STUDY PROTOCOL

Study Protocol for the Effects of Artificial Intelligence(AI)-Supported Automated Nutritional Interventionon Glycemic Control in Patients with Type 2 DiabetesMellitus

Rie Oka . Akihiro Nomura . Ayaka Yasugi . Mitsuhiro Kometani .

Yuko Gondoh . Kenichi Yoshimura . Takashi Yoneda

Received: February 4, 2019 / Published online: March 15, 2019� The Author(s) 2019

ABSTRACT

Introduction: Nutritional intervention is effec-tive in improving glycemic control in patientswith type 2 diabetes but requires large inputs ofmanpower. Recent improvements in photoanalysis technology facilitated by artificialintelligence (AI) and remote communicationtechnologies have enabled automated evalua-tions of nutrient intakes. AI- and mobile-

supported nutritional intervention is expectedto be an alternative approach to conventionalin-person nutritional intervention, but with lesshuman resources, although supporting evi-dence is not yet complete. The aim of this studyis to test the hypothesis that AI-supportednutritional intervention is as efficacious as thein-person, face-to-face method in terms ofimproving glycemic control in patients withtype 2 diabetes.Methods: This is a multicenter, unblinded,parallel, randomized controlled study compar-ing the efficacy of AI-supported automatednutrition therapy with that of conventionalhuman nutrition therapy in patients with type2 diabetes. Patients with type 2 diabetes mainlycontrolled with diet are to be recruited andrandomly assigned to AI-supported nutritiontherapy (n = 50) and to human nutrition ther-apy (n = 50). Asken, a mobile application whosenutritional evaluation has been already vali-dated to that by the classical method ofweighted dietary records, has been speciallymodified for this study so that it follows therecommendations of Japan Diabetes Society(total energy restriction with proportion ofcarbohydrates to fat to protein of 50–60, 20, and20–30%, respectively).Planned Outcomes: The primary outcome isthe change in glycated hemoglobin levels frombaseline to 12 months, and this outcome is tobe compared between the two groups. The sec-ondary outcomes are changes in fasting plasma

Enhanced Digital Features To view enhanced digitalfeatures for this article go to https://doi.org/10.6084/m9.figshare.7798436

R. Oka (&) � A. Nomura � M. Kometani � Y. Gondoh� T. YonedaDepartment of Cardiovascular and InternalMedicine, Kanazawa University Graduate School ofMedical Science, Kanazawa, Japane-mail: ririoka@goo.jp

R. OkaDepartment of Internal Medicine, Hokuriku CentralHospital, Toyama, Japan

A. Nomura � K. YoshimuraInnovative Clinical Research Center (iCREK),Kanazawa University Hospital, Kanazawa, Japan

A. NomuraCureApp Institute, Karuizawa, Japan

A. YasugiAsken Medical Division, WIT Co., Ltd., Tokyo, Japan

T. YonedaInstitutes of Liberal Arts and Science, KanazawaUniversity, Kanazawa, Japan

Diabetes Ther (2019) 10:1151–1161

https://doi.org/10.1007/s13300-019-0595-5

glucose, plasma lipid profile, body weight, bodymass index, waist circumference, blood pres-sures, and urinary albumin excretion. Theresults of this randomized controlled trial willfill the gap between the demand for support ofAI in nutritional interventions and the scientificevidence on its efficacy.Trial Registration: UMIN000032231.

Keywords: Artificial intelligence (AI); Behaviorchange; Nutritional therapy; Remote healthcare

INTRODUCTION

Successful diabetes care requires not only theproper medications but also lifestyle modifica-tions in terms of diets and activity [1, 2]. Amongthe variety of weight loss methods and multi-faceted interventions on lifestyle, nutritionalintervention has the largest statistically signifi-cant effect on glycemic control in patients withtype 2 diabetes [3]. Due to the progressive nat-ure of type 2 diabetes [4], early nutritionalinterventions, such as before insulin injectionsare needed, may be very effective [5]. Priorstudies demonstrating significant decreases inglycosylated hemoglobin (HbA1c) during a1-year nutritional intervention involved fre-quent contact between patients and registereddietitians, with, for example, patients seenindividually on a weekly basis in the LookAHEAD trial [6] and on a once-monthly basis inthe MEDITA Randomized Trial [7]. Such intenseinterventions are difficult to maintain in clini-cal settings because both patients and providersreport lack of time [8] and healthcare resourcesare limited [9].

