Article Test

Introduction
Diabetes is one of the most challenging health problems worldwide.1 People with diabetes require continuous treatment, including dietary guidelines; however, this is not always sufficient to stabilize their blood glucose levels. A high degree of variability in blood glucose levels, along with hyper- and hypoglycemic events, pose significant health risks.2 It is essential to manage these factors as it may help prevent or delay the microvascular and macrovascular complications associated with diabetes 3-7 Blood glucose stability leads to fewer hospitalizations and, therefore, substantial reductions in diabetes management costs, while increased glycemic variability is associated with increased healthcare expenses.8-10

Regular measurement of blood glucose levels became commonplace in the second half of the 20th century. The first glucose meters for home use were marketed in 1981 and began to actively replace urine and in-office glucose checks, which were neither accurate nor prompt.11,12 Frequent blood glucose level checks provide patients with the knowledge to optimize insulin dosing. Over the years, counting carbohydrates before meals and glucose meter usage have become the standard methods of diabetes self-care.3,12 However, blood glucose monitoring and carbohydrate counting have serious limitations.3,12-15 These approaches have two major disadvantages: patients are unable to accurately count carbohydrates in the food consumed 16,17; and blood glucose meters provide information on blood glucose levels at only that particular point of time and do not indicate the direction or rate of change in blood glucose levels.18,19 These can lead to erroneous conclusions, poor diabetes management, decreased self-care capacity, and, ultimately, diabetes distress and burnout.12,20

Furthermore, carbohydrate counting and multiple finger-stick blood check measurements throughout the day are not always accurate to predict glycemic variability.21 Identical meals can cause variability in postprandial glucose levels in different people.22 Protein and fat content also affect postprandial glucose levels. However, since the impact differs from person to person, there is no standardized approach to measure fat and protein content.23 Therefore, patients need to evaluate how their bodies react to different meals.

The development and use of continuous glucose monitoring (CGM) partially solves the problems associated with glucose meters.24,25 The development and use of CGM partially solves the problems associated with glucose meters.19 These CGM readings are wirelessly sent to the users smartphone and then aggregated in a data storage. Apple Health application, Dexcom cloud, and Nightscout cloud are examples of such storage applications. Apple Health is a health data storage system pre-installed on iPhones. It gathers data on blood pressure measurements, glucose levels, and physical activity. The Dexcom cloud service aggregates data from produced sensors. Nightscout is an open-source web application that accumulates diabetes management data from all connected devices, including CGM sensors.

Glycemic stability increases with continuous and intermittent CGM use.26 However, the exact causes of glycemic fluctuations may remain vague for individuals with diabetes.18,19,27 With CGM, patients can observe fluctuations in glucose levels as a result of their lifestyle decisions. To determine the exact causes of fluctuation in blood glucose levels, people with diabetes must recollect data pertaining to their meals, exercise, insulin injection, and other lifestyle factors such as stress levels or sleep duration; this can prove to be a challenging task.12

In addition to CGM technology developments, time in range (TIR) has been introduced as a new and important clinical metric for diabetes management and is used to monitor glucose variability.5,28-31 The International Consensus defines TIR as blood glucose levels of 70–180 mg/dL (3.9–10 mmol/L), with a goal of 70% of readings in a day within the target range for most people with type 1 or type 2 diabetes, and 50% for those at a higher risk of hypoglycemia.19,28,31 TIR is a clinical tool that, alongside A1C, is included in the American Diabetes Association’s guidelines and allows clinicians to monitor patients’ glucose level variability throughout the day.3 Regular TIR monitoring and correlated lifestyle adjustments lead to a reduction in diabetes complication risks and mortality.3-5,32,33

