Improving Glycemic Control in Type 2 Diabetes Using Mobile Applications and e-Coaching: A Mixed Treatment Comparison Network Meta-Analysis

Abstract

Objectives:

This study compared the effectiveness of glycemic control among usual care, care management using a mobile application (app), and management using an app with additional e-coaching for patients with type 2 diabetes mellitus (T2DM) using a mixed treatment comparison (MTC) network meta-analysis (NMA).

Methods:

A systematic search for published randomized controlled trials (RCTs) was conducted, which included Pubmed, Web of Science, Cochrane Central Register of Controlled Trials, CINAL, Koreamed, KMbase, and ScienceOn, until October 2020. Among the 10,391 studies identified after removing duplicates, 14 RCTs were finally included in the MTC NMA. Data extraction and methodological quality assessment using version 2 of the Cochrane tool for assessing the risk-of-bias in randomized trials (RoB 2) was performed. The comparative efficacy was analyzed using the random-effects NMA based on a frequentist model by the intervention group and main outcome variables.

Results:

At the 3-month follow-up after each intervention, a comparison of the P-scores revealed the app plus e-coaching intervention to be the most effective method for reducing the HbA1c level in a homogeneous gender ratio group (P-score 0.92). At the 6-month follow-up period, app intervention was the best in reducing the HbA1c level in the homogeneous gender ratio and under 60 years of age group (P-score 1.00).

Conclusions:

Based on MTC analysis using the data from published RCTs, mobile apps or apps with e-coaching interventions for T2DM patients were more effective in improving the HbA1c values, FBS, and hypoglycemia frequency than usual care. Nevertheless, further research will be needed to clarify the effects of adding e-coaching to the app.

Study registration:

Research Registry UIN (reviewregistry780)

Keywords

type 2 diabetes glycemic control mobile applications e-coaching mixed treatment comparison

Introduction

Diabetes mellitus (DM) is one of the fastest-growing chronic diseases worldwide. Nevertheless, the disease is treatable, and its consequences can be prevented or delayed by regular blood sugar monitoring, management of complications, diet control, regular physical activity, and diabetes medication.¹ Furthermore, the possibility of complications after DM’s onset can be reduced by healthy lifestyle changes, such as maintaining a healthy body weight, eating a healthy diet, maintaining regular exercise, not smoking, and reduced alcohol consumption.²

In this respect, the appropriate self-monitoring of the blood glucose (SMBG) helps achieve their HbA1c target and modify their lifestyle habits.^3,4 On the other hand, despite the evidence-based benefits of SMBG for type 2 DM (T2DM) management, there are limitations in the timeliness of adequate intervention input, sufficient frequency, and accuracy.^5,6 The American Diabetes Association (ADA)’s clinical practice recommendations state that when prescribing SMBG, physicians should ensure that the patients receive education and regular evaluations of the technique, results, and ability to use data from SMBG to adjust therapy.⁷

Recently, the development of information technology (IT), including technology-assisted self-management support systems, has allowed technical health-applicants to use web-based interventions to adhere to their healthy lifestyle habit, manage blood glucose levels, prevent diabetes complications, and improve the quality of life (QOL) of patients with T2DM.^3,7 Some evidence exists regarding the benefits of Web- and mobile phone-based diabetes management.^8,9 Moreover, the effects of adding e-coaching in various ways to strengthen the effects of web-based systems have been studied.^10-13 Nevertheless, few studies have compared the effectiveness between usual care, application management, and management with added e-coaching to a mobile application (app) for patients with T2DM.

Therefore, this study compared the effectiveness of the following 3 interventions using a mixed treatment comparison (MTC) network meta-analysis (NMA): (1) usual care, (2) care using an app to perform SMBG on a mobile phone or device, and (3) using an app to perform SMBG on a mobile phone or device and managed with additional e-coaching for patients with T2DM.

Methods

Research Question

This study examined what is effective in improving blood sugar control in T2DM among apps only, using apps with e-coaching, and usual care. On the other hand, there is still no consensus on the definition of e-coaching, but after reviewing the previous articles, the definition of e-coaching is defined as medical assistance provided based on the patient’s information collected in the app in various ways, such as direct coaching, remote coaching, and automatic coaching.¹⁴

The PICO related to this research question were as follows.

Population: type 2 DM in adults

Intervention:

(1) app (that is, care using an app to perform SMBG on a mobile phone or device) (T2),

(2) app plus e-coaching (that is, care using an app and managed with additional e-coaching from healthcare providers) (T3)

Comparator: usual care (T1)

Outcomes:

(1) glycated hemoglobin (HbA1c)

(2) fasting blood sugar (FBS)

(3) hypoglycemia frequency

Search Strategy

The literature database search included the core database Pubmed (https://www.ncbi.nlm.nih.gov/pubmed), Web of Science (https://apps.webofknowledge.com), Cochrane Central Register of Controlled Trials (https://www.cochranelibrary.com/advanced-search), CINAL (https://www.riss.kr/foreign/cinahl.do) and 3 domestic Korean databases, Koreamed (https://www.koreamed.org), Korean Medicine Database (KMBASE) (https://kmbase.medric.or.kr) and ScienceOn. (https://scienceon.kisti.re.kr/main/mainForm.do). The search keywords using both subject headings and keywords according to population and intervention: Diabetes Mellitus, Type2 [mesh], telemedicine [mesh], self-care [mesh], smartphone [mesh], cell phone [mesh], counseling [mesh], feedback [mesh], health promotion [mesh], and disease management [mesh]. The literature search was performed up to October 2020, and there were no language restrictions (Appendix 1).

Inclusion/Exclusion Criteria

Studies that met the following criteria were included.

(1) Population: type 2 DM in adults,

(2) Intervention: mobile apps or mobile apps plus personalized e-coaching,

(3) Study design: randomized controlled trial.