Recent technologies for remote healthcarehas a potential to reduce the burden of bothpatients and healthcare providers regardingdietary management [10, 11]. Patients can makedaily dietary records using mobile devices andcommunicate online with dieticians withoutthe need for hospital visits. Dietary self-moni-toring in itself is considered to be an effectivestrategy for effecting behavior change [12], andadherence to dietary self-monitoring is higherwhen an electronic diary is used than whenpaper records are kept with or without

feedbacks [13]. Moreover, improvements inphoto analysis technology that often includesartificial intelligence AI) has enabled the auto-matic identification of food items and evalua-tion of their nutritional values, by which real-time dietary feedbacks are provided [14]. AI-and mobile-supported nutritional interventionis expected to overcome time and location bar-riers in human interventions and boost themotivation of patients with type 2 diabetes toadhere to their diet.

With this background, our aim is todemonstrate the efficacy of automated nutri-tional intervention by use of a mobile phoneand AI on reducing glycemia in patients withtype 2 diabetes. In the study described here, weadopt a mobile application whose nutritionalevaluation has been already validated to that bythe classical method of weighted dietary records[15]. The hypothesis of this study is that AI-supported nutritional intervention is as effica-cious as in-person, face-to-face method inimproving glycemic control in patients withtype 2 diabetes.

METHODS

Study Design

This is a multicenter, unblinded, parallel, ran-domized controlled study aimed at comparingthe efficacy of AI-supported automated nutri-tion therapy with that of conventional humannutrition therapy in patients with type 2 dia-betes. The study protocol was approved by theInstitutional Review Board of KanazawaUniversity on April 13, 2018 (IRB no. 2623-3);all recruited patients are required to providewritten informed consent. This trial is registeredin the University Hospital Medical InformationNetwork (UMIN-CTR; https://www.umin.ac.jp/ctr; UMIN ID: UMIN000032231), and is beingconducted in accordance with the Declarationof Helsinki of 1964, as revised in 2013. Patientrecruitment commenced on April 13, 2018 andis scheduled to continue through to April 13,2020.

1152 Diabetes Ther (2019) 10:1151–1161

Study Setting and Participants

Participants are recruited from patients attend-ing the outpatient clinics of the Department ofDiabetes and Endocrinology at KanazawaUniversity Hospital, Hokuriku Central Hospital,and Houju Memorial Hospital. Those patientswho are obese or overweight and have type 2diabetes mainly controlled with diet areeligible.

Inclusion criteria are (1) age C 20 years; (2)body mass index (BMI) cutoff for Asiansof[ 23 kg/m2, which is the World HealthOrganization cutoff for Asians requiring publichealth action [16]; (3) physician diagnosis oftype 2 diabetes confirmed by clinical data (e.g.,documentation of fasting plasma glucose [FPG]of C 126 mg/dL, a positive oral glucose toler-ance test of[ 200 mg/dL, or HbA1c C 6.5%); (4)owning and using a mobile phone in their dailylives; and (5) able to understand study proce-dures and provide written informed consent.

Exclusion criteria include (1) use of insulinand/or glucagon-like peptide-1 within the past6 months; (2) use of agents causing weight loss;(3) HbA1c C 7.5%; (4) renal dysfunction (e.g.,urinary protein[1 g /g creatinine and/or anestimated glomerular filtration rate of\45 mL/min/1.73 m2); (5) uncontrolled thyroid disease;(6) uncontrolled hypertension (diastolic bloodpressure[100 mm Hg and/or systolic bloodpressure[160 mm Hg) and/or secondaryhypertension; (7) oral or injected steroid usewithin the past 3 months; (8) pregnancy orplanned pregnancy during the study period; (9)history of myocardial infarction, severe uncon-trolled heart failure, unstable angina, transientischemic attack, or stroke in the past 6 months;(10) receiving treatments for malignancies; (11)diagnosed eating disorder; and (12) medical orpsychological condition(s) which would com-promise the ability to meet protocol require-ments in the opinion of the investigator.

