Kayo Waki
Associate professor
Make your next move.
Prepare your application
Review drafts before you send them.
Explore related searches
Articles (10)
Background A high intake of dietary fiber has been shown to improve glycemic control and decrease hyperinsulinemia in people living with type 2 diabetes (T2D). T2D patients in Japan consume less than the recommended amount of fiber. Based on findings from a formative study, we developed an artificial intelligence (AI)-powered mobile health (mHealth) intervention, FiberMore, that uses the theory of planned behavior to help T2D patients increase their dietary fiber intake by enhancing their perceived behavioral control and attitude toward fiber consumption. Objective We aimed to assess the efficacy of FiberMore in improving the dietary fiber intake of T2D patients by conducting a pilot randomized controlled trial. In addition, we want to explore the efficacy of FiberMore in reducing HbA 1c of T2D patients via improvement in dietary fiber intake. Methods This is a randomized, single-blinded, multicenter study targeting 80 T2D patients from 3 institutions in Japan with a 2-week run-in, a 12-week intervention, and a 12-week observation. The intervention group is given access to FiberMore throughout the 12-week intervention period. A core feature of FiberMore is AI-powered meal photo logging using a fine-tuned GPT-4o (OpenAI) model, which analyzes the nutrient content of meals and delivers personalized, real-time feedback on fiber content. In addition, FiberMore provides personalized fiber goal setting and supports participants in identifying barriers to increasing fiber intake, along with corresponding coping strategies (labeled as “solutions” to the participant), through an AI chatbot. The AI chatbot also assesses participants’ emotional attitudes toward eating more fiber and delivers relevant educational content on dietary fiber. The control group receives a sham intervention focused on salt reduction, consisting of educational content delivered at 3 time points during the intervention period and records their daily efforts in salt reduction in a diary. The 12-week intervention period will be followed by a 12-week observational period to investigate the sustainability of the intervention’s effects. The primary outcome is between-group difference in the change of dietary fiber intake at 12 weeks. The secondary outcomes include HbA1c, other clinical measures, measurements of behavior changes, and assessment of participants’ satisfaction and perceived usefulness of the intervention. Results Recruitment began on February 12, 2025, and ended on September 1, 2025. We anticipate that the intervention period will conclude in December 2025 and the observation period will conclude in March 2026. As of September 22, 2025, a total of 72 participants have been officially enrolled and randomized. Conclusions: There are currently no mHealth dietary interventions that specifically focus on increasing fiber intake in Japan, highlighting the novelty of this intervention. This trial will generate important evidence on the efficacy, feasibility, and safety of an AI-based mHealth intervention for enhancing dietary fiber intake and glycemic control in free-living individuals with T2D. Furthermore, as a pilot study, it will offer valuable insights into the development of AI as a promising tool for accurate, low-burden dietary assessment.
Year:
2025
Background Diabetes affects millions worldwide. Primary care physicians provide a significant portion of care, and they often struggle with selecting appropriate medications. Objective This study aimed to develop a model that accurately predicts what drug an endocrinologist would prescribe based on the current measurements. The goal was to create a system that would assist nonspecialists in choosing medications, thereby potentially improving diabetes treatment outcomes. Based on the performance of previous studies, we set a performance target of achieving a receiver operating characteristic area under the curve (ROC-AUC) above 0.95. Methods A transformer-based encoder-decoder model predicts whether 44 types of diabetes drugs will be prescribed. The model uses sequences of age, sex, history for 12 laboratory tests, and prescribed drug history as inputs. We assessed the model using the electronic health records from 7034 patients with diabetes seeing endocrinologists between 2012 and 2022 at the University of Tokyo Hospital. We assessed model performance trained on data subsets spanning different time periods (2, 5, and 10 years) using micro- and macro-averaged ROC-AUC on a hold-out test set comprising data solely from 2022. The model’s performance was compared against LightGBM. Results The model trained on data from the past 5 years (2017‐2021) yielded the best predictive performance, achieving a microaverage (95% CI) ROC-AUC of 0.993 (0.992-0.994) and a macroaverage (95% CI) ROC-AUC of 0.988 (0.980-0.993). The model achieved an ROC-AUC above 0.95 for 43 out of 44 drugs. These results surpassed the predefined performance target and outperformed both previous studies and the LightGBM model’s microaverage ROC-AUC of 0.988 (0.985-0.990) in terms of prediction accuracy. Furthermore, training the model with short-term data from the past 5 years yielded high accuracy compared to using data from the past 10 years, suggesting that learning from more recent prescribing patterns might be advantageous. Conclusions The proposed model demonstrates the feasibility of accurately predicting the next prescribed drugs. This model, trained from the past prescriptions of endocrinologists, has the potential to provide information that can assist nonspecialists in making diabetes-treatment decisions. Future studies will focus on incorporating important factors such as prescription contraindications and constraints to enhance safety, as well as leveraging large-scale clinical data across multiple hospitals to improve the generalizability of the model.