In addition to TIR, time below range (TBR) and time above range (TAR) can be calculated from CGM data. These metrics characterize hyperglycemic and hypoglycemic episodes, respectively.19 An increase in TIR and a concomitant decrease in TBR can be considered as evidence of effective and safe glucose management.29
CGM-based applications can be used to track time within the target range.34,35 These software tools allow TIR analysis in addition to keeping a food diary, thereby enabling people with diabetes to evaluate their responses to the food they consume and make better decisions about their future meals and insulin timing or doses. There is growing evidence of improvement in patient engagement, self-management, and glycemic stability among individuals with diabetes using mobile applications.36-41

We hypothesized that an application that enables the patient to analyze the effects of routine events, such as meals, physical activity, and insulin dosing, on glucose levels with CGM, would reduce glycemic variability; this may lead to an increase in TIR together with a decrease in TBR and TAR more considerably than a CGM-only approach. To test this hypothesis, we assessed the impact of such application usage in individuals with type 1 and type 2 diabetes mellitus.

Materials and methods
Study design and selection criteria
The ‘Undermyfork’ application was chosen for this investigation. This free application is distributed via the Apple App Store and Google Play. It combines meal photos and insulin data with glucose readings from CGM. It allows users to comprehend which meal-related decisions drive them out of the target glucose range and which ones allow them to remain within the target range. As a result of using the application, a library of events such as meals and corresponding bolus insulin dosing data are created. By automatically connecting this library to glucose data, individuals with diabetes can see the relationship between logged events and changes in glucose values. The application receives CGM readings automatically from the Apple Health, Nightscout, and Dexcom clouds.

With the application, people with diabetes take photos of each meal and optionally adds keywords (tags) with the meal description and the meal context data (for example, time of the day or preceding physical activity). The Insights section of the app allows users to review meal photos classified by two-hours postprandial TIR, analyze how different lifestyle strategies (e.g., meal choices, insulin dosing, and timing) affect postprandial TIR after similar foods, and identify meals associated with high and low glycemic variability. Based on this section, people with diabetes can evaluate their responses to the food choices and make more informed decisions about future meals, insulin timing, and dosing.

For the retrospective cohort study, data were collected from the applications cloud database. The database contains records of individual glucose levels, photos of meals, and insulin dose information. Informed consent was obtained when study participants signed up to use the application for the first time. The user data were anonymized, and each user was under a depersonalized ID number.

We defined the start point of application use as the day when the first meal photo was added. The eligibility criteria were type 1 or type 2 diabetes diagnoses. Data from 155 patients who used the application while wearing CGM devices were investigated. All these 155 patients, taken from a cohort of 1,365 users with type 1 and type 2 diabetes, met the following selection criteria. They had to have at least 70% complete CGM data in three 14-day periods: 1) 14 days prior to the use of the mobile application (before period); 2) 14 days after the first use of the mobile application (after period); 3) 14 days prior to the before period (control period). All CGM readings were placed into a single framework, with the average glucose level during the five-minute periods as the basic data point. Missing data in CGM readings were filled via interpolation if the missed period was less than 60 minutes.42

The TIR, TAR, and TBR were calculated for each user and each period. TIR was defined as the quotient of the division of the number of CGM readings in the 70–180 mg/dL (or 3.9–10 mmol/L) range to the total number of readings. TAR and TBR were defined as the quotient of the division of the number of CGM readings more than 180 mg/dL (or 10 mmol/L) and less than 70 mg/dL (or 3.9 mmol/L), respectively, to the total number of readings. A TIR of over 70% approximates an A1C of less than 7%, which is associated with a significant reduction in long-term complications of diabetes 3,30,33 We selected users with a TIR of less than 70% in the before period to investigate changes in this parameter only for the users at risk of delayed complications.

Active use of the application was defined as at least five days (35%) with meal uploads for 14 days with a start point on the first day of application usage. Thus, specific selection criteria were as follows: 1) diagnosis of type 1 or 2 diabetes mellitus; 2) 70% CGM readings in each of the periods under review; 3) at least five days with photos uploaded to the application at the after period; and 4) TIR in before period less than 70%. The overall number of selected users was 21 (Figure 1). The resulting TIR, TAR, and TBR for each eligible user in the cohorts are presented in Table 1 and Figure 2.