Interventions

Patients are allocated randomly to either AI-and mobile-supported nutrition therapy or in-person nutrition therapy by human dieticians

for 1 year. The principle of nutrition therapy inboth interventions is total energy restriction,based on the recommendations of Japan Dia-betes Society (JDS), Japan Atherosclerosis Soci-ety (JAS), and Japan Society for the Study ofObesity (JASSO); the ideal body weight (IBW) iscalculated as IBW = height (m)2 9 22 (kg/m2)[17, 18], and the total dietary energy (kcal/day)is estimated as 25–35 kcal/kg IBW/day. Physi-cians determine the total dietary caloric intakeaccording to the daily activity of each patient[17, 18]: 25–30 kcal/kg IBW for subjects withlow-level activity, 30–35 kcal/kg IBW for sub-jects with usual-level activity, and C 35 kcal/kgIBW for subjects with heavy-level activity. Therecommended macronutrient distribution is setto be 50–60% from carbohydrate sources, B 20%from protein sources, and 20–30% from fatsources [17].

Experimental Group: AI-SupportedNutrition Therapy

The mobile phone app used for this study iscalled ‘‘Asken’’ and is one of the most popularJapanese apps for behavior change amongindividuals aspiring to lose weight. The criteriaof its nutritional feedback system has beenspecially modified for this study so that it fol-lows the recommendations set by the JDS. Oneof the main features of this app is the avail-ability of various food-logging interfaces, mostnotably an AI-powered photo analysis systemlinked with its extensive database of over100,000 menus both on the market and home-made. Once a photo of a hamburger combo istaken, for example, the system instantly iden-tifies the frame of each item as well as its menuand serving amount (Fig. 1). After confirmationby the patient, during which process the patientcan re-select the correct type or brand of ham-burger from 82 types/brands in the database,the app calculates the energy and nutrientintakes based on the Standard Tables of FoodComposition in Japan 2015, seventh revisededition [19]. The validity and accuracy ofAsken’s estimated total energy intake andmacronutrients, such as carbohydrates, fat, andprotein, have been demonstrated in an earlier

Diabetes Ther (2019) 10:1151–1161 1153

study [15]. Following each food logging, Askeneventually shows how the user has satisfied his/her recommended range of daily intake of eachnutrient with a graph (Fig. 2) and delivers anindividualized message by AI based on theDietary Reference Intakes for Japanese 2015[20]. Because this whole process from menuidentification with photo analysis to generationof dietary feedbacks is fully unmanned andautomated, the patients can learn their detailednutritional conditions anywhere and anytime.These dietary messages and feedbacks are alsodelivered by a female Japanese character ‘‘Miki’’;there are more than 200,000 different patternsfor her wording and facial expressions. ‘‘Miki’’always remains compassionate and empatheticto the patients even when they eat a lot or skip arecording of food intake (Fig. 3).

The patients also have access to peer supportfrom other Asken users through the in-app diary

function. Weights and steps can be automati-cally recorded in the app by its data-sharingfunction with other apps, such as GoogleFit andApple Healthcare. Patients are instructed to logonto the Asken app and input meals and snacksat least once a week.

