Study design
The Trial to Slow the Progression of Diabetes (TRIPOD) is designed as a randomized, open-labelled, controlled, multi-center, superiority trial with three parallel arms. A total of 339 adults with sub-optimally controlled T2DM will be block randomized according to a 1:1:1 allocation ratio to the three arms. The study intervention will last for 24 months (104 weeks) and the primary outcome is the mean change in HbA1c level at Month 12 from baseline as measured by blood tests. This protocol conforms to the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines (Additional file 1 and Fig. 1).
Study setting and eligibility criteria
Adults (aged 21 to 70) with sub-optimally controlled T2DM from eleven Singapore Health Services (SingHealth) referral sites (two specialist diabetes centers and nine polyclinics) will refer themselves directly to the study team from Duke-NUS Medical School, who will manage and execute all study-related procedures, including the study visits (i.e., training sessions and follow-up assessments), at Duke-NUS Medical School or other study venues, if available. SingHealth is Singapore’s largest public healthcare group and is in a collaborative academic medicine partnership with Duke-NUS Medical School. Collectively, the SingHealth referral sites serve both low- and high-income patients, with all Singaporean citizens and permanent residents entitled to government subsidies. All eligibility criteria will be self-declared.
Inclusion criteria
- Diagnosed with T2DM with sub-optimal diabetes control as defined by a HbA1c level of between 7.5% and 11.0% (inclusive) at the most recent test taken within the past three calendar months.
- Not on insulin.
- On at least one oral glucose-lowering drug.
- Aged 21 to 70 at last birthday
- Singapore citizen or permanent resident.
- Able to read, write, and communicate in English.
- Own a personal smartphone and comfortable with using apps.
Exclusion criteria
- Pregnant or lactating.
- History of chronic kidney disease.
- Undergone dialysis for treatment of kidney failure.
- History of cardiovascular disease.
- History of stroke.
- History of blood diseases.
- History of chronic liver disease.
- Undergone chemotherapy, radiation therapy, or immunotherapy for cancer treatment in the past five years.
- Undergone blood transfusion in the past three months.
- Taking systemic corticosteroids.
- History of bariatric surgery or extensive bowel resection.
- Any major surgery in the past year.
- Unable to walk up 10 stair steps (individual steps, not floors) without stopping/difficulty.
Conditional eligibility criterion
As the DMP has a physical activity component, patients will complete the Physical Activity Readiness Questionnaire (PAR-Q) [23] to detect individuals who may be at risk if they increase their physical activity. If patients answer “Yes” to any PAR-Q question, they will be required to consult their physician and provide an approval note from the physician to be able to participate in the study.
Participant timeline and study arms
Interested patients will be directed to take an online screener questionnaire to assess their eligibility to join the study. All eligible patients will be invited to complete an online baseline questionnaire and attend a training session conducted by the study team where they will be briefed on the study and sign informed consent (i.e., enrolment). During the training session, participants will have their baseline anthropometric measurements taken by the study team, their blood drawn by trained phlebotomists for HbA1c tests, and their arm allocation revealed to them (i.e., recruitment). All patients will receive arm-specific participant booklets containing information about the study design, timeline, visits, recommended activities, interventions, and payouts. Participants in both the intervention arms will receive the DMP along with training on how to use the study devises and apps. In addition, participants in the DMP plus M-POWER Rewards arm will be given details about the M-POWER Rewards program and will sign a participant oath declaring that all activity data that they will provide solely represents their own efforts without attempts to cheat. Research has revealed that such oaths might reduce the probability that individuals will engage in dishonest behavior [24].
Participants will remain in their respective study arms throughout the 24-month (104-week) study. Follow-up questionnaires and assessments will be administered within 8-week window periods at Month 6 (Weeks 22-29), Month 12 (Weeks 48-55), Month 18 (Weeks 74-81), and Month 24 (Weeks 100-107). All follow-up assessments will be conducted at Duke-NUS Medical School (or other study venues, if available) where anthropometric measurements will be taken and blood samples collected for HbA1c tests.
Figure 2 illustrates the flow diagram for participants.
Arm 1: Usual care
Participants in the usual care arm will continue to receive usual care at their diabetes clinics throughout the study. In order to better identify the effect of the DMP and M-POWER Rewards interventions, we will present the same diabetes self-management recommendations to participants in all arms. Usual care arm participants will be encouraged to perform the following recommended activities meant to improve glycemic control during the course of the study:
- Learn more about diabetes self-management.