Procedures
We chose three time-based cohorts. A 28-day period prior to the first meal photo added to the application was divided into two groups: 1) before cohort, two weeks preceding the start point (days from -14 to -1); 2) control cohort, two weeks preceding the before cohort (days from -28 to -15). An after cohort included glucose level data from the first to the 14th day of application use.

Data Analysis
The data were gathered for 16.5 months, from October 1, 2020, to March 19, 2022, and processed using Spyder (the Scientific Python Development Environment), a free and open-source distribution of the Python programming language.43 We used the Pandas library for data analysis and SciPy and scikit-posthocs packages to compute statistics.
There were three paired periods; thus, the significance of the differences between the TIRs of each user during all three periods (continuous measures) was assessed using Friedmans two-way analysis of variance by ranks with pairwise comparison.44 We conducted a pairwise comparison to examine the mean differences within the periods using the Nemenyi test. Statistical significance was accepted at an adjusted p-value < .05.

Results
The study population included 21 accounts. Ninety-one percent of patients (n = 19) had type 1 diabetes mellitus and 9% (n = 2) had type 2 diabetes mellitus. Table 2 presents the postprandial blood glucose level characteristics of the users in the cohorts.

The mean ± standard deviation (SD) TIRs during control, before, and after periods were 47.5% ± 18.1, 46.2% ± 18.3, and 57.3% ± 19.0), respectively. The mean ± SD TARs during the same periods were 49.5% ± 20.1, 50.5% ± 20.5, and 38.5% ± 21.3, respectively. The corresponding mean ± SD TBRs were 3.0% ± 3.7, 3.2% ± 4.1, and 4.2% ± 6.2, respectively. The number of days with meal photo uploads had a mean of 10, and the mean ± SD of meal photos uploaded per day was 1.8 ± 1.0. The mean ± SD differences between the before and after cohort groups were 11.1% ± 14.0 in TIR and -12.0% ± 14.9 in TAR. The mean percentage of CGM data available after interpolation was 90.0% ± 6.8, 97.0% ± 2.3, and 96.8% ± 3.9 in the control, before, and after periods, respectively. The effects of mobile-based interventions on glycemic parameters are presented in Table 3.

Time in range
A significant difference between the three measurements was observed while comparing the change of the TIRs within each period (related- samples Friedman’s test χr2  = 15.4, p-value = .0004). After performing the Nemenyi pairwise test for multiple comparisons of mean rank sums, no significant difference was observed between the control and before periods (p-value = .526). On the contrary, when the before and control periods were compared with the after period, significant differences were observed in both comparisons (p- value = .001; p-value = .019, respectively).

Time above range
The TAR evolution was also tested using Friedman’s χr2 criterion. A significant difference was found between the periods (χr2 = 16.1, p = .0003). The Nemenyi test did not reveal a significant difference between the control and before periods (p = .348). However, when comparing pairs before and after and control and after periods, significant differences were observed (p-value = .001; p- value = .029, respectively).

Time below range
No statistical difference was observed within the groups in TBR evolution (Friedman’s test χr2 = 2.2, p = .333).

DISCUSSION
This retrospective cohort study aimed to test the hypothesis of an association between photo- based food diary mobile application usage and an increase in TIR along with a corresponding decrease in TAR and TBR in patients with type 1 and type 2 diabetes mellitus. There was a statistically and clinically significant increase in time in range with a corresponding reduction in time above range. In addition, there was no statistical difference in glycemic parameters between the control and before periods. Thus, the effect discovered is suggested to be determined by the application usage, not by CGM-related improvement in blood glucose levels.

The mobile app provided users with a review of their postprandial blood glucose level excursions in response to different meals. With this information, they were able to grasp their personal reactions to the food consumed and, consequently, adjust their subsequent meal choices. These results show that an application that provides a better understanding of glucose level fluctuations after different meal- related decisions may help patients increase their TIR. Such comprehension in people living with diabetes may increase their involvement in the decision-making process, leading to positive behavioral changes, including development of lifelong healthy eating habits. These improvements result in a decrease in glycemic variability with the potential to reduce the risk of diabetes complications.

