exercise reduces glucose levels, whereas strength training tends to maintain more stable glycaemia responses [19]). The Diabetes Cockpit app lacked intensity metrics, limiting quantification in our dataset. Future studies should include objective intensity data to clarify its impact. Also limiting is the reliance on potentially inaccurate app‐based energy data and the absence of systematic documentation on activity type and intensity. Another limitation is the lack of carbohydrate intake data or insulin doses administered from the participants. With data of this kind, our current analysis would be significantly more meaningful.
5. Conclusion
This retrospective real‐world analysis confirms that physical activity and exercise are generally safe for adults with T1D, maintaining stable glycemic control over time. While higher energy expenditure slightly increased hypoglycemia and variability, absolute risks remained low. These results emphasize individualized management to safely incorporate exercise. App‐based real‐world data provide valuable insights into activity, exercise, and glucose control, supporting digital tools for personalized diabetes care.
Author Contributions
S.S., J.S., L.S. and O.M. were involved in the conception and design of the study/analysis. S.S., J.S., L.S. and O.M. were involved in the conduct of the study/analysis. K.B., S.S., L.S. and O.M. were involved in the analysis and interpretation of the results. K.B. and S.S. wrote the first draft of the manuscript, and all authors edited, reviewed, and approved the final version of the manuscript. O.M. is the guarantor of this work and, as such, had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding
The authors have nothing to report.
Conflicts of Interest
K.B. has received research support from Medtronic and Sinocare. U.L.‐S. has received Conference travel support from Boehringer, Sanofi, Novartis, Amgen, Lilly, Novo Nordisk AT, Fresenius, Astra Zeneca, Daiichi Sankyo. O.M. has received clinical trial support from Sêr Cymru II COFUND Fellowship/European Union, Novo Nordisk A/S, Novo Nordisk AT, Abbot Diabetes Care, Sanofi, Dexcom, Team Novo Nordisk, SAIL, Maisels Brauerei, Medtronic AT, EFSD/EASD, Falke, BISp, Perfood, Ypsomed, and Sinocare; received presenters' honoraria form Medtronic AT, Medtronic Int., Eli Lilly, Novo Nordisk, Sanofi, TAD Pharma, ADA, Diatec, Berufsverband deutscher Internist*innen, Dexcom, Insulet, AstraZeneca, and Ypsomed; received conference travel support from Novo Nordisk A/S, Novo Nordisk AT, Novo Nordisk UK, Medtronic AT, Sanofi, EASD, OEDG, and DDG, and is a member in the Advisory Board of Sanofi, TAD Pharma, Dexcom, perfood, Medtronic. S.S. has received a doctoral scholarship from the German National Academic Foundation (Studienstiftung des deutschen Volkes) as well as conference and conference travel support from the German Diabetes Association, the European Association for the Study of Diabetes, and mylife Diabetescare AG. Research support was received by mylife Diabetescare AG, Medtronic, and Sinocare. L.S. is Co‐founder and CEO of Syntactiq Dynamics FlexCo, Vienna, Austria. H.S. has received personal fees (lecture fees, advisory boards) from Amgen, Amarin, Bayer, Boehringer Ingelheim, Daiichi Sankyo, Eli Lilly, Novartis and NovoNordisk, Sanofi and Sobi. M.T. has received personal fees (lecture fees, advisory boards) from Abbott, Eli Lilly, Medtronic, Sanofi, and Ypsomed, and support for conference attendance from Eli Lilly and KelCon GmbH. The other authors declare no conflicts of interest.
Supporting information
Table S1: Comparison of CGM‐derived glucose metrics as median and IQR across physical activity quartiles.
Figure S1: Glucose Metrics and Energy Expenditure Across Groups
Acknowledgements
We would like to express our gratitude to all users of the Diabetes Cockpit application who voluntarily provided their data. Open Access funding provided by Universitat Graz.
Bünzel K., Sanfilippo S., Schierbauer J., et al., “Long‐Term Associations Between Energy Expenditure and CGM‐Derived Metrics in Adults With Type 1 Diabetes: A Retrospective Analysis of the Syntactiq Cockpit Database,” Diabetes, Obesity and Metabolism 28, no. 9 (2026): 8714–8718, 10.1111/dom.71003.