- Engage in at least 150 minutes of moderate-to-vigorous exercise per week.
- Monitor weight at least once a week and aim to achieve a healthy BMI (<23 kg/m2).
- Monitor blood glucose with a glucometer at least three times a week on separate days and aim to achieve post-meal (two hours after meals) readings within 4.0–10.0 mmol/L.
- Take diabetes medication as prescribed during usual care.
Participants in the usual care arm will also be encouraged to achieve two health goals by the end of the 24-month (104-week) study: (1) reduction of ≥1 percentage point in HbA1c level from baseline and (2) lose ≥5% of initial body weight at baseline for those with BMI ≥23 kg/m2 at baseline or, for those with BMI <23 kg/m2 at baseline, maintain healthy weight.
Arm 2: Diabetes Management Package (DMP)
Similar to the usual care arm, participants in the DMP intervention arm will continue to receive usual care at their diabetes clinics and will be encourage to achieve the same HbA1c level reduction and weight loss goals as those provided to the usual care arm. In addition, as part of the DMP, participants will receive the following apps, devices, and recommendations:
- M-POWER smartphone app: designed and created for this study, the M-POWER app will serve as a one-stop portal for participants to monitor their diabetes self-management activities and progress throughout the study. The app syncs and displays relevant data from all apps and study device accounts provided to participants. Participants will also be able to view study-relevant medical information, specifically HbA1c test results and weight measured at assessments, and personalized HbA1c level reduction and weight loss goals on the M-POWER app. Personal progress for all activities will be illustrated in graphical form for easy comprehension. Besides displaying individual progress, the app will compare personal results with the average results of all participants within their respective arm (descriptive norms), along with congratulatory or motivating messages for satisfactory or inadequate performance, respectively (injunctive norms). The participant’s current and best streaks (number of consecutive weeks where the target has been met) for all components will also be shown on the app.
- GlycoLeap digital lifestyle and education program: a 24-week, education and behavior change program designed and produced by KKT Technology Pte. Ltd. (Holmusk, Singapore) for patients with T2DM in Singapore. It is delivered through a smartphone app, Glyco, and comprises interactive, educational lessons and quizzes and human health coaching.
Lessons and quizzes: participants are given a diabetes self-management education curriculum in the form of 24 lessons accessible from the Glyco app. At the end of each lesson, participants will be presented with a quiz containing multiple choice questions to test their understanding and retention of lesson content. Quiz scores ≥80% confer a pass grade and the scores of the first passed attempts or the latest results will be displayed on the M-POWER app. Participants can revisit the health lessons and retake the quizzes at any time throughout the study.
Personal health coaching and support: participants will receive personalized advice, guidance, and positive motivation from a team of health coaches through the Glyco app during the 24-week GlycoLeap program. Having a personal coach introduces accountability, which is an important driver of behavior change. All health coaches are qualified and accredited dietitians in Singapore. To improve the value of coaching, health coaches will provide personalized coaching based on data that participants have input in the Glyco app (see below), including food logs. Participants can create food logs by using the Glyco app to take photographs of their meals. These photos will be reviewed and rated based on nutritional quality and portion sizing by the health coaches.
- Pedometer for physical activity tracking: participants will be provided with a Fitbit™ pedometer (Fitbit™ Inc., San Francisco, CA, USA), access to an anonymous Fitbit™ account that will be created for the study, and recommended to engage in at least 150 minutes of moderate-to-vigorous physical activity (MVPA) whilst targeting for at least 500 Fitbit™ active minutes weekly (see Discussion for explanation). Participants will also be encouraged to sync their Fitbit™ devices with their Fitbit™ study accounts wirelessly at least once a week. Active minutes from participants’ Fitbit™ accounts will be displayed on the M-POWER app.
- Weighing scale for weight monitoring: participants will be provided with a basic bathroom weighing scale with a digital display and recommended to weigh themselves at least once a week and log their weight measurements manually on the Glyco app. Weight logs that have been entered on the Glyco app, along with personalized weight goals, will be displayed on the M-POWER app.
- Glucometer for blood glucose monitoring: participants will be provided with an CONTOUR™ PLUS ONE glucometer (Ascensia Diabetes Care Holdings AG, Basel, Switzerland) and recommended to take at least three post-meal (two hours after meals) measurements on separate days a week and aim achieve readings within 4.0–10.0 mmol/L. Participants will be encouraged to upload their data by wirelessly syncing their glucometers with the Glyco app at least once a week for updated glucometer readings to be displayed on the M-POWER app.