Previous studies have shown that fluctuations in blood glucose levels after the same meal may vary considerably from person to person.22 Since there is no standardized way of eating for diabetes 9, self- education and decision-making skills are essential factors in reducing glycemic variability. Evidence suggests that the level of self-care is higher among diabetes application users.41,45 Food diaries and CGM data trackers have been shown to be the most relevant features of these applications.45,46 Such approaches provide individuals with diabetes with practical tools for nutrition self-care, which is an essential part of diabetes management.47

Recent studies have suggested that, in addition to overall average blood glucose levels, the duration of hyperglycemic episodes also affects the initiation and progression of diabetes complications.48 A decrease in TAR is shown to reduce the risk of cardiovascular diseases and neuropathic complications.47-49 Lifestyle changes associated with an increase in glycemic stability can improve the quality of life of people with diabetes. There was no correlation between application use and the time the participants spent below the target blood glucose range, although the time below range was relatively low in all groups. This is also consistent with prior literature, which states that TIR has a strong correlation with TAR but is not correlated with TBR.50,51

It is important to note that this study had several limitations. First, since all the data were anonymized, we did not know the sex, age, race, duration of diabetes, or continuance of CGM use for the participants. Second, this study only included people with smartphones who are proficient in mobile applications and, simultaneously, were wearing CGMs. This indicates that the research participants may not be representative of people with diabetes in low-income socioeconomic groups. Third, there is no information on the long-term effects of the application usage. Finally, since this was a pilot study, another limitation was the small sample size which included mostly people with type 1 diabetes. We plan to conduct further research on the effects of this mobile application on TIR, TAR, and TBR in a broader population of people with type 1 and type 2 diabetes. An additional future research direction should be a collaboration with a healthcare team to review the user data and give feedback on lifestyle decisions and medication adjustments. It is important to test our hypothesis for a longer period of time, up to six months, with a possible follow-up in a year, to check the sustainability of changes in patients outcomes. To further improve the quality of the investigation, an increase in sample size is required, along with an independent control group of patients who have not received the intervention.

Looking ahead, as more and more digital therapeutics tools also in the field of diabetes become available on the market, the research aiming to reveal their synergies and effectiveness requires collaboration between healthcare professionals and the scientific community. More studies are needed to compare various applications between each other and in combination with other tools. The results of these studies will be recommendations that healthcare professionals can use to track their patients, make treatment decisions, and also improve the people with diabetes outcomes. Not only should time in range but also A1C be among the primary clinical endpoints for future research. Clinics that keep patients under observation regularly already collect data on A1C. The next step is to overlay this data with efficacy findings from different diabetes management applications to evaluate what is the best for the patient.

CONCLUSION
This study demonstrated how a mobile application that includes a digital collection of meal photos combined with postprandial continuous glucose monitoring data led to an increase in time in range and reduced time above range. This is likely related to changes in eating and insulin dosing in response to the insights from the data and corresponding report. These results are important as increased time in range is associated with reduction in microvascular complications of diabetes and improved quality of life. Further research is required to confirm the results of this study. The long-term outcomes of partial substitution of carbohydrate counting with the visualization of individual retrospective postprandial continuous glucose monitoring data should be the subject of additional research.