The exclusion criteria were as follows: (1) those who do not have type 2 DM; (2) studies without using mobile apps as an intervention; (3) Studies that report only outcomes related to application use, such as application satisfaction and utilization rate; (4) studies in which meta-analysis was impossible due to the insufficient reporting of outcome values; (5) study designs excluding a randomized controlled study (RCT), that is, non-RCT, such as systematic review (SR), qualitative research, trial protocols, research letters, and conference abstracts.

Data Extraction and Assessment of Bias

Two reviewers extracted the data from each study independently. The data were reviewed for consistency between the 2 extractors, and disagreements were resolved by consensus. For each study, the data extraction details included the selection criteria, population characteristics, the type of intervention, outcome measures, study results, and follow-up period.

In addition, 2 researchers independently evaluated the methodological quality of the studies included using version 2 the Cochrane tool for assessing the risk-of-bias in randomized trials (RoB 2), in which 5 bias domains were evaluated: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported results.¹⁵ The response options of the risk-of-bias judgments for each domain and overall were “low risk of bias,” “some concerns,” or “high risk of bias”.¹⁵ In particular, the overall risk-of-bias judgment of studies was graded as follows: (1) “low risk of bias” when a low risk of bias was determined for all domains; (2) “some concerns” if at least one domain was assessed as raising some concerns, but not to be at high risk of bias for any single domain; or (3) “high risk of bias” when a high risk of bias was reached for at least one domain or the study judgment included some concerns in multiple domains.

Statistical Analysis

The random-effects network meta-analysis (NMA) was conducted based on a frequentist model by the intervention group and main outcome variables, with mean difference (MD) or odds ratio (OR) as the effect size between the interventions: usual care (T1) vs apps (T2) vs apps plus e-coaching (T3). NMA can evaluate multiple interventions that may, or may not, have been compared directly in clinical trials, thereby providing a hierarchical ranking for multiple groups without direct comparisons.¹⁵ The 2 basic assumptions of NMA were transitivity and consistency. The hypothesis of transitivity is that it allows a statistical comparison between A and B when obtaining the head-to-head results of A vs C and B vs C, respectively.¹⁶ In addition, the consistency between the estimates of the treatment effect obtained from a direct and indirect estimate is important to the assumption that maintains the NMA, so the difference was tested based on the statistics on the difference between the direct and indirect estimates using the automated generation of node splitting models to verify the consistency, and check for consistency with the Q static and P-value.^15,17 The indicators that assess the presence of statistical heterogeneity in NMA were I² and the Q static and P-value. The I² values were <25%, 25%-50%, and >50%, indicating low, moderate, and high heterogeneity, respectively. P values less than .1 indicated significant heterogeneity and inconsistency.¹⁶ As a result of the NMA, subgroup analysis was performed according to the gender ratio and age group when heterogeneity and inconsistency were found.

Network forest plots were used to show the pooled effect estimates and their associated uncertainty for all interventions compared to the reference treatment. The plot also displayed the 95% confidence interval (CI) along with the estimates and the P-score used to rank all therapies based on network estimates. In this study, higher values indicated better treatments.

R (version 3.6.1) was used as the software for statistical analysis, and the functions included in the “netmeta” package were utilized.¹⁶ A statistical hypothesis test was performed based on the significance level of .05.

Results

Selected Studies

Fourteen RCTs were included in the meta-analysis (Figure 1, Appendix 2-3). The follow-up period ranged from 3 months to 1 year. Table 1 lists the demo and clinical characteristics of each study. The studies included were conducted in several different countries, such as China, the Republic of Korea, Norway, the USA, the UK, and Canada. Various apps were used for each study included. Some apps automatically transmit the blood glucose values through Bluetooth. There were differences in the additional functions such as nutrition and exercise and inputting the blood glucose values for each app. Furthermore, there was a difference in the e-coaching method, namely the coaching subject, method, and frequency. The coaching practitioners were members of a variety of medical teams including physician, nurse, and/or dietician, and it was done at a frequency of any time to once a month using messages or phone calls (Tables 1 and 2).

Figure 1.

Flow diagram of study selection.

Table 1.

Characteristics of the Included Studies.