Comparison Group: Human NutritionTherapy

In Japan, nutrition therapy managed by humandietitians is covered by health insurance pro-grams for patients with diabetes. The principleof nutrition therapy of a comparison group isthe same as that of the experimental group,namely, total caloric restriction with macronu-trient distribution as recommended by the JDS[17]. At baseline, dietitians estimate the currentenergy and nutrient intakes on a face-to-face

Fig. 1 Artificial intelligence (AI)-powered photo analysis of a meal. Deep learning AI analyzes the photo of the entire mealand identifies the frame of each item as well as its menu and serving amount

1154 Diabetes Ther (2019) 10:1151–1161

interview. Participants and dietitians worktogether to design feasible strategies to improvedietary habits. Patients are provided with prin-ted materials for describing their plans andrecording their meals. After the first interview atbaseline, human dietitians continue to evaluatethe achievement status of the initial plans basedon the paper records and to provide individualfeedbacks once or more times within 3 months.Patients have a one-to-one meeting for C 20 miwith a human dietitian during each visit.

Outcomes

The primary outcome of this study is the changein HbA1c levels during 1 year of intervention,with the outcomes to be compared between thetwo groups. HbA1c is an accepted diagnosticparameter by which to monitor long-term gly-cemic control [21]. Secondary outcomes areabsolute changes in (1) FPG; (2) plasma lipidprofile, total cholesterol, triglycerides, high-density lipoproteins, and low-density lipopro-teins; (3) body weight, BMI, and waist circum-ference; (4) systolic and diastolic blood pressure;and (5) urinary albumin excretion. (6) Recorded

data on the Asken app (frequencies of log-ons,weight, daily intakes of total energy and nutri-ents) and (7) costs of the interventions in eco-nomic terms are also evaluated.

Blood Sampling and AnthropometricMeasurements

Venous blood samples are drawn from theantecubital vein after an overnight fast by atrained nurse and transported to the centrallaboratory of each hospital for analysis. HbA1cis measured using a high-performance liquidchromatography method. Lipids are measuredusing enzymatic analytical chemistry methods,and plasma glucose is assessed using the glucoseoxidase method. Resting blood pressure ismeasured in the sitting position with an auto-matic device after C 5 min of rest. Measure-ments of BMI and waist circumference areconducted according to published methods[22].

Fig. 2 Evaluation of energy and nutrient intakes. Askenshows how the patient has satisfied his/her recommendedrange of daily intake of each nutrient with a graph on thebasis of data input by the patient. Translation of text inarea A (from top to bottom): energy, protein, lipids,

carbohydrates, calcium, iron, vitamin A, vitamin E, vitaminB1, B2, vitamin C, dietary fiber, saturated fatty acids, andsalt. In area B, red boxes indicate excessive intake, greenboxes indicate appropriate intake, and blue boxes indicatedeficiencies

Diabetes Ther (2019) 10:1151–1161 1155

Concomitant Medication

In principle, treatments for diabetes, hyperten-sion, and dyslipidemia are not changed duringthe study period in order to avoid effects on theefficacy assessments of this study. If it isinevitable to add a new antidiabetic drug to thetreatment regimen, a full description of thechange and the reason and timing are describedin the case report form (CRF). If insulin therapyis started, the trial will be discontinued at thattime.

User Privacy

User privacy is preserved from two directions,the app and humans. First, the app is equippedwith security policy and measures according tothe JIPDEC PrivacyMark� system, which is awidespread certification for companies manag-ing personal information in Japan. The app’soperating company, WIT Co. Ltd., has a secureddata server in Japan to deal with all of the user

information safely and appropriately while dis-closing in the privacy policy what kinds of dataare collected for the service and how and bywhom the data can be used. In addition, theapp data are stored and managed anonymously,kept re-identifiable only by the limited numberof authorized researchers. Data managers storethe file for such re-identification separatelyfrom the study data itself with password-pro-tected access systems in an area with a limitedaccess.

Data Monitoring and Safety

Clinical data of the participants will be extrac-ted from electronic medical records andanonymously provided on an electronic CRF(eCRF) to the data manager. All entries in theCRFs should be backed up with the relevantsource data, including informed consent forms,medical history, laboratory data, and recordeddata on the app.

Monitoring of this research will be con-ducted by Innovative Clinical Research Center,Kanazawa University (iCREK). Monitors willensure that the investigational team complieswith the study protocol and that data andadverse events (AEs) are accurately recorded inthe CRFs. Due to the low risk character of theintervention in this study, only spontaneouslyreported AEs will be reported, and the investi-gators will immediately take appropriate mea-sures when AEs occur.