References
1.    Saeedi P, Petersohn I, Salpea P, et al. Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from    the    International    Diabetes    Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract. 2019;157:107843. doi:10.1016/j.diabres.2019.107843
2.    Monnier L, Colette C, Wojtusciszyn A, et al. Toward defining the threshold between low and high glucose variability in diabetes. Diabetes Care. 2017;40(7):832-838. doi:10.2337/dc16-1769
3.    Glycemic Targets: Standards of Medical Care in Diabetes-2020. Diabetes Care. 2020;43(Suppl 1):S66-S76. doi:10.2337/dc20-S006
4.    Schade DS, Lorenzi GM, Braffett BH, et al. Hearing impairment and type 1 diabetes in the diabetes control and complications trial/epidemiology of diabetes interventions and complications (DCCT/EDIC) cohort. Diabetes Care. 2018;41(12):2495-2501. doi:10.2337/dc18- 0625
5.    Albernaz PL. Hearing loss, dizziness, and carbohydrate Metabolism. Int Arch Otorhinolaryngol.    2016;20(3):261-270. doi:10.1055/s-0035-1558450
6.    Gregg EW, Sattar N, Ali MK. The changing face of diabetes complications. Lancet Diabetes Endocrinol.    2016;4(6):537-547. doi:10.1016/S2213-8587(16)30010-9
7.    Fasil A, Biadgo B, Abebe M. Glycemic control and    diabetes    complications    among    diabetes mellitus patients attending at University of Gondar Hospital, Northwest Ethiopia. Diabetes Metab Syndr Obes. 2019;12:75-83. doi:10.2147/DMSO.S185614
8.    Hirsch JD, Morello CM. Economic impact of and treatment options for type 2 diabetes. Am J Manag Care. 2017;23(13 Suppl):S231-S240.
9.    Foundations of Care and Comprehensive Medical Evaluation. Diabetes Care. 2016;39 Suppl 1:S23- 35. doi:10.2337/dc16-S006
10.    American Diabetes Association. Economic costs of diabetes in the U.S. in 2017. Diabetes Care. 2018;41(5):917-928. doi:10.2337/dci18-0007
11.    Clarke SF, Foster JR. A history of blood glucose meters and their role in self-monitoring of diabetes mellitus. Br J Biomed Sci. 2012;69(2):83-93. doi:10.1080/09674845.2012.12002443
12.    Weinstock RS, Aleppo G, Bailey TS, et al. The role of blood glucose monitoring in diabetes management. Arlington (VA): American Diabetes Association; 2020 Oct. doi: 10.2337/db2020-31
13.    Gunst J, De Bruyn A, Van den Berghe G. Glucose control in the ICU. Curr Opin Anaesthesiol. 2019;32(2):156-162. doi:10.1097/ACO.0000000000000706
14.    Beck RW, Connor CG, Mullen DM, Wesley DM, Bergenstal RM. The fallacy of average: how using HbA1c alone to assess glycemic control can be misleading. Diabetes Care. 2017;40(8):994-999. doi:10.2337/dc17-0636
15.    Azhar A, Gillani SW, Mohiuddin G, Majeed RA. A systematic review on clinical implication of continuous    glucose    monitoring        in    diabetes management.    J    Pharm    Bioallied    Sci. 2020;12(2):102-111. doi:10.4103/jpbs.JPBS_7_20
16.    Brazeau AS, Mircescu H, Desjardins K, et al. Carbohydrate counting accuracy and blood glucose variability in adults with type 1 diabetes. Diabetes Res Clin Pract. 2013;99(1):19-23. doi:10.1016/j.diabres.2012.10.024
17.    McArdle PD, Mellor D, Rilstone S, Taplin J. The role of carbohydrate in diabetes management. Pract    Diabetes.    2016;33(7):237-42. doi:10.1002/pdi.2048
18.    Petrie JR, Peters AL, Bergenstal RM, Holl RW, Fleming GA, Heinemann L. Improving the clinical value and utility of CGM systems: issues and recommendations : A joint statement of the European Association for the Study of Diabetes and the American Diabetes Association Diabetes Technology Working Group. Diabetologia. 2017;60(12):2319-2328.   doi:10.1007/s00125- 017-4463-4