Study ID		1.1	1.2	1.3	1.4	1.5	1.6	1.7	1.8	1.9	2.1	2.2	2.3	3.1	3.2
First author, year		Quinn (2011)	Nagrebetsky (2013)	karhula (2015)	Wayne (2015)	Lee (2017)	Kim (2019)	Sun (2019)	Yu (2019)	Yang (2020)	Kleinman (2017)	Alanzi (2018)	Kusnanto (2019)	Holmen (2014)	Torbjornsen (2014)
Country		USA	UK	Finland	Canada	Malaysia	Korea	China	China	Korea	India	Saudi Arabia	Indonesia	Norway	Norway
Follow up period (month)		12	6	12	6	3	6	6	6	3	6	6	3	12	4
No. of patients (baseline)	T1	56	7	70	49	40	82	47	47	97	46	10	15	50	50
	T2										44	10	15	51	51
	T3	62	7	180	48	45	90	44	45	150				50	50
Age:mean (SD)	T1	53.2 (8.4)	60 (13)	65.5 (9.6)	53.3 (11.9)	553.77 (8.03)	60.0 (8.4)	668.04	56.2 (8.4)	60.6 (10.2)	48.0 (9.5)			55.9 (12.2)	55.9 (12.2)
	T2										48.8 (9.0)			58.6 (11.8)	58.6 (11.8)
	T3	52 (8.0)	56 (8)	66.6 (8.2)	53.1 (10.9)	53.24 (7.29)	60.0 (8.4)	67.9	50.3 (10.4)	54.1 (10.1)				57.4 (12.1)	57.4 (12.1)
Male/Female (n)	T1	28/28	5/2	40/30	10/39	16/24	39/43	18/29	501 pt	45/52	27/19	7/3	6/9	30/20	30/20
	T2										36/8	8/2	7/8	34/17	34/17
	T3	31/31	5/2	81/99	17/31	24/21	50/40	19/25	31/14	80/70				25/25	25/25
T2DM duration (year (sd))	T1	9.0 (7.0)	2.3			10.04 (7.64)	12.5 (7.3)	11.52 (7.73)			8.5			9.4 (5.5)	9.4 (5.5)
	T2										10			11.2 (7.3)	11.2 (7.3)
	T3	8.2 (5.3)	3			7.91 (4.81)	13.2 (8.0)	11.19 (6.39)						9.6 (8.4)	9.6 (8.4)
BMI(kg/m²):mean(SD)	T1	34.3 (6.3)	32.4 (6.2)	30.9 (5.7)	37.00 (7.92)	30.28 (5.05)	25.8 (4.1)	23.3	25.3 (3.0)	25.7 (3.9)	28.0 (4.2)			32.0 (6.0)	32.0 (6.0)
	T2										29.7 (6.0)			32.4 (6.5)	32.4 (6.5)
	T3	35.8 (7.1)	33.4 (7.1)	31.1 (5.4)	33.74 (6.70)	29.20 (5.98)	25.5 (3.2)	23.6	26.0 (3.9)	26.3 (3.7)				30.7 (5.6)	30.7 (5.6)
Waist circumference (cm):mean (SD)	T1				113.88 (17.04)				93.2 (9.1)	87 (9.8)
	T2
	T3				112.11 (14.50)				91.5 (8.6)	89.5 (8.9)
HbA1c (%):mean (SD)	T1	9.2 (1.7)	8.19 (3.34)	7.2	8.89 (1.30)	8.79 (1.15)	7.8 (0.7)	7.88 (0.64)	8.7 (1.9)	7.9 (0.8)	9.1 (1.1)		8.18 (1.02)	8.3 (1.2)	8.3 (1.2)
	T2					8.79 (1.15)					9.4 (1.2)		8.74 (1.34)	8.1 (1.1)	8.1 (1.1)
	T3	9.9 (2.1)	8.01 (3.16)	7.25	8.69 (1.32)		7.7 (0.7)	7.84 (0.73)	8.3 (1.9)	8 (0.8)				8.2 (1.1)	8.2 (1.1)

Table 2.

Type of Intervention in the Included Studies.

Study ID	Author (year)	Application			E-coaching
Study ID	Author (year)	Type	Automatic transfer of blood glucose values to the app	Functions	Who	Method	Frequency
1.1	Quinn (2011)	Mobile diabetes management software application (MDMA)		Enter diabetes self-care data (blood glucose values, carbohydrate intake, medications, other diabetes management information).	Physicians	Telephone call, messages	Monthly
1.1	Quinn (2011)	Mobile diabetes management software application (MDMA)		Provide patients with automated, real-time educational, behavioral, and motivational messages specific to the entered data.	Physicians	Telephone call, messages	Monthly
1.2	Nagrebetsky (2013)	t+ Diabetes, t+ Medical	√	Request the patient take at least 6 blood glucose tests per week (of which at least 3 measure the fasting glucose).	Nursing staff	Text messages and telephone calls	Monthly
1.2	Nagrebetsky (2013)	t+ Diabetes, t+ Medical	√	Provide patients with real-time graphical feedback on their glucose levels.	Nursing staff	Text messages and telephone calls	Monthly
1.3	karhula (2015)	Mobile personal health record (PHR) app	√	Upload measurements (blood glucose, blood pressure, weight, and steps), Receive feedback and send reminders.	Health coaches	Telephone call	Every 4 to 6 weeks
1.4	Wayne (2015)	Connected wellness platform (CWP)		Track key indicators (blood sugar levels, exercise, frequency/duration/intensity, food intake, and mood)	Health coach	Message, telephone call, in-person meeting	Any time
1.4	Wayne (2015)	Connected wellness platform (CWP)		Encourage the patients to use the system daily	Health coach	Message, telephone call, in-person meeting	Any time
1.5	Lee (2017)	Mobile diabetes sofware application	√	Request measurements of the blood glucose 6 times daily during the Ramadan period, or more often if they wish.	Multi-disciplinary health care team consisting of pharmacists, physicians, and nurses
				Automatically generate a message alert to inform the researcher in the event that there are 3 consecutive hypoglycaemic readings during the study (≤3.9 mmol/L)
				Provide patients with advice on diabetes management, medication adherence, and lifestyle modification.
1.6	Kim (2019)	mDiabetes application	√	Setting specific goals to improve the patients’ dietary habit (eating regularly, limiting alcohol, limiting sodium, and limiting refined grains and added sugar)	Physicians
				Glucose monitoring (Immediate feedback, feedback on hyper-/hypoglycaemia, adherence to medication, and insulin dose)
				Diet control (caloric counting and feedback, nutrition information, and management of individual goals)
				Exercise control (step count monitoring and management of individual goals)
				Encourage patients to record their readings daily.
				Upload the glucometer data.
1.7	Sun (2019)	mHealth management app	√	Send medical advice to the patients and reminders to monitor their glucose levels.	Medical teams	Personal messaging app, telephone	Every 2 weeks (dietary recommendations: monthly)
				Input the daily dietary intake.
				Send the daily dietary records of each patient to the dietician and send the dietician’s dietary recommendations to the patients.
				Diabetic education to adopt a better lifestyle.
1.8	Yu (2019)	Diabetes-carer	√	Self-management (enter the daily recordings of the SMBG, diet, physical activities, etc.)	Clinicians	Voice and message patterns	Real-time communication
				Patient community (share their achievements and experiences with other patients)
				Real-time communication between patients and clinicians
				Automatically generate a message to the patient and notify the clinicians when the blood glucose value exceeds the safety value.
				Uploads the daily SMBG.
1.9	Yang (2020)	Hicare smart K	√	Requires input of the mealtime when the SMBG was checked	Physicians of the primary care clinic	Messages, telephone call	At least once a week
				Requests the patient inputs their blood pressure at least once a week
				Input the patient’s weight data.
				Depending on whether or not patients have met their SMBG goals, send messages of praise/encouragement, advice on dietary control and exercise, and advice on the regular glucose checkups.
				Reminders
2.1	Kleinman (2017)	Gather health		Data visualization
				Ongoing support to increase self-care behaviors
				Facilitate collaborative care decisions
				Smart mobile diabetes management module (SMDMM)
2.2	Alanzi (2018)	Saudi Arabia networking aiding diabetes (SANAD) system	√	Social networking module (SNM)
				Cognitive-behavioral therapy module (CBT-M)
				Notify the patient 6 times a day: at 06.00 (breakfast), 09.00 (morning snack), 12.00 (lunch), 15.00 (afternoon snack), 16.00 (physical activity/diabetic foot gymnastics), and 18.00 (dinner).
2.3	Kusnanto (2019)	DM-calendar app	√	Remind the patient of a blood sugar check once a month.
2.3	Kusnanto (2019)	DM-calendar app	√	Diabetes diary app including visual graphs, trend reports, and feedback.
3.1	Holmen (2014)	Few touch application (FTA)	√	Food habit registration system	Diabetes specialist nurse	Telephone call	Each month for 4 months, 5 in total
				Physical activity registration system
				Personal goal-setting system
				General information system
				Blood glucose data management system
3.2	Torbjornsen (2014)	Few touch application (FTA)	√	Food habits data management system	Diabetes specialist nurse	Telephone call	Five times in the first 4 months, an average of 20 minutes each
				Physical activity data management system
				Personal goal-setting system
				General diabetes information look-up system