Participant Timeline

The schedule for enrollment, interventions, andassessments are shown in Fig. 4. Participants ineither intervention group are seen at least onceevery 3 months for the collection of study out-come data. Data collection is performed duringthe patient’s routine visits for diabetes checks.

Sample Sizes

The primary endpoint of this study is thechange of HbA1c from baseline to endpoint(12 months of intervention). Based on thedecrease in HbA1c in prior studies involving

Fig. 3 Dietary messages and feedbacks which appear onAsken. Feedbacks are delivered by a female Japanesecharacter ‘‘Miki,’’ who is always compassionate andempathetic with patients. This figure is taken from theUS version of Asken app only for reference; the study isconducted with the Japanese version

1156 Diabetes Ther (2019) 10:1151–1161

mobile-supported interventions [23], we expectthat the mean change in HbA1c level will be0.3% with a standard deviation of 0.3%. At asignificance level of 0.05 and a detection powerof 80%, the number of cases per group necessaryto reach significant difference in change inHbA1c between two groups is calculated to be74 cases. Assuming a 20% dropout rate, thetarget number of cases for each group is 50cases, with a total of 100 participants in thisstudy.

Randomization and Blinding

We use a web-based computerized blockedrandomization with blocks of six to assign par-ticipants to either two modes of nutritiontherapy. Block randomization is performed ineach center separately. Randomization is con-ducted centrally and independently of investi-gators, and participants in each group aretreated equally by their physicians. Althoughblinding of medical staff and participants to thestudy allocation is impossible, staff who areresponsible for data collection are blinded.

Statistical Analyses

The primary objective of this study is to confirmthat AI-supported, fully-automated nutritionalintervention is not be inferior to in-person

nutrition therapy by human dieticians withregard to HbA1c change during a 1-year period.Non-inferiority is considered confirmed withthe use of a prespecified non-inferiority marginof 0.2%. This is established if the lower limit ofthe 95% confidence interval of the difference inHbA1c change between the two groups (AI-supported group minus human-supportedgroup) is 0.2%. An independent t test is used tocompare the two groups for the change inHbA1c, which is calculated by subtracting thevalue of the baseline from the value after 1 yearof intervention. The changes in secondarycontinuous outcomes of this study are alsoanalyzed using an independent t test. Toaccount for non-dietary factors, such as theaddition of new drugs, compliance to medica-tions, and exercise habits, multiple regressionanalyses for changes in continuous outcomesare performed using these non-dietary factors ascovariates.

To estimate the trajectories of repeatedmeasurements for the study period, we usemixed-models. Differences between two groups(main effect) and the interaction between studygroups and study month are assessed for out-comes, including changes in HbA1c, weight,BMI, lipid parameters, and blood pressure. Allmixed models are adjusted for sex, age, baselineweight, and baseline HbA1c. Mixed modelsenable effective use of all available data evenwith missing values and, therefore, are

Fig. 4 Time schedule of the study. Patients who satisfy theeligibility criteria are randomized to two groups (1:1). Inboth groups, the principle of nutrition therapy is totalenergy restriction with macronutrient distributions

recommended by Japan Diabetes Society (JDS). Partici-pants are seen once every 3 months for the collection ofstudy outcome data. The observation period for bothgroups is 12 months

Diabetes Ther (2019) 10:1151–1161 1157

considered an intention-to-treat method ofanalysis. Analyses are conducted using SPSSsoftware version 24.0 for Windows (IBM Corp.,Armonk, NY, USA). All tests are conducted astwo-sided a = 0.05, and 95% confidence inter-vals are calculated.

Participant Withdrawal

Participants who discontinue using Asken or toseek advice according to the planned consulta-tions with dieticians will continue to have datacollected from their routine diabetes clinic vis-its, unless they specifically withdraw consentfor this study.