- Pill tracker for medication adherence: participants will be provided with a RX Cap™ pill tracker (DoseSmart™ Inc., San Francisco, CA, USA), access to an anonymous DoseSmart™ account that will be created for the study, and recommended to use the pill tracker with the oral glucose-lowering drug that has been assigned for the study and to take their medication as prescribed. For medication assignment by the study team, Metformin will be the first choice as most T2DM patients on oral glucose-lowering drug will be expected to have been prescribed Metformin. If participants are not on Metformin, the oral glucose-lowering drug with the most frequent dosing will be assigned instead. Participants will be encouraged to upload their data by wirelessly syncing their pill trackers with their DoseSmart™ study account at least once a week for their medication adherence to be displayed on the M-POWER app.
Arm 3: DMP and M-POWER Rewards program (DMP plus M-POWER Rewards)
Participants in the DMP plus M-POWER intervention arm will also continue to receive usual care at their diabetes clinics, the same HbA1c reduction and weight loss goals as those provided to the other two arms. In addition to the DMP intervention, participants will be entitled to the M-POWER Rewards program. Our rewards strategy takes advantage of rebates to address loss aversion and a mix of near-term and longer term goals to address present bias. Participants can earn up to 1,000 M-Points (1 M-Point is equivalent to SGD1) over the two-year study period for performing specific care processes according to study recommendations (Table 1) and for achieving HbA1c and weight loss goals (Table 2). The rewards scheme offers more M-Points for meeting targets that are more challenging but more likely to reap better health benefits. M-Points can be redeemed in the form of financial rebates for approved non-inpatient, healthcare-related expenses incurred during the study period. Approved expenses include clinic and outpatient visits, laboratory tests, medications, medical devices (e.g., glucometers), consumables (e.g., glucose test strips), and other approved health-monitoring devices (e.g., physical activity trackers, weighing scales). To redeem their accrued M-Points, participants will submit photos of their receipts via the M-POWER app to the study team for approval. Participants may view their earned, redeemed, and balance M-Points on the M-POWER app and all redemptions will be issued in cash.
Table 1 M-POWER rewards scheme for performing recommended activities.
|
Category and App/ Device
|
Criteria
|
M-Points
|
Terms and Conditions
|
Award
|
Max over 104 weeks
|
Bonus
|
All Arm 3 participants will be credited with a bonus at the start of their study
|
8 bonus M-Points
|
8
|
Can only be claimed if participants earn ≥12 M-Points throughout the study
|
Health Literacy,
Glyco app
|
Complete the GlycoLeap lesson quizzes and achieve a score of 80% or higher to pass
|
1 M-Point per quiz passed
|
24
|
Quizzes are taken on the Glyco app and can be retaken until a passing grade is achieved
|
Weight Monitoring,
basic weighing scale and Glyco app
|
Weekly weigh-ins
|
1 M-Point per week
|
104
|
Weigh-ins must be self-reported weekly through the Glyco app
|
Physical Activity,
Fitbit pedometer and Fitbit app
|
Achieving Fitbit’s active minutes
|
3 M-Points if ≥500 Fitbit active minutes per week
2 M-Points if 400–499 Fitbit active minutes per week
|
312
|
Data must be uploaded via the Fitbit app
|
Blood Glucose Monitoring,
Ascensia Diabetes Care CONTOUR PLUS ONE glucometer and Glyco app
|
Weekly post-meal glucose measurements (taken two hours after each meal) need to be within range: 4.0–10.0 mmol/L
|
2 M-Points per week if ≥3 post-meal measurements (with each measurement taken on a separate day) per week between 4.0–10.0mmol/L
|
208
|
Data must be uploaded via the Glyco app
|
Medication Adherence,
DoseSmart pill tracker and DoseSmart app
|
Compliance is considered met when the medication is taken based on both prescribed number of times per day and prescribed time of the day (if applicable)
|
1 M-Point per week if 100% compliant to medication schedule within the week
|
104
|
Data must be uploaded via the DoseSmart app
|
Table 2 M-POWER Rewards scheme for achieving HbA1c and weight loss goals.