19.    Danne T, Nimri R, Battelino T, et al. International consensus on use of continuous glucose monitoring. Diabetes Care.  2017;40(12):1631-1640. doi:10.2337/dc17-1600
20.    Moström P, Ahlén E, Imberg H, Hansson PO, Lind M. Adherence of self-monitoring of blood glucose in persons with type 1 diabetes in Sweden. BMJ Open Diabetes Res Care. 2017;5(1):e000342. doi:10.1136/bmjdrc-2016-000342
21.    Zhong VW, Crandell JL, Shay CM, et al. Dietary intake and risk of non-severe hypoglycemia in adolescents with type 1 diabetes. J Diabetes Complications.    2017;31(8):1340-1347. doi:10.1016/j.jdiacomp.2017.04.017
22.    Zeevi D, Korem T, Zmora N, et al. Personalized nutrition by prediction of glycemic responses. Cell. 2015;163(5):1079-1094. doi:10.1016/j.cell.2015.11.001
23.    Bell KJ, Smart CE, Steil GM, Brand-Miller JC, King B, Wolpert HA. Impact of fat, protein, and glycemic index on postprandial glucose control in type 1 diabetes: implications for intensive diabetes management in the continuous glucose monitoring era. Diabetes Care. 2015;38(6):1008-1015. doi:10.2337/dc15-0100
24.    Stone JY, Bailey TS. Benefits and limitations of continuous glucose monitoring in type 1 diabetes. Expert Rev Endocrinol Metab. 2020;15(1):41-49. doi:10.1080/17446651.2020.1706482
25.    Ziegler R, Heinemann L, Freckmann G, Schnell O, Hinzmann R, Kulzer B. Intermittent use of continuous glucose monitoring: expanding the clinical value of CGM. J Diabetes Sci Technol. 2021;15(3):684-694. doi:10.1177/1932296820905577
27.    James S, Perry L, Gallagher R, Lowe J. Diabetes educators: perceived experiences, supports and barriers to use of common diabetes-related technologies.    J    Diabetes    Sci    Technol. 2016;10(5):1115-1121. doi:10.1177/1932296816660326
28.    Gabbay MAL, Rodacki M, Calliari LE, et al. Time in range: a new parameter to evaluate blood glucose control in patients with diabetes. Diabetol Metab Syndr. 2020;12:22. doi: 10.1186/s13098- 020-00529-z
29.    Battelino T, Danne T, Bergenstal RM, et al. Clinical targets for continuous glucose monitoring data interpretation: recommendations from the International Consensus on Time in Range. Diabetes Care.    2019;42(8):1593-1603. doi:10.2337/dci19-0028
30.    American Diabetes Association. 6. Glycemic targets: standards of medical care in diabetes- 2021. Diabetes Care. 2021;44(Suppl 1):S73-S84. doi:10.2337/dc21-S006
31.    Hirsch IB, Battelino T, Peters AL, Chamberlain JJ, Aleppo G, Bergenstal RM. Role of continuous glucose monitoring in diabetes treatment. Arlington, VA: American Diabetes Association; 2018.
32.    Lanspa MJ, Krinsley JS, Hersh AM, et al. Percentage of time in range 70 to 139 mg/dl is associated with reduced mortality among critically ill patients receiving IV insulin infusion. Chest. 2019;156(5):878-886. doi:10.1016/j.chest.2019.05.016
33.    Vigersky RA, McMahon C. The relationship of hemoglobin A1C to time-in-range in patients with diabetes. Diabetes Technol Ther. 2019;21(2):81-85. doi:10.1089/dia.2018.0310
34.    Cutruzzolà A, Irace C, Parise M, et al. Time spent in target range assessed by self-monitoring blood glucose associates with glycated hemoglobin in insulin treated patients with diabetes. Nutr Metab Cardiovasc Dis. 2020;30(10):1800-1805. doi:10.1016/j.numecd.2020.06.009
35.    Dovc K, Battelino T. Evolution of diabetes technology. Endocrinol Metab Clin North Am. 2020;49(1):1-18. doi:10.1016/j.ecl.2019.10.009