Abbreviations: SMBG, self-monitoring of blood glucose.

Improving Glycemic Control: Glycated Hemoglobin (HbA1c)

Thirteen studies were included in the NMA that reported or could calculate the mean difference and SD between before and after for each group (Table 3).

Table 3.

Comparative Effect on HbA1c.

Follow-up (M)		No.	Author (year)	Outcomes by intervention									Comparison	MD	95% CI
				T1			T2			T3					95% CI
				n	Mean	SD	n	Mean	SD	n	Mean	SD			Lower	Upper
HbA1c
3	Overall
		3.2	Torbjornsen (2014)	39	−0.39	0.18	40	−0.23	0.12	39	−0.41	0.15	T2 vs T1	0.058	−0.336	0.452
		1.5	Lee (2017)	40	−0.22	1.48	-	-	-	45	−1.06	1.66	T3 vs T1	−0.224	−0.467	0.019
		1.8	Yu (2019)	47	−1.20	0.40	-	-	-	45	−1.30	0.40	T3 vs T2	−0.282	−0.676	0.112
		1.9	Yang (2020)	94	−0.28	0.08	-	-	-	145	−0.63	0.07
	Subgroup: gender ratio
	Homogeneous	3.2	Torbjornsen (2014)	39	−0.39	0.18	40	−0.23	0.12	39	−0.41	0.15	T2 vs T1	0.152	0.096	0.208
		1.8	Yu (2019)	47	−1.20	0.40	-	-	-	45	−1.30	0.40	T3 vs T1	−0.036	−0.106	0.034
													T3 vs T2	−0.188	−0.244	−0.132
	Heterogeneous	1.5	Lee (2017)	40	−0.22	1.48	-	-	-	45	−1.06	1.66
	Heterogeneous	1.9	Yang (2020)	94	−0.28	0.08	-	-	-	145	−0.63	0.07	T3 vs T1	−0.477	−0.898	−0.056

6 (ITT*)	Overall
		2.2	Alanzi (2018)	10	−0.05	0.31	9	−0.6	0.35	-	-	-	T2 vs T1	−0.604	−1.045	−0.163
		2.1	Kleinman (2017)	46	−0.80	1.60	44	−1.50	1.10	-	-	-	T3 vs T1	−0.140	−0.407	0.128
		1.6	Kim (2019)	82	−0.06	0.10	-	-	-	90	−0.40	0.09	T3 vs T2	0.464	−0.051	0.980
		1.7	Sun (2019)	47	−0.62	1.00	-	-	-	44	−1.07	0.89
		3.3	Yu (2019)	47	−1.10	0.40	-	-	-	45	−1.10	0.30
		1.4	Wayne (2015)	64	−0.97	1.40	-	-	-	67	−0.64	1.04
	Subgroup: gender ratio
	Homogeneous	2.2	Alanzi (2018)	10	−0.05	0.31	9	−0.6	0.35	-	-	-	T2 vs T1	−0.602	−1.016	−0.188
		2.1	Kleinman (2017)	46	−0.80	1.60	44	−1.50	1.10	-	-	-	T3 vs T1	-0.042	−0.343	0.261
		1.7	Sun (2019)	47	−0.62	1.00	-	-	-	44	−1.07	0.89	T3 vs T2	0.560	0.048	1.073
		3.3	Yu (2019)	47	−1.10	0.40	-	-	-	45	−1.10	0.30
		1.4	Wayne (2015)	64	−0.97	1.40	-	-	-	67	−0.64	1.04
	Subgroup: gender ratio & age
	Homogeneous & <60 years	2.2	Alanzi (2018)	10	−0.05	0.31	9	−0.6	0.35	-	-	-	T2 vs T1	−0.585	−0.867	−0.303
		2.1	Kleinman (2017)	46	−0.80	1.60	44	−1.50	1.10	-	-	-	T3 vs T1	0.050	−0.125	0.225
		3.3	Yu (2019)	47	−1.10	0.40	-	-	-	45	−1.10	0.30	T3 vs T2	0.635	0.303	0.967
		1.4	Wayne (2015)	64	−0.97	1.40	-	-	-	67	−0.64	1.04
6 (PP*)	Overall
		2.2	Alanzi (2018)	10	−0.05	0.31	9	−0.06	0.35	-	-	-	T2 vs T1	−0.605	−1.053	−0.157
		2.1	Kleinman (2017)	46	−0.08	1.60	44	−1.50	1.10	-	-	-	T3 vs T1	−0.232	−0.508	0.044
		1.6	Kim (2019)	71	0.00	0.10	-	-	-	80	−0.40	0.09	T3 vs T2	0.373	−0.153	0.899
		1.7	Sun (2019)	47	−0.62	1.00	-	-	-	44	−1.07	0.89
		3.3	Yu (2019)	47	−1.10	0.40	-	-	-	45	−1.10	0.30
		1.4	Wayne (2015)	49	−0.76	1.39	-	-	-	48	−0.82	1.05
	Subgroup: gender ratio
	Homogeneous	2.2	Alanzi (2018)	10	−0.05	0.31	9	−0.06	0.35	-	-	-	T2 vs T1	−0.593	−0.931	−0.256
		2.1	Kleinman (2017)	46	−0.08	1.60	44	−1.50	1.10	-	-	-	T3 vs T1	−0.118	−0.351	0.115
		1.7	Sun (2019)	47	−0.62	1.00	-	-	-	44	−1.07	0.89	T3 vs T2	0.475	0.065	0.885
		3.3	Yu (2019)	47	−1.10	0.40	-	-	-	45	−1.10	0.30
		1.4	Wayne (2015)	49	−0.76	1.39	-	-	-	48	−0.82	1.05
12		1.1	Quinn (2011)	51	−0.70	0.20	-	-	-	56	−1.90	0.20	T2 vs T1	−0.376	−1.358	0.606
		3.1	Holmen (2014)	41	−0.16	0.17	39	−0.31	0.18	40	−0.15	0.22	T3 vs T1	−0.443	−1.063	0.177
		1.3	Karhula (2015)	61	0.18	0.09	-	-	-	156	0.04	0.07	T3 vs T2	−0.067	−1.049	0.915