Year:
2025
Background A 12-week pilot of the StepAdd mobile health (mHealth) behavior change intervention based on social cognitive theory (SCT) saw an 86.7% increase in mean daily step counts among patients with type 2 diabetes. Due to the lack of exploration of theoretical implications in mHealth intervention studies, there is a need to understand the mechanism underlying the behavioral change to inform the future design of digital therapeutics. Objective This study aimed to examine the SCT drivers underlying the mean increase in exercise among Japanese patients with type 2 diabetes who participated in the StepAdd intervention. Methods This is a post hoc analysis of data collected in the single-arm pilot study of the 32 patients who completed the StepAdd intervention. The StepAdd app uses self-mastery and coping strategies to increase self-efficacy and thus increase walking. Self-mastery was measured by the goal completion (GC) rate, which is the percentage of days in which patients met these adapting goals. The use of coping strategies was measured by the strategy implementation (SI) rate, which is the percentage of days in which patients applied their selected coping strategies. We assessed correlations between GC, SI, and self-efficacy to increase walking via linear regression and analyzed relationships via structural equation modeling. Results We found statistically significant support for the SCT approach, including a correlation coefficient (ρ) of 0.649 between step increase and GC rate ( P <.001); a ρ of 0.497 between the coping SI rate and self-efficacy increase ( P =.004); a ρ of 0.446 between GC rate and self-mastery increase ( P =.01); and a ρ of 0.355 between self-regulation increase and step increase ( P =.046), giving us insight into why the behavior intervention succeeded. We also found significant correlations between self-efficacy for barriers and self-efficacy for task-specific behavior (ρ=0.358; P =.04), as well as self-regulation and self-efficacy for task-specific behavior (ρ=0.583; P <.001). However, a cross-lagged panel modeling analysis found no significant evidence that changes in self-efficacy preceded behavior changes in line with SCT. Conclusions Self-mastery and coping strategies contributed to the walking behavior change in StepAdd, supporting the SCT model of behavior change. Future research is needed to better understand the causal pathways proposed by SCT.
Year:
2025
UNSTRUCTURED Reduced or delayed medical follow-ups have been reported during the COVID-19 pandemic, which may lead to worsening clinical outcomes for patients with diabetes. The Japanese government granted special permission for medical institutions to use telephone consultations and other remote communication modes during the COVID-19 pandemic. We aimed to evaluate changes in the frequency of outpatient consultations, glycemic control, and renal function among patients with type 2 diabetes before and during the COVID-19 pandemic. This is a retrospective single-cohort study conducted in Tokyo, Japan, analyzing results for 3035 patients who visited the hospital regularly. We compared the frequency of outpatient consultations attended (both in person and via telemedicine phone consultation), glycated hemoglobin A1c (HbA1c), and estimated glomerular filtration rate (eGFR) among patients with type 2 diabetes mellitus during the 6 months from April 2020 to September 2020 (ie, during the COVID-19 pandemic) with those during the same period of the previous year, 2019, using Wilcoxon signed rank tests. We conducted a multivariate logistic regression analysis to identify factors related to the changes in glycemic control and eGFR. We also compared the changes in HbA1c and eGFR from 2019 to 2020 among telemedicine users and telemedicine nonusers using difference-in-differences design. The overall median number of outpatient consultations attended decreased significantly from 3 (IQR 2-3) in 2019 to 2 (IQR 2-3) in 2020 (P<.001). Median HbA1c levels deteriorated, though not to a clinically significant degree (6.90%, IQR 6.47%-7.39% vs 6.95%, IQR 6.47%-7.40%; P<.001). The decline in median eGFR was greater during the year 2019-2020 compared to the year 2018-2019 (–0.9 vs –0.5 mL/min/1.73 m2; P=.01). Changes in HbA1c and eGFR did not differ between patients who used telemedicine phone consultations and those who did not. Age and HbA1c level before the pandemic were positive predictors of worsening glycemic control during the COVID-19 pandemic, whereas the number of outpatient consultations attended was identified as a negative predictor of worsening glycemic control during the pandemic. The COVID-19 pandemic resulted in reduced attendance of outpatient consultations among patients with type 2 diabetes, and these patients also experienced deterioration in kidney function. Difference in consultation modality (in person or by phone) did not affect glycemic control and renal progression of the patients.