STRENGTHS AND LIMITATIONS

This study is a randomized controlled trial withthe aim to assess the efficacy of AI-supportedautomated nutrition intervention utilizingmobile technology on glycemic control inpatients with type 2 diabetes mellitus. Thestudy was designed to fill the information gapbetween the demand for mobile technologiespromoting behavior change in eating habitsand the scientific evidence supporting its effi-cacy compared to the conventional in-personinterventions by humans [24].

The accuracy and validity of dietary evalua-tion is one of the strengths of this study. Thedish-based nutritional evaluation performed bythe Asken app has been validated to theweighed dietary records, which is the gold-s-tandard method for individual level dietaryassessment [25]. The correlation coefficientsbetween those two methods have been reportedto be 0.87 for total energy and C 0.75 formacronutrients, which are comparable or evenhigher than the correlation coefficients betweenthe weighed dietary records and the food fre-quency questionnaire [15]. Because Asken hasdeveloped this database of [ 100,000 fooditems, which can even differentiate one kind offood at one shop from the similar kind atanother place, those who often eat out or eattake-outs can more easily gain access to accuratenutritional evaluations.

Personalization of goal-settings and feedbacksare two additional strengths of AI-supportednutritional intervention and key components forthe easy use of healthcare apps [26, 27]. Askenrequests each patient to set his/her personalizedgoal for weight and behavior, and this function isexpected to foster the intrinsic motivation, simi-lar to other apps for self-management of chronicdiseases [28, 29]. The feedbacks are tailor-madebased on the self-tracking data of each individual.Moreover, patients can customize food menus fortheir preferences, create their own items in theapp, and cite a routine menu repeatedly onAsken. In their study analyzing the database fromthe app ‘‘Lose it,’’ Serrano et al. reported thatgreater customization of the app is associatedwith a higher likelihood of engagement with theapp and positive health outcomes [28]. Personal-ization of goals and feedbacks realized by use of AIand mobile technology is in line with the ‘‘per-sonalized approach to diabetes management’’that was recently highlighted in a joint positionstatement of the American Diabetes Associationand the European Association for the Study ofDiabetes [31].

There are some limitations to this study. First,because participants are limited to those whoown and use a mobile phone, the results may notbe generalized to generations with relativelylower information and communication technol-ogy literacy. Second, people who adhere to thehealthcare apps are likely to have a deliberativestyle rather than intuitive style when makinghealth-related decisions. Third, due to theunblinded design, the patients in the AI armwould know that they are getting the experi-mental intervention, which may result in betteroutcomes in the AI intervention than the control.Finally, Asken collects data not only on daily dietsbut also activities such as step counts. Indepen-dent contribution of nutritional intervention is tobe evaluated by accounting for the effects of otherwearable devices monitoring activities.

ACKNOWLEDGEMENTS

The authors thank the participants of the study,Chiyoko Takata and Yukiko Takamiya for their

1158 Diabetes Ther (2019) 10:1151–1161

secretarial supports, and Ms. Eri Matsuzaki atKagawa Nutrition University for her commentson methods.

Confidentiality. The study ID number isassigned at the time of registration and is usedwhen handling samples and information rela-ted to the study to maintain anonymity. Usingthe study ID number as a key, electronic medi-cal records, eCRFs, administrative forms, andrecorded data on the app are linked. The liststhat link study ID numbers to identifyinginformation such as names or other personalidentifiers are stored in a separate, locked filewith password-protected access systems in anarea with a limited access. All records thatcontain names or other personal identifiers,such as informed consent forms, are storedseparately from the study records.

Funding. Sponsorship for this study andarticle processing charges are funded by PublicResearch of Kanazawa University Hospital.This study is also supported in part by Grants-in-Aid for Scientific Research from the Ministryof Education, Science, Sports, and Culture ofJapan (17K09199).

Authorship. All named authors meet theInternational Committee of Medical JournalEditors (ICMJE) criteria for authorship for thisarticle, take responsibility for the integrity ofthe work as a whole, and given their approvalfor this version to be published.