CI, confidence interval; MD, mean difference; T1, usual-care (control), T2, mobile application, T3, mobile application plus e-coaching.

ITT, ITT was used when both ITT and PP are present; PP, PP was used when both ITT and PP are present.

Four studies reported the outcome of the 3-month follow-up period, and the analysis results are as follows. Looking at the estimate and 95% CI for the mean difference between interventions, that is, T2 vs T1, T3 vs T1, and T3 vs T2, the 95% CI was zero. Therefore, there was no significant difference between the interventions for HbA1c reduction.

The P-scores were next compared. The T3 intervention was the best in reducing the HbA1c (P-score 0.94) (Figure 2a). On the other hand, since heterogeneity existed (I² = 96%, Q = 11.72, P-value = .0028), subgroup analysis was performed by distinguishing whether the gender ratio was homogeneous in the groups. In the homogeneous gender ratio groups, there was a significant difference between the interventions for HbA1c reduction, excluding T3 vs T1. A comparison of the P-scores revealed the T3 intervention to be the best for HbA1c reduction (P-score = 0.92), and there was no heterogeneity (I² = 0%, Q = 0.0, P-value = 1.000) or inconsistency (Q = 0.80, P-value = .3711) issues that could be interpreted as the final meta-analysis result (Figure 2b). In the heterogeneous gender ratio groups, there was a significant difference between the interventions for HbA1c reduction. A comparison of the P-scores revealed the T3 intervention to be the best for HbA1c reduction (P-score = 0.99), and there was no heterogeneity (I² = 51.8%, Q = 2.08, P-value = .1497) that could be interpreted as the final meta-analysis result (Figure 2c).

Figure 2.

Network forest plot of HbA1c. (a) total (3 months) (b) homogeneous gender ratio (3 months) (c) heterogeneous gender ratio (3 months) (d) total (6 months, ITT) (e) homogeneous gender ratio (6 months, ITT) (f) homogeneous gender ratio and <60 years old (6 months, ITT) (g) homogeneous gender ratio (6 months, PP) (h) total (12 months).

Six studies reported the outcome of the 6-month follow-up period, and the analysis results were as follows. Intention to treat (ITT) analysis revealed no significant difference between the interventions for HbA1c reduction. A comparison of the P-scores revealed the T2 intervention to be the best in reducing HbA1c (P-score = 0.98) (Figure 2d). On the other hand, because heterogeneity existed (I² = 87.8%, Q = 32.75, P-value = <.0001), subgroup analysis was performed by distinguishing whether the gender ratio was homogeneous between the groups. In the homogeneous gender ratio groups, there was a significant difference between the interventions for HbA1c reduction, excluding T3 vs T1. A comparison of the P-scores showed that the T2 intervention was the best for HbA1c reduction (P-score = 0.99) (Figure 2e). The heterogeneity was reduced somewhat (I² = 59.7%, Q = 7.44, P-value = .059), but it still existed based on the statistical significance. Therefore, other subgroup analyses were performed based on the age group. The homogeneous gender ratio and under 60 years of age group had a significant difference between the interventions for HbA1c reduction, except for T3 vs T1. Moreover, there was no heterogeneity (I² = 11.3%, Q = 2.25, P-value = .324) that could be interpreted as the final meta-analysis result (Figure 2f).

On the other hand, the per-protocol (PP) analysis results showed that the T2 intervention was the best for HbA1c reduction in the homogeneous gender ratio group (P-score = 0.99) (Figure 2g).

An analysis of the 3 studies reporting the outcome of the 12-month follow-up period showed no significant difference between the interventions in terms of the HbA1c reduction. Although the T3 intervention was the best in reducing the HbA1c level (P-score = 0.74), heterogeneity (I² = 99.7%, Q = 660.94, P-value = <.0001) and inconsistency (Q = 26.62, P-value= <.0001) not resolved by subgroup analysis were found (Figure 2h).