|
Bonus M-Points awarded at each follow-up assessment
|
Criteria for earning M-Points
|
If BMI ≥23 kg/m2 at baseline
|
If BMI <23 kg/m2 at baseline
|
60 bonus M-Points
|
Attain a ≥1.0 percentage point decrease in HbA1c level from baseline
or
Lose ≥5% of weight from baseline weight
|
Attain a ≥1.0 percentage point decrease in HbA1c level from baseline
|
30 bonus M-Points
|
Maintain HbA1c level or attain up to 1.0 percentage point decrease in HbA1c level from baseline
or
Maintain weight or lose up to 5% weight from baseline weight
|
Maintain HbA1c level or attain up to 1.0 percentage point decrease in HbA1c level from baseline
|
Outcome measures
Primary outcome
The primary outcome is mean change in HbA1c level at Month 12 from baseline. Decreases in HbA1c level have been associated with risk reductions in diabetes-related clinical complications or mortality and diabetes randomized trials frequently use mean reduction in HbA1c level as a study outcome [5, 25].
Secondary outcomes
- Mean change in HbA1c level at Months 6, 18, and 24 from baseline.
- Mean change in weight at Months 6, 12, 18, and 24 from baseline.
- Mean change in systolic and diastolic blood pressure at Months 6, 12, 18, and 24 from baseline.
- Proportion of participants that had insulin treatment initiated by their diabetes care physician by Months 6, 12, 18, and 24.
- Mean change in self-reported physical activity at Months 6, 12, 18, and 24 from baseline as assessed using the Global Physical Activity Questionnaire (GPAQ) [26].
- Mean change in self-reported weight monitoring at Months 6, 12, 18, and 24 from baseline as assessed by frequency of self-weighing.
- Mean change in self-reported blood glucose monitoring at Months 6, 12, 18, and 24 from baseline as assessed by frequency of self-testing.
- Mean change in self-reported medication adherence at Months 6, 12, 18, and 24 from baseline as assessed by frequency of taking diabetes medications as prescribed.
- Mean change in diabetes self-management at Months 6, 12, 18, and 24 from baseline as assessed using the Diabetes Self-Management Questionnaire (DSMQ) [27].
- Mean change in sleep quality at Months 6, 12, 18, and 24 from baseline as assessed using the Pittsburgh Sleep Quality Index (PSQI) [28].
- Mean change in work productivity and daily activity impairment at Months 6, 12, 18, and 24 from baseline as assessed using a modified Work Productivity and Activity Impairment: Specific Health Problem instrument (WPAI:SHP) [29].
- Mean change in health utility index at Months 6, 12, 18, and 24 from baseline as assessed using the 5-level EQ-5D instrument [30].
- ICERs based on HbA1c level will be determined by calculating the incremental cost per percentage point unit reduction in HbA1c level at Month 12 (primary endpoint).
- ICERs based on QALYs will be determined by converting ICER based on HbA1c level into incremental cost per QALY gained.
Sample size
Data from Bilger et al., 2017 [31] showed a standard deviation (SD) of HbA1c levels averaged across usual care and intervention groups being 1.2% at both baseline and after a 6-month intervention and a correlation of 0.4 between the two time points. We assume an SD of 1.2% in HbA1c at Month 12 (primary endpoint) and a correlation of 0.2 between baseline and Month 12 measurements. In order to detect a mean difference of 0.5% between the usual care and DMP arms and 0.5% between the DMP and DMP plus financial incentive arms, with Analysis of Covariance (ANCOVA) adjustment for baseline HbA1c and multiplicity adjustment by the closed testing procedure [32], a sample size of 90 per group is needed to have 80% power at two-sided 5% familywise type 1 error rate. To allow for 20% attrition at Month 12, we will recruit 113 per arm, or 339 in total.
Randomization
Participants will be randomized according to a 1:1:1 allocation ratio to the three arms, using stratified randomization with random permuted blocks within strata. The block size will be determined by a statistician generating the randomization list and will not be disclosed to the investigators and other study team members who have contact with study participants. Three stratification factors will be used: diabetes center (specialist clinic or polyclinic), gender, and dichotomized HbA1c level at baseline (7.5–9.2% or 9.3–11.0%). For allocation concealment, sequentially numbered, opaque, and sealed randomization envelopes will be used for the randomization assignment for all participants.