Author Contributions. MK and TY con-ceived the study. AN and KY provided advice onthe study design and statistical analysis. YGparticipated in the discussion of study designand wrote the study plan. RO drafted themanuscript and AY critically revised themanuscript. All authors read and approved thefinal manuscript.

Compliance with Ethics Guidelines. TheInstitutional Review Board of the KanazawaUniversity approved the study protocol on April13, 2018 (IRB no. 2623-3) in accordance withthe ethical standards of the Declaration of Hel-sinki and current legal regulations in Japan. All

procedures are in accordance with the ethicalstandards of the institutional and nationalcommittees responsible for human experimen-tation and with the Helsinki Declaration of1964, as revised in 2013. Substantial amend-ments of the study protocol must be approvedby the IRBs. The investigator provides eachpotential subject with an informed consentform that has been approved by the IRB,explains the contents of the study in detail bothin writing and orally, and obtains the subject’swritten consent of his/her own free will.

Disclosures. Rie Oka has received a speakerfee from Novartis. Takashi Yoneda has receiveda research grant from Daiichi-Sankyo andspeaker fees from Astellas, Novaritis, and Sanofi.Ayaka Yasugi is an employee of WIT Co., Ltd.,Japan. Akihiro Nomura, Mitsuhiro Kometani,Yuko Gondoh, and Kenichi Yoshimura havenothing to declare.

Data Availability. Data sharing is notapplicable to this article because it describes astudy protocol.

Open Access. This article is distributedunder the terms of the Creative CommonsAttribution-NonCommercial 4.0 InternationalLicense (http://creativecommons.org/licenses/by-nc/4.0/), which permits any non-commercial use, distribution, and reproductionin any medium, provided you give appropriatecredit to the original author(s) and the source,provide a link to the Creative Commons license,and indicate if changes were made.

REFERENCES

1. American Diabetes Association. 1. Strategies forimproving care. Diabetes Care. 2016;39[Suppl1]:S6–S12.

2. Knowler WC, Barrett-Connor E, Fowler SE, et al.Reduction in the incidence of type 2 diabetes withlifestyle intervention or metformin. N Engl J Med.2002;346:393–403.

3. Pastors JG, Warshaw H, Daly A, Franz M, KulkarniK. The evidence for the effectiveness of medical

Diabetes Ther (2019) 10:1151–1161 1159

nutrition therapy in diabetes management. Dia-betes Care. 2002;25:608–13.

4. Eckel RH, Kahn SE, Ferrannini E, et al. Obesity andtype 2 diabetes: what can be unified and what needsto be individualized? Diabetes Care.2011;34:1424–30.

5. Beebe C. Body weight issues in preventing andtreating type 2 diabetes. Diabetes Spectrum.2003;16:261–6.

6. Pi-Sunyer X, Blackburn G, Brancati FL, et al.Reduction in weight and cardiovascular disease riskfactors in individuals with type 2 diabetes: one-yearresults of the Look AHEAD trial. Diabetes Care.2007;30:1374–83.

7. Esposito K, Maiorino MI, Petrizzo M, Bellastella G,Giugliano D. The effects of a mediterranean diet onthe need for diabetes drugs and remission of newlydiagnosed type 2 diabetes: follow-up of a random-ized trial. Diabetes Care. 2014;37:1824–30.

8. Garden R. Expanding clinical empathy: an activistperspective. J Gen Intern Med. 2009;24:122–5.

9. Kluge EH. Resource allocation in healthcare:implications of models of medicine as a profession.MedGenMed. 2007;9:57.

10. Grock S, Ku JH, Kim J, Moin T. A review of tech-nology-assisted interventions for diabetes preven-tion. Curr Diab Rep. 2017;17:107.

11. Bonoto BC, de Araujo VE, Godoi IP, et al. Efficacy ofmobile apps to support the care of patients withdiabetes mellitus: a systematic review and meta-analysis of randomized controlled trials. JMIRMhealth Uhealth. 2017;5:e4.

12. Burke LE, Wang J, Sevick MA. Self-monitoring inweight loss: a systematic review of the literature.J Am Diet Assoc. 2011;111:92–102.