Handling Editor: Richard Donnelly
References
- 1. Galassetti P. and Riddell M. C., “Exercise and Type 1 Diabetes (T1DM),” Comprehensive Physiology 3, no. 3 (2013): 1309–1336, 10.1002/CPHY.C110040. [DOI] [PubMed] [Google Scholar]
- 2. Colberg S. R., Sigal R. J., Yardley J. E., et al., “Physical Activity/Exercise and Diabetes: A Position Statement of the American Diabetes Association,” Diabetes Care 39, no. 11 (2016): 2065–2079, 10.2337/DC16-1728. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Barnard R. J. and Youngren J. F., “Regulation of Glucose Transport in Skeletal Muscle 1,” FASEB Journal 6, no. 14 (1992): 3238–3244, 10.1096/FASEBJ.6.14.1426762. [DOI] [PubMed] [Google Scholar]
- 4. Ergin B. D., Ergin I., and Gökşen D., “Fear of Hypoglycemia and Longer Disease Duration Associated With Physical Activity Avoidance in Children and Adolescents With Type 1 Diabetes Mellitus,” Journal of Clinical Research in Pediatric Endocrinology 15, no. 3 (2023): 238–247, 10.4274/JCRPE.GALENOS.2023.2022-9-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Bünzel K., Sanfilippo S., Moser O., et al., “Head‐To‐Head Comparison of Two Automated Insulin Delivery Systems in Children With Type 1 Diabetes During a 2‐Week Summer Camp: An Exploratory Prospective Study,” Hormone Research in Pædiatrics Epub ahead of print. (2025): 1–9, 10.1159/000548290. [DOI] [PubMed] [Google Scholar]
- 6. Turner L. V., Marak M. C., Gal R. L., et al., “Associations Between Daily Step Count Classifications and Continuous Glucose Monitoring Metrics in Adults With Type 1 Diabetes: Analysis of the Type 1 Diabetes Exercise Initiative (T1DEXI) Cohort,” Diabetologia 67, no. 6 (2024): 1009–1022, 10.1007/S00125-024-06127-2. [DOI] [PubMed] [Google Scholar]
- 7. Chevance G., Golaszewski N. M., Tipton E., et al., “Accuracy and Precision of Energy Expenditure, Heart Rate, and Steps Measured by Combined‐Sensing Fitbits Against Reference Measures: Systematic Review and Meta‐Analysis,” JMIR mHealth and uHealth 10, no. 4 (2022): e35626, 10.2196/35626. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Fernández‐Verdejo R., Sanchez‐Delgado G., and Ravussin E., “Energy Expenditure in Humans: Principles, Methods, and Changes Throughout the Life Course,” Annual Review of Nutrition 156 (2025): 52, 10.1146/annurev-nutr-062122. [DOI] [PubMed] [Google Scholar]
- 9. Apple Inc ., Apple Health ( App Store , 2024). [Google Scholar]
- 10. “Welcome to Nightscout — Nightscout Documentation,” accessed December 11, 2025, https://nightscout.github.io/.
- 11. “Syntactiq.ai ‐ Advancing Diabetes Research Through Data,” accessed November 19, 2025, https://syntactiq.ai/.
- 12. Riddell M. C., Gallen I. W., Smart C. E., et al., “Exercise Management in Type 1 Diabetes: A Consensus Statement,” Lancet Diabetes and Endocrinology 5, no. 5 (2017): 377–390, 10.1016/S2213-8587(17)30014-1. [DOI] [PubMed] [Google Scholar]
- 13. Yardley J., Mollard R., MacIntosh A., et al., “Vigorous Intensity Exercise for Glycemic Control in Patients With Type 1 Diabetes,” Canadian Journal of Diabetes 37, no. 6 (2013): 427–432, 10.1016/j.jcjd.2013.08.269. [DOI] [PubMed] [Google Scholar]
- 14. Codella R., Terruzzi I., and Luzi L., “Why Should People With Type 1 Diabetes Exercise Regularly?,” Acta Diabetologica 54, no. 7 (2017): 615–630, 10.1007/S00592-017-0978-X. [DOI] [PubMed] [Google Scholar]
- 15. Kilpatrick E. S., Rigby A. S., Goode K., and Atkin S. L., “Relating Mean Blood Glucose and Glucose Variability to the Risk of Multiple Episodes of Hypoglycaemia in Type 1 Diabetes,” Diabetologia 50, no. 12 (2007): 2553–2561, 10.1007/S00125-007-0820-Z. [DOI] [PubMed] [Google Scholar]
- 16. Glycemic Goals and Hypoglycemia: Standards of Care in Diabetes‐2025,” Diabetic Care 48, no. 1 (2025): S128–S145, 10.2337/dc25-S006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Hong K. R., Hwang I. W., Kim H. J., Yang S. H., and Lee J. M., “Apple Watch 6 vs. Galaxy Watch 4: A Validity Study of Step‐Count Estimation in Daily Activities,” Sensors 24, no. 14 (2024): 4658, 10.3390/S24144658. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Freckmann G., Wehrstedt S., Eichenlaub M., et al., “A Comparative Analysis of Glycemic Metrics Derived From Three Continuous Glucose Monitoring Systems,” Diabetes Care 48, no. 7 (2025): 1213–1217, 10.2337/dc25-0129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Särnblad S., Ponsot E., Leprêtre P. M., and Kadi F., “Acute Effects of Aerobic Continuous, Intermittent, and Resistance Exercise on Glycemia in Adolescents Males With Type 1 Diabetes,” Pediatric Diabetes 22, no. 4 (2021): 610–617, 10.1111/PEDI.13194. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Comparison of CGM‐derived glucose metrics as median and IQR across physical activity quartiles.
Figure S1: Glucose Metrics and Energy Expenditure Across Groups