Participant recruitment, retention, withdrawal, discontinuation, and adverse event reporting
Participant recruitment
Eleven SingHealth referral sites (two specialist diabetes centers and nine polyclinics) will serve as referral sites for this study. Patients attending regular diabetes care at these referral sites will be recruited through various advertising avenues at the referral sites, newspaper advertising, and online sites. All recruitment materials will briefly explain the study design, list key eligibility requirements, and direct interested participants to the study website that contains additional information about the study. The website will also contain contact information of the study team and serve as a means for interested patients to contact the study team via a contact form. Interested patients may assess their eligibility by completing an online eligibility screener on the website and prospective participants will be requested to provide their contact details and consent to be contacted. If the patient did not report having a HbA1c level within the eligibility range or having taken a HbA1c test within the past three calendar months, the patient will be presented with the option to have the study team arrange for the patient to retake the screener at a future date after the next regular, clinic-scheduled HbA1c test has been taken. If a prospective participant answers “Yes” to any PAR-Q question [23], they will be required to provide an approval note from a physician to be able to participate in the study (conditional eligibility). Patients who answered “No” to all questions will be deemed physically fit to engage in physical activity as recommended in the DMP. The study team will contact all prospective participants to provide additional details about the study, answer any questions that patients might have, obtain photographs of the patients’ oral glucose-lowering drug to verify their medication and clinic, a unique personal link to complete the baseline questionnaire, and schedule a baseline training session. Prospective participants will attend training sessions at Duke-NUS Medical School (or other study venues, if available) where they will be briefed on the study, enrolled, provide baseline measurements, recruited, and taught how to use the study devices, apps, and the M-POWER Rewards program. As an incentive to join the study and for compensation for their time, successfully recruited participants will receive SGD10 in cash at the end of the training session. Figure 2 illustrates the flow diagram for participants.
Participant retention
Participants will receive SGD30 at each assessment for successfully completing assessments within their respective window periods. Participants in the usual care arm and the DMP incentive arm will receive SGD150 and SGD70, respectively, as forms of “fairness” payouts for not receiving the DMP and/or incentives entitled to other arms. Table 3 lists the participant payouts by arm.
Table 3 Participant payouts.
|
Type of payout
|
Condition
|
Payouts
|
Arm 1
(UC)
|
Arm 2
(DMP)
|
Arm 3
(DMP with
M-POWER)
|
Attending training session
|
Attend the training session
|
SGD10
|
SGD10
|
SGD10
|
Completing assessments
|
Complete assessments at Months 6,12,18, and 24 within their respective window periods
|
SGD30
per assessment
|
SGD30
per assessment
|
SGD30
per assessment
|
Study device data upload
|
Each device (pedometer, glucometer, and pill tracker) needs to contain at least one entry within the first seven calendar days from the first Monday (inclusive) of the month to receive payouts for that device for that month
The data must be uploaded successfully
|
NA
|
SGD2 per device per month*
|
SGD2 per device per month*
|
M-POWER Rewards
|
As per M-POWER Rewards scheme (Tables 1 and 2)
|
NA
|
NA
|
Max 1,000
M-Points (SGD1,000)
over 104 weeks
|
Fairness payout
|
Complete both Month 12 and Month 24 assessments within their respective window periods
Awarded upon completion of the Month 24 assessment
|
SGD150 for entire study†
|
SGD70 for entire study‡
|
NA
|
* SGD6 per month for three devices (pedometer, glucometer, and pill tracker) in total.
† For not receiving DMP, incentives for study device data upload, and M-POWER Rewards.
‡ For not receiving M-POWER Rewards.
|
Participant burden will be minimized by limiting participant in-person visits to five sessions (one baseline and four follow-up assessments), although additional visits may be necessary to disburse cash payouts or provide replacement devices. Malfunctioned or misplaced devices will be replaced at a subsidized rate (depending on budget) to enable continued participation of participants in the intervention arms.
Participant withdrawal and discontinuation
Participants are free to withdraw from the study at any time by informing the study team or the investigators on their decision to withdraw. Data that has been collected until the time of their withdrawal will be stored and analyzed. Participants may be discontinued from the study due to one or more of the following reasons:
- They become pregnant.
- Upon voluntarily informing their doctor that they are participating in this study, their doctor decides that continuing participation could be harmful and informs us in the process.
- Failure to follow the instructions of the study team or investigators.
Participants who develop any of the exclusion criteria 2–13 during the course of the study will not be withdrawn from the study unless they choose to withdraw voluntarily.