Fasting Blood Sugar (FBS) Level

Four studies that reported or could calculate the mean difference and SD between before and after for each group were included in the NMA (Table 4).

Table 4.

Comparative Effect on the Fasting Blood Sugar, Hypoglycemia Frequency and Change in BMI.

Follow-up (M)	No.	Author (year)	Outcomes by intervention									Comparison	MD	95% CI
			T1			T2			T3					95% CI
			n	Mean	SD	n	Mean	SD	n	Mean	SD			Lower	Upper
Fasting blood sugar
3	1.5	Lee (2017)	40	0.06	3.42	-	-	-	45	−0.01	3.57	T2 vs T1	−1.540	−2.724	−0.356
	1.8	Yu (2019)	47	0.18	0.69	-	-	-	45	−1.36	0.46	T3 vs T1	−0.889	−1.878	0.100
	1.9	Yang (2020)	94	0.10	0.20	-	-	-	145	−1.10	0.31	T3 vs T2	0.651	−0.892	2.194
6	1.6	Kim (2019)	82	0.70	0.20	-	-	-	90	0.10	0.20
6	1.8	Yu (2019)	47	0.24	0.79	-	-	-	45	−0.65	0.45	T3 vs T1	−0.716	−0.994	−0.438
Hypoglycemia frequency
6	1.2	Nagrebetsky (2013)	92	27	-	-	93	26	92	27	-	T3 vs T1	0.861	0.502	1.474
	1.6	Kim (2019)	47	10	-	-	45	6	47	10	-
	1.8	Yu (2019)	7	0	-	-	7	1	7	0	-

CI, confidence interval; MD, mean difference; T1, usual-care(control); T2, mobile application; T3, mobile application plus e-coaching.

ITT, ITT was used when both ITT and PP are present; PP, PP was used when both ITT and PP are present.

An analysis of 3 studies reporting the outcome of a 3-month follow-up period showed significant difference between the interventions in terms of the FBS reduction, and a comparison of the P-score revealed the T3 intervention to be the best (P-score = 0.96) (Figure 3a). On the other hand, in the subgroup analysis of heterogeneous gender ratio groups performed due to heterogeneity (I² = 79.1%, Q = 9.57, P-value = .0084), there was no significant difference between the interventions for FBS reduction. A comparison of the P-scores revealed the T3 intervention to be the best for FBS reduction (P-score = 1.00), and there was no heterogeneity (I² = 54.8%, Q = 2.21, P-value = .1368) that could be interpreted as the final meta-analysis result (Figure 3b).

Figure 3.

Network forest plot of fasting blood sugar and hypoglycemia frequency. (a) FBS (3 months) (b) FBS, heterogeneous gender ratio (3 months) (c) FBS (6 months) (d) hypoglycemia frequency (6 months).

An analysis of the 2 studies reporting the outcome of the 6-month follow-up period showed a significant difference between interventions in FBS reduction, and the P-score comparison revealed the T3 intervention to be the best (P-score = 1.00). However, heterogeneity existed. (I² = 77.8%, Q = 4.49, P-value = .0340) (Figure 3c).

Hypoglycemia Frequency

An analysis of 3 studies reporting the outcome of the 6-month follow-up period showed no statistically significant difference between the interventions in terms of the hypoglycemia frequency, but a comparison of the P-score showed that the T3 intervention was the best (P-score = 0.71). In addition, there was no heterogeneity (I² = 0%, Q = 1.21, P-value = .5474) (Table 4, Figure 3d).

Risk of Bias

The quality of the 14 selected RCT was assessed using RoB 2. The overall risk of bias was low in 57% (8/14), with some concerns in 36% (5/14). Domain1, the risk of bias arising from the randomization process, showed some concern in 21% (3/14). Domain 5, the risk of bias in the selection of the reported result, revealed some concern in 29% (4 /14) (Figure 4).

Figure 4.

Risk of bias assessment summary (RoB 2.0).

Discussion

Wearables or portable devices and smartphones provide patients with timely health information, particularly personalized advice, and by medical professionals to encourage healthy behavior.¹⁷ In addition, the use of new apps with digital health software and sensors for DM management is spreading rapidly.¹⁷ The effectiveness of SMBG with or without additional tools was consistent in some previous studies.^18,19 In addition, in previous SRs, the mobile app had a significant effect on the management of HbA1c in DM patients, but some SRs concluded that the function of the health care professionals was important for the apps to exert their effects.^9,20-24 The present study compared the effectiveness of usual care with that of an app with or without e-coaching interventions in patients with T2DM management using MTC methodology. An analysis, according to the follow-up period, showed that apps with or without e-coaching interventions were more effective treatments for reducing the HbA1c values, FBS, and hypoglycemia frequency than usual care. On the other hand, there was insufficient evidence to conclude that either an app or an app with e-coaching is definitively superior in this MTC analysis. In a previous SR, although personal health coaching programs were effective in improving glycemic control, they argued that further studies on the specific program components, training, and supervision approaches would be needed.²⁵ The possible causes of the research findings are as follows. The apps might not function optimally due to incomplete development. One systematic review examined the developmental considerations of mobile apps for diabetes self-management. All the developed apps had the SMBG function, but some did not consider behavioral health theory, user involvement, the involvement of clinical experts, data analytics, education, and reminder functions.²⁶ Therefore, there may be qualitative differences between apps that support SMBG for DM patients. Furthermore, a study analyzing the critical management functions, such as decision support and alerts for SMBG in T2DM in 2019, claimed that quality assurance mechanisms, such as certification, are needed because some problems with apps may have missed opportunities to improve the treatment and health outcomes.²⁷