13. Burke LE, Conroy MB, Sereika SM, et al. The effectof electronic self-monitoring on weight loss anddietary intake: a randomized behavioral weight losstrial. Obesity (Silver Spring). 2011;19:338–44.

14. Boushey CJ, Spoden M, Zhu FM, Delp EJ, Kerr DA.New mobile methods for dietary assessment: reviewof image-assisted and image-based dietary assess-ment methods. Proc Nutr Soc. 2017;76:283–94.

15. Matsuzaki E, Michie M, Kawabata T. Validity ofnutrient intakes derived from an internet websitedish-based dietary record for self-management ofweight among Japanese women. Nutrients.2017;9:1058.

16. WHO Expert Consultation. Appropriate body-massindex for Asian populations and its implications forpolicy and intervention strategies. Lancet2004;363:157–163.

17. Haneda M, Noda M, Origasa H, et al. Japaneseclinical practice guideline for diabetes 2016. Dia-betol Int. 2018;9:1–45.

18. Teramoto T, Sasaki J, Ueshima H, et al. Treatment—therapeutic lifestyle modification. J AtherosclerThromb. 2008;15:109–15.

19. Standard Tables of Food Composition in Japan.2015. https://www.mext.go.jp/en/policy/science_technology/policy/title01/detail01/1374030.htm.Accessed Feb 2019.

20. Ministry of Health, Labour and Welfare. Dietaryreference intakes for Japanese. Daiichi Shuppan:Tokyo; 2015.

21. American Diabetes Association. 6. Glycemic targets:standards of medical care in diabetes—2019. Dia-betes Care. 2019;42:S61–S70.

22. Oka R, Miura K, Sakurai M, et al. Comparison ofwaist circumference with body mass index for pre-dicting abdominal adipose tissue. Diabetes Res ClinPract. 2009;83:100–5.

23. Free C, Phillips G, Galli L, et al. The effectiveness ofmobile-health technology-based health behaviourchange or disease management interventions forhealth care consumers: a systematic review. PLoSMed. 2013;10:e1001362.

24. Wang Y, Xue H, Huang Y, Huang L, Zhang D. Asystematic review of application and effectivenessof mHealth interventions for obesity and diabetestreatment and self-management. Adv Nutr.2017;8:449–62.

25 Gibson R. Principles of nutritional assessment.Oxford University Press: Auckland; 2005

26. Lentferink AJ, Oldenhuis HK, de Groot M, Polstra L,Velthuijsen H, van Gemert-Pijnen JE. Key compo-nents in eHealth interventions combining self-tracking and persuasive eCoaching to promote ahealthier lifestyle: a scoping review. J Med InternetRes. 2017;19:e277.

27. Zhao J, Freeman B, Li M. Can mobile phone appsinfluence people’s health behavior change? Anevidence review. J Med Internet Res. 2016;18:e287.

28. Waterlander W, Whittaker R, McRobbie H, et al.Development of an evidence-based mHealth weightmanagement program using a formative researchprocess. JMIR Mhealth Uhealth. 2014;2:e18.

1160 Diabetes Ther (2019) 10:1151–1161

29. Ramanathan N, Swendeman D, Comulada WS,Estrin D, Rotheram-Borus MJ. Identifying prefer-ences for mobile health applications for self-moni-toring and self-management: focus group findingsfrom HIV-positive persons and young mothers. Int JMed Inform. 2013;82:e38–46.

30. Serrano KJ, Yu M, Coa KI, Collins LM, Atienza AA.Mining health app data to find more and less

successful weight loss subgroups. J Med InternetRes. 2016;18:e154.

31. Inzucchi SE, Bergenstal RM, Buse JB, et al. Man-agement of hyperglycemia in type 2 diabetes: apatient-centered approach. Position statement ofthe American Diabetes Association (ADA) and theEuropean Association for the Study of Diabetes(EASD). Diabetes Care 2012; 35:1364–79.

Diabetes Ther (2019) 10:1151–1161 1161