Adverse event reporting
Before commencing the study, participants will be informed that they should report the occurrence of any potential adverse events during the course of the study to the study team. The study team will ask participants about potential adverse events during the follow-up assessments. In the event that the study team is informed of any serious adverse event (SAE), the principal investigator will notify the SingHealth Centralized Institutional Review Board (CIRB) by submitting the SAE Reporting Form within the stipulated time frame. The notifying and reporting requirements depend on the severity, nature and causality of the event and specific procedures as delineated by SingHealth will be followed [33].
Data collection
Survey data
The screener questionnaire will be administered online via the study website. The baseline, Month 6, 12, 18, and 24 questionnaires will be provided to participants through unique, personal Qualtrics™ (Qualtrics International Inc., Provo, UT, USA) links sent via email. All questionnaires will contain questions to assess secondary outcomes, including the DSMQ, GPAQ, PSQI, a modified WPAI:SHP, and 5-level EQ-5D survey instruments, and the Brief Illness Perception Questionnaire (BIPQ), which has been validated in English for diabetes [34, 35]. Additionally, the baseline questionnaire includes questions on socioeconomic characteristics, while the follow-up questionnaires include questions on program evaluation.
Health outcomes
Measurements for health outcomes will be recorded at the training session (baseline) and at the Month 6, 12, 18, and 24 follow-up assessments conducted at Duke-NUS Medical School (or other study venues, if available). HbA1c tests will be conducted using the high-performance liquid chromatographic method via the VARIANT™ II TURBO Hemoglobin Testing System (Bio-Rad Laboratories Inc., Hercules, CA, USA) on the blood samples collected at all study visits. No blood samples will be stored after the HbA1c tests have been completed. With the exception of height (Seca 217 Mobile Stadiometer, Seca Deutschland), which will only be measured at baseline, weight (Seca 869 Mobile Floor Scale, Seca GmBH, Germany), and blood pressure (Welch-Allyn 420 Spot Vital Signs BP Monitor,Welch Allyn, NY, USA) will be measured at all study visits. For height and weight, duplicate measurements will be recorded, along with a third measurement if the first two readings are unequal. Duplicate measurements will also be recorded for systolic and diastolic blood pressure, unless the difference between consecutive readings is >10 mmHg and >5 mmHg, respectively, in which case up to four additional readings may be obtained.
Study devices and apps
Participants in the two intervention arms will wirelessly sync data from their pedometer, glucometer, and pill tracker with their anonymous Fitbit™, Glyco, and DoseSmart™ study accounts, respectively. Weight logs and quiz scores captured on the Glyco app will be stored in participants’ Glyco accounts. In order to display updated participant activities and M-Points on the M-POWER app, the study platform will pull data from all the anonymous study accounts on a daily basis via API and automatically calculate and award M-Points daily.
Data for sensitivity analysis and cost-effectiveness analysis
Medical data such as HbA1c test results taken during usual care at the clinics and diabetes medications prescribed and purchased, including insulin initiation dates, will be obtained from the medical records and pharmacy bills of all participants. HbA1c test results from these medical records will be used in sensitivity analyses. Billing data from inpatient, outpatient, pharmacy, and emergency departments will be collected to estimate net costs and cost offsets for the cost-effectiveness analysis. The costs of program delivery will be determined by capturing data on all relevant (non-sunk) labor costs, materials and supplies, contracted services (including costs for GlycoLeap), and M-POWER Rewards payouts.
Data management and monitoring
During the study, all data will be stored on secure servers at Duke-NUS Medical School and is backed up daily. All data containing personal identifiers will be encrypted with password-protection. All physical research data including consent forms, data entry forms, and password-protected portable hard disk drives containing backup data will be stored in locked cabinets at Duke-NUS Medical School. Only de-identified data will be shared with statisticians for data analysis and only the investigators and study team directly involved with the study will have access to the data. The research data will be kept for at least ten years after research completion and securely destroyed upon the publication of all pertinent research studies or reports.
This study is subject to reviews and/or audits by the SingHealth Research Quality Assurance Unit. These reviews or audits may be conducted routinely, triggered by the SingHealth CIRB, or upon an investigator’s request. No termination guidelines or rules have been defined for this study.