The other was the heterogeneity and non-standardization of the e-coaching method. There also appears to be an issue about whether e-coaching involving healthcare professionals does not have overwhelming superiority over several functions automatically notified by apps and portals. In the RCTs included in this study, there were differences between studies on who coached, how they coached, and how often they coached. Moreover, it appears that these differences did not properly reflect the role of e-coaching. In other words, e-coaching guidelines, such as the type of application, coaching method, and frequency, will be needed to measure the effectiveness of e-coaching accurately. Accordingly, in 2020, the ADA newly announced the operation of the Certified Diabetes Care and Education Specialist (CDCES)’ through consensus with experts, such as clinicians, nutritionists, nurses, and pharmacists, and stated that in the future, experts who have CDCES would be active in the fields of T2DM management, prediabetes, and diabetes prevention.²⁸

In addition, some reports raised privacy issues while discussing whether the apps comply with essential data protection principles.²⁹ Despite these multiple environments and some limitations to web- and mobile phone-based diabetes management or plus e-coaching by healthcare professionals, the scope of digital health, including artificial intelligence and deep learning, is expanding into the diagnosis of diabetes, glucose management, and diagnosis and management of diabetes-related complications.^30-34

This study had some limitations. It was impossible to conduct an analysis that considers the diversity in mobile app types and e-coaching methods due to the limited number of studies included. However, several studies are being conducted in this field. Eventually, the limitations will be overcome, and the most preferential diabetes management method will be derived.

Conclusions

According to the MTC analysis using the data from published RCT, mobile apps or apps with e-coaching for T2DM management are more effective in reducing the HbA1c values, FBS, and hypoglycemia frequency than usual care in a 3- or 6-month follow up period. On the other hand, further research will be needed to clarify the difference between adding e-coaching to the app and its long-term effectiveness.

Supplemental Material

sj-pdf-1-dst-10.1177_19322968211010153 – Supplemental material for Improving Glycemic Control in Type 2 Diabetes Using Mobile Applications and e-Coaching: A Mixed Treatment Comparison Network Meta-Analysis

Supplemental material, sj-pdf-1-dst-10.1177_19322968211010153 for Improving Glycemic Control in Type 2 Diabetes Using Mobile Applications and e-Coaching: A Mixed Treatment Comparison Network Meta-Analysis by Min Kyung Hyun, Jang Won Lee, Seung-Hyun Ko and Jin Seub Hwang in Journal of Diabetes Science and Technology

Footnotes

Abbreviations

AMA, American Diabetes Association; App, application; CI, confidence interval; DM, diabetes mellitus; FBS, fasting blood sugar; HbA1c, glycated hemoglobin; IT, information technology; ITT, intention to treat; MD, mean difference; MTC, mixed treatment comparison; NMA, network meta-analysis; OR, odds ratio; PP, per-protocol; QOL, quality of life; RCTs, randomized controlled trials; RoB 2, risk-of-bias in randomized trials; SD, Standard deviation; SMBG, self-monitoring of the blood glucose; SR, systematic review; T2DM, type 2 diabetes mellitus; T1, usual care; T2, app (that is, care using an app to perform SMBG on a mobile phone or device); T3, app plus e-coaching (that is, care using an app and managed with additional e-coaching from healthcare providers).

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of Dongguk University, Gyeongju (DGU IRB 20190025-02).

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (No. HI19C0784)

ORCID iDs

Min Kyung Hyun

Jang Won Lee

Seung-Hyun Ko

Jin Seub Hwang

Supplemental Material

Supplemental material for this article is available online.

References

World Health Organization (WHO). Diabetes. WHO; 2020. https://www.who.int/news-room/fact-sheets/detail/diabetes

Zheng

Ley

FB.

Global aetiology and epidemiology of type 2 diabetes mellitus and its complications. Nat Rev Endocrinol. 2018;14(2):88-98. doi:10.1038/nrendo.2017.151

Dennis

Rhodes

, et al. Technology-assisted self-monitoring of lifestyle behaviors and health indicators in diabetes: qualitative study. JMIR Diabetes. 2020;5(3):e21183. doi:10.2196/21183

Tomah

Mahmoud

Mottalib

, et al. Frequency of self-monitoring of blood glucose in relation to weight loss and A1C during intensive multidisciplinary weight management in patients with type 2 diabetes and obesity. BMJ Open Diabetes Res Care. 2019;7(1):e000659. doi:10.1136/bmjdrc-2019-000659

Hoogendoorn

Schechter

Llabre

Walker

Gonzalez

JS.

Distress and type 2 diabetes self-care: putting the pieces together. Ann Behav Med. Published online September 11, 2020. doi:10.1093/abm/kaaa070

Winkley

Upsher

Stahl

, et al. Psychological interventions to improve self-management of type 1 and type 2 diabetes: a systematic review. Health Technol Assess. 2020;24(28):1-232. doi:10.3310/hta24280

American Diabetes Association. 7. Diabetes technology: standards of medical care in diabetes—2019. Diabetes Care. 2019;42(suppl 1):S71-S80. doi:10.2337/dc19-S007

Mao

Lin

Wen

Chen

Impact and efficacy of mobile health intervention in the management of diabetes and hypertension: a systematic review and meta-analysis. BMJ Open Diabetes Res Care. 2020;8(1):e001225. doi:10.1136/bmjdrc-2020-001225

Guo

Zhang

The efficacy of mobile phone apps for lifestyle modification in diabetes: systematic review and meta-analysis. JMIR MHealth UHealth. 2019;7(1):e12297. doi:10.2196/12297

10.

Yousuf

Reintjens

Slipszenko

, et al. Effectiveness of web-based personalised e-Coaching lifestyle interventions. Neth Heart J. 2019;27(1):24-29. doi:10.1007/s12471-018-1200-7

11.

Long

Ganetsky

Canamucio

Dicks

Heisler

Marcus

SC.

Effect of peer mentors in diabetes self-management vs usual care on outcomes in us veterans with type 2 diabetes: a randomized clinical trial. JAMA Netw Open. 2020;3(9):e2016369. doi:10.1001/jamanetworkopen.2020.16369

12.