36.    Hui CY, Walton R, McKinstry B, Jackson T, Parker R, Pinnock H. The use of mobile applications to support self-management for people with asthma: a systematic review of controlled studies to identify features associated with clinical effectiveness and adherence. J Am Med Inform Assoc.    2017;24(3):619-632. doi:10.1093/jamia/ocw143
37.    Haase J, Farris KB, Dorsch MP. Mobile applications to improve medication adherence. Telemed J E Health. 2017;23(2):75-79. doi:10.1089/tmj.2015.0227
38.    Choi A, Lovett AW, Kang J, Lee K, Choi L. Mobile applications to improve medication adherence: existing apps, quality of life and future directions. Adv Pharmacol Pharm. 2015;3(3):64-74. doi:10.13189/app.2015.030302
39.    Liang X, Wang Q, Yang X, et al. Effect of mobile phone intervention for diabetes on glycaemic control: a meta-analysis. Diabet Med. 2011;28(4):455-463.    doi:10.1111/j.1464- 5491.2010.03180.x
40.    Yang Y, Lee EY, Kim HS, Lee SH, Yoon KH, Cho JH. Effect of a mobile phone-based glucose- monitoring and feedback system for type 2 diabetes management in multiple primary care clinic settings: cluster randomized controlled trial. JMIR Mhealth Uhealth. 2020;8(2):e16266. doi:10.2196/16266
41.    Hou C, Carter B, Hewitt J, Francisa T, Mayor S. Do mobile phone applications improve glycemic control (HbA1c) in the self-management of diabetes? A Systematic Review, Meta-analysis, and GRADE of 14 Randomized Trials. Diabetes Care. 2016;39(11):2089-2095. doi:10.2337/dc16- 0346
42.    Fonda SJ, Lewis DG, Vigersky RA. Minding the gaps in continuous glucose monitoring: a method to repair gaps to achieve more accurate glucometrics. J Diabetes Sci Technol. 2013;7(1):88-92. doi:10.1177/193229681300700110
43.    Raybaut P. Spyder-documentation. Available online at: pythonhosted org. 2009
44.    Glantz SA Primer of biostatistics (fourth edn.). New York: McGraw-Hill; 1997. pp. 350–360
45.    Kebede MM, Pischke CR. Corrigendum: popular diabetes apps and the impact of diabetes app use on self-care behaviour: a survey among the digital community of persons with diabetes on social media. Front Endocrinol (Lausanne). 2019;10:220. doi:10.3389/fendo.2019.00220
46.    Boyle L, Grainger R, Hall RM, Krebs JD. Use of and beliefs about mobile phone apps for diabetes self-management: surveys of people in a hospital diabetes clinic and diabetes health professionals in New Zealand. JMIR Mhealth Uhealth.
2017;5(6):e85. doi:10.2196/mhealth.7263
47.    American Diabetes Association. Standards of medical care in diabetes--2014. Diabetes Care. 2014;37   Suppl   1:S14-80.   doi:10.2337/dc14- S014
48.    Stehouwer CDA. Microvascular dysfunction and hyperglycemia: a vicious cycle with widespread consequences. Diabetes. 2018;67(9):1729-1741. doi:10.2337/dbi17-0044
49.    Schwingshackl L, Hoffmann G. Comparison of the long-term effects of high-fat v. low-fat diet consumption on cardiometabolic risk factors in subjects with abnormal glucose metabolism: a systematic review and meta-analysis. Br J Nutr. 2014;111(12):2047-2058. doi:10.1017/S0007114514000464
50.    Rodbard D. Metrics to evaluate quality of glycemic control: comparison of time in target, hypoglycemic, and hyperglycemic ranges with risk indices. Diabetes Technol Ther. 2018;20(5):325- 334. doi:10.1089/dia.2017.0416
51.    Beck RW, Riddlesworth TD, Ruedy K, et al. Continuous glucose monitoring versus usual care in patients with type 2 diabetes receiving multiple daily insulin injections: a randomized trial. Ann Intern Med. 2017;167(6):365-374. doi:10.7326/M16-2855.