Survey data
Each completed screener questionnaire attempt will generate a unique ID that will be traceable to individual prospective participants. Similarly, each personal Qualtrics™ link for baseline and follow-up questionnaires will contain the participant’s unique study ID, ensuring that we can differentiate survey entries completed by different participants. Data validation will be implemented to ensure that responses are provided for all questions and that questionnaire entries are only flagged as complete after the participant reaches the end of the questionnaire. For the baseline questionnaire, if a participant completes the same questionnaire more than once by mistake, each entry will be recorded separately and timestamped, allowing us to assess the multiple entries. Only one entry from each participant will be used for data analysis. For the follow-up questionnaires, participants will only be permitted to complete each questionnaire once. Survey data from all questionnaires will be downloaded by the study team regularly and checked for completion and any errors or inconsistencies.
Health outcomes
Measurements taken at the study visits will be recorded on hardcopy printouts. All data will be converted into electronic data using double data entry by two different individuals and verified for consistency. Any discrepancies will be resolved by referring to source documents. At the study visits, phlebotomists will verify participants’ identities before drawing blood and will label blood samples with the participants’ study IDs only. HbA1c results will not contain any personal identifiers and will be shared with the study team after each batch of blood samples are processed.
Study devices and apps
To encourage participants in the incentive arms to engage with the study components, weekly inactivity notifications that list the components that they have not been engaging with will be emailed to participants. Participants who are not engaging with the study components will be contacted by the study team, who will offer their assistance and ensure that participants are not experiencing problems with the devices or apps. Study team members will also note down those who indicate that they do not wish to perform any activity(s) and will not contact these participants for the purpose of troubleshooting. At each follow-up assessment, study team members will also verify that participants are using the devices as instructed and respond to any queries that participants might have on usage of the devices or apps. Besides the aforementioned measures to reduce data loss, participants in both intervention arms will be incentivized to upload their data: participants will be given SGD2 per device per month for syncing their pedometers, glucometers, and pill trackers with their anonymous study accounts (Table 3).
As participants in the DMP plus M-POWER arm may be tempted to cheat in order to earn more M-Points, they will sign a participant oath at the end of the training session and each time they collect their M-Points redemption pay-outs. This might help to improve data validity as such oaths could reduce the likelihood of cheating [24]. The study team will also manage and back up data from the devices and apps and regularly check the data for inconsistencies that may suggest cheating.
Statistical methods
Preliminary descriptive analyses
Preliminary descriptive analysis and the patterns of missing data/drop-out rate in each arm will be examined. The statistician conducting the analysis will be blinded to the arm assignments when conducting preliminary descriptive analyses. The arm assignments will be disclosed only after the final database lock. A detailed statistical analysis plan will be prepared before the final database lock.
Primary analysis
Mean change in HbA1c level at Month 12 is the primary outcome. The primary analysis will be performed on a modified intention-to-treat population, including participants who have both baseline and Month 12 HbA1c level data. A linear regression model with HbA1c level at Month 12 as the dependent variable and an intercept, HbA1c level at baseline (continuous variable), indicator variables for participant who received DMP alone, participants who received DMP plus M-POWER, and indicator variables for stratification factors (gender and diabetes center) as independent variables will be performed. Using this model, a test for the global null hypothesis of all three arms having equal mean HbA1c level at Month 12 will be performed (Null Hypothesis 1: Coefficient of DMP plus M-POWER = Coefficient of DMP = 0), followed by tests for three pairwise hypotheses, comparing mean HbA1c level at Month 12 in DMP alone vs. usual care (Null Hypothesis 2: Coefficient of DMP = 0), DMP plus M-POWER vs. usual care (Null Hypothesis 3: Coefficient of DMP plus M-POWER = 0), and DMP plus M-POWER vs. DMP alone (Null Hypothesis 4: Coefficient of DMP plus M-POWER = Coefficient of DMP). We will also conduct a sensitivity analysis with further adjustment for dichotomized HbA1c levels at baseline (7.5–9.2% vs. 9.3–11.0%). The differences in the primary outcome between study arms will be presented along with corresponding 95% confidence intervals. Following the closed testing procedure [32] for controlling for multiple comparisons involving three groups, only if tests of both the global null hypothesis and a pairwise null hypothesis reach statistical significance at 0.05 level will the pairwise null hypothesis be rejected.