McLeod

Stanley

Signal

, et al. Impact of a comprehensive digital health programme on HbA(1c) and weight after 12 months for people with diabetes and prediabetes: a randomised controlled trial. Diabetologia. 2020; 63(12):2559-2570.

13.

Saslow

Moskowitz

Mason

, et al. Intervention enhancement strategies among adults with type 2 diabetes in a very low-carbohydrate web-based program: evaluating the impact with a randomized trial. JMIR Diabetes. 2020;5(3):e15835.

14.

Veen

Bovendeert

JFM

Backx

FJG

Huisstede

BMA

. E-coaching: new future for cardiac rehabilitation? A systematic review. Patient Educ Couns. 2017;100(12):2218-2230. doi:10.1016/j.pec.2017.04.017

15.

Sterne

JAC

Savović

Page

, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019;366:l4898. doi:10.1136/bmj.l4898

16.

Harrer

Cuijpers

Furukawa

Ebert

. Doing meta-analysis in R: a hands-on guide. https://bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/2019.

17.

Han

King

Klonoff

, et al. Digital diabetes congress 2019. J Diabetes Sci Technol. 2019;13(5):979-989. doi:10.1177/1932296819872107

18.

Williams

Parsons

Dunseath

Stephens

Luzio

Owens

DR.

The impact of structured self-monitoring of blood glucose on glycaemic variability in non-insulin treated type 2 diabetes: The SMBG study, a 12-month randomised controlled trial. Diabetes Metab Syndr. 2020;14(2):101-106. doi:10.1016/j.dsx.2020.01.006

19.

Parsons

Luzio

Harvey

, et al. Effect of structured self-monitoring of blood glucose, with and without additional TeleCare support, on overall glycaemic control in non-insulin treated Type 2 diabetes: the SMBG Study, a 12-month randomized controlled trial. Diabet Med. 2019;36(5):578-590. doi:10.1111/dme.13899

20.

Vieira

Kobayasi

Pereira

Zaia

Sasaki

SU.

Impact of technology use in type 2 diabetes distress: a systematic review. World J Diabetes. 2020;11(10):459-467. doi:10.4239/wjd.v11.i10.459

21.

Sun

Malcolm

Wong

Shorr

Doyle

MA.

Improving glycemic control in adults and children with type 1 diabetes with the use of smartphone-based mobile applications: a systematic review. Can J Diabetes. 2019;43(1):51-8.e3. doi:10.1016/j.jcjd.2018.03.010

22.

IXY

Kee

JCY

Threapleton

, et al. Effectiveness of smartphone technologies on glycaemic control in patients with type 2 diabetes: systematic review with meta-analysis of 17 trials. Obes Rev. 2018;19(6):825-838. doi:10.1111/obr.12669

23.

Hou

Diao

Hewitt

Carter

Mobile phone applications and self-management of diabetes: a systematic review with meta-analysis, meta-regression of 21 randomized trials and GRADE. Diabetes Obes Metab. 2018;20(8):2009-2013 doi:10.1111/dom.13307

24.

Martos-Cabrera

Velando-Soriano

Pradas-Hernández

, et al. Smartphones and apps to control glycosylated hemoglobin (HbA1c) level in diabetes: a systematic review and meta-analysis. J Clin Med. 2020;9(3):693. doi:10.3390/jcm9030693

25.

Pirbaglou

Katz

Motamed

Pludwinski

Walker

Ritvo

Personal health coaching as a type 2 diabetes mellitus self-management strategy: a systematic review and meta-analysis of randomized controlled trials. Am J Health Promot. 2018;32(7):1613-1626. doi:10.1177/0890117118758234

26.

Adu

Malabu

Callander

Malau-Aduli

BS.

Considerations for the development of mobile phone apps to support diabetes self-management: systematic review. JMIR MHealth UHealth. 2018;6(6):e10115. doi:10.2196/10115

27.

Lum

Jimenez

Huang

, et al. Decision support and alerts of apps for self-management of blood glucose for type 2 diabetes. JAMA. 2019;321(15):1530-1532. doi:10.1001/jama.2019.1644

28.

Powers

Bardsley

Cypress

, et al. Diabetes self-management education and support in adults with type 2 diabetes: a consensus report of the American Diabetes Association, the Association of Diabetes Care and Education Specialists, the Academy of Nutrition and Dietetics, the American Academy of Family Physicians, the American Academy of PAs, the American Association of Nurse Practitioners, and the American Pharmacists Association. JAAPA. 2020;33(7):1-20. doi:10.1097/01.JAA.0000668828

29.

Huckvale

Prieto

Tilney

Benghozi

P-J

Car

Unaddressed privacy risks in accredited health and wellness apps: a cross-sectional systematic assessment. BMC Med. 2015;13:214.

30.

Zhu

Herrero

Georgiou

Deep learning for diabetes: a systematic review. IEEE J Biomed Health Inform. Published online November 24, 2020; doi:10.1109/jbhi.2020.3040225

31.

Yang

Wang

, et al. Self-management among type 2 diabetes patients via the WeChat application: a systematic review and meta-analysis. J Clin Pharm Ther. 2021;46(1):4-16. doi:10.1111/jcpt.13264

32.

Tejedor

Woldaregay

Godtliebsen

Reinforcement learning application in diabetes blood glucose control: a systematic review. Artif Intell Med. 2020;104:101836. doi:10.1016/j.artmed.2020.101836

33.

Tan

Kyaw

Smith

Tan

Tudor Car

Use of smartphones to detect diabetic retinopathy: scoping review and meta-analysis of diagnostic test accuracy studies. J Med Internet Res. 2020;22(5):e16658. doi:10.2196/16658

34.

Silva

Lee

Forbes

Demmer

Barton

Enticott

Use and performance of machine learning models for type 2 diabetes prediction in community settings: a systematic review and meta-analysis. Int J Med Inform. 2020;143:104268. doi:10.1016/j.ijmedinf.2020.104268

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.26 MB