If there are substantially different drop-out rates or different drop-out patterns among the three arms, a general linear model for repeated measures will be performed for the primary analysis. This model simultaneously models HbA1c level at baseline, Month 6 and Month 12 as the dependent variables and includes interactions between indicator variables for DMP alone and Month 6 visit, DMP plus M-POWER and Month 6 visit, DMP alone and Month 12 visit, DMP plus M-POWER and Month 12 visit as independent variables. The model will also adjust for visit (indicator variable for the Month 6 visit and indicator variable for the Month 12 visit) and randomization stratification variables (gender and diabetes center). An unstructured matrix will be used to model the residual variance-covariance structure within participant. The model does not include main effect terms for the intervention variables, and thus constrains the estimated group means of baseline HbA1c levels to be identical across the three randomized groups (Fitzmaurice et al., 2011). This model specification helps control for the variation in baseline HbA1c level arising by chance among the randomized groups. Using this model, a test for the global null hypothesis of all three arms having equal mean HbA1c level at Month 12 will be performed (Null Hypothesis 1: Coefficient of interaction between indicator of DMP alone and indicator of Month 12 visit = Coefficient of interaction between indicator of DMP plus M-POWER and indicator of Month 12 visit = 0), followed by tests for three pairwise hypotheses, comparing mean HbA1c level at Month 12 in DMP alone vs. usual care (Null Hypothesis 2: Coefficient of interaction between indicator of DMP alone and indicator of Month 12 visit = 0), DMP plus M-POWER vs. usual care (Null Hypothesis 3: Coefficient of interaction between indicator of DMP plus M-POWER and indicator of Month 12 visit = 0), and DMP plus M-POWER vs. DMP alone (Null Hypothesis 4: Coefficient of interaction between indicator of DMP alone and indicator of Month 12 visit = Coefficient of interaction between indicator of DMP plus M-POWER and indicator of Month 12 visit). If the missing data patterns do not necessitate using a general linear model for repeated measures as a primary analysis, this analysis will be conducted as a sensitivity analysis.
Secondary effectiveness analyses
Quantitative outcomes
Secondary quantitative outcomes (weight, blood pressure, GPAQ total physical activity score, weight monitoring frequency, blood glucose monitoring frequency, diabetes medication adherence frequency, DSMQ sum score, global PSQI score, percent overall work impairment and percent activity impairment due to diabetes and related health problems using a modified WPAI:SHP, health utility index using 5-level EQ-5D ) will be analyzed using a similar strategy applied to the primary outcome. A linear regression model will be used to model the secondary quantitative outcome as the dependent variable and an intercept, the outcome at baseline (quantitative variable), indicator variables for participants who received DMP alone, participants who received DMP plus M-POWER, and indicator variables for stratification factors (gender, diabetes center and dichotomized HbA1c level at baseline) as independent variables will be performed. Additional analyses will also be performed to evaluate the intervention effects that account for insulin progression/medication changes and other potential effect modifiers, mediators, covariates, and program engagement metrics.
Binary outcomes
Secondary binary outcomes (e.g., proportion of participants that had insulin treatment initiated by their diabetes care physician) will be analysed using a generalized linear model with a log link function and binomial distribution (log-binomial regression model). The model will include as independent variables the follow variables: an intercept, indicator variables for participant who received DMP alone, participants who received DMP plus M-POWER, and indicator variables for stratification factors (gender, diabetes center and dichotomized HbA1c level at baseline).
Secondary cost-effectiveness analyses
The net cost and cost-effectiveness analysis will be performed from a third-party payer’s perspective using an Activity Based Costing approach which the PI has performed in many prior studies [36-38]. Using this approach, the costs of program delivery will be determined by capturing data on all non-sunk labor costs, materials and supplies, contracted services (including costs for GlycoLeap), and M-POWER Rewards payouts. Billing data from inpatient, outpatient, pharmacy, and emergency departments will be collected. We will compute the net costs of each arm and the incremental cost per unit reduction in HbA1c at Month 12 (primary endpoint) as compared to the next most costly intervention. We will also compute the incremental cost per QALY gained based on published studies/models that quantify the relationship between reductions in HbA1c at Month 12 and QALYs gained. Results will be compared to established benchmarks for cost-effectiveness and those of other interventions aimed at reducing HbA1c levels. As a secondary analysis, we will also quantify incremental costs per QALY gained based on the 5-level EQ-5D health utility index captured at Month 12 through questionnaires and assumptions about the duration of any benefits attained during the study. One-way and n-way sensitivity analyses and cost-effectiveness acceptability curves that graphically present the probability that each intervention arm is incrementally cost-effective (and potentially cost saving) for a range of willingness-to-pay metrics that a decision maker may consider will also be presented.