Year:
2025
BACKGROUND A strong association exists between sleep duration and glycemic control in patients with type 2 diabetes (T2D), yet convincing evidence of a causal link remains lacking. Improving sleep is increasingly emphasized in clinical T2D treatment guidance, highlighting the need for effective, scalable sleep interventions that can affordably serve large populations through mobile health (mHealth). OBJECTIVE This study aims to pilot an intervention that extends sleep duration by modifying bedtime behavior, assessing its efficacy among short-sleeping (≤6 hours per night) patients with T2D, and establishing robust evidence that extending sleep improves glycemic control. METHODS This randomized, single-blinded, multicenter study targets 70 patients with T2D from 9 institutions in Japan over a 12-week intervention period. The sleep extension intervention, BedTime, is developed using the Theory of Planned Behavior (TPB) and focuses on TPB’s constructs of perceived and actual behavioral control (ABC). The pilot intervention combines wearable actigraphy devices with SMS text messaging managed by human operators. Both the intervention and control groups will use an actigraphy device to record bedtime, sleep duration, and step count, while time in bed (TIB) will be assessed via sleep diaries. In addition, the intervention group will receive weekly bedtime goals, daily feedback on their bedtime performance relative to those goals, identify personal barriers to an earlier bedtime, and select strategies to overcome these barriers. The 12-week intervention period will be followed by a 12-week observational period to assess the sustainability of the intervention’s effects. The primary outcome is the between-group difference in the change in hemoglobin A 1c (HbA 1c ) at 12 weeks. Secondary outcomes include other health measures, sleep metrics (bedtime, TIB, sleep duration, total sleep time, and sleep quality), behavioral changes, and assessments of the intervention’s usability. The trial commenced on February 8, 2024, and is expected to conclude in February 2025. RESULTS Patient recruitment ended on August 29, 2024, with 70 participants enrolled. The intervention period concluded on December 6, 2024, and the observation period ended on February 26, 2025, with 70 participants completing the observation period. The data analysis is currently underway, and results are expected to be published in July 2025. CONCLUSIONS This trial will provide important evidence on the causal link between increased sleep duration and improved glycemic control in short-sleeping patients with T2D. It will also evaluate the efficacy of our bedtime behavior change intervention in extending sleep duration, initially piloted with human operators, with the goal of future implementation via an mHealth smartphone app. If proven effective, this intervention could be a key step toward integrating sleep-focused mHealth into the standard treatment for patients with T2D in Japan. CLINICALTRIAL Japan Registry of Clinical Trials jRCT1030230650; https://jrct.niph.go.jp/latest-detail/jRCT1030230650 INTERNATIONAL REGISTERED REPORT DERR1-10.2196/64023
Year:
2024
BACKGROUND The iAide2 (Tokai) physical activity monitoring system includes diverse measurements and wireless features useful to researchers. The iAide2’s sleep measurement capabilities have not been compared to validated sleep measurement standards in any published work. OBJECTIVE We aimed to assess the iAide2’s sleep duration and total sleep time (TST) measurement performance and perform calibration if needed. METHODS We performed free-living sleep monitoring in 6 convenience-sampled participants without known sleep disorders recruited from within the Waki DTx Laboratory at the Graduate School of Medicine, University of Tokyo. To assess free-living sleep, we validated the iAide2 against a second actigraph that was previously validated against polysomnography, the MotionWatch 8 (MW8; CamNtech Ltd). The participants wore both devices on the nondominant arm, with the MW8 closest to the hand, all day except when bathing. The MW8 and iAide2 assessments both used the MW8 EVENT-marker button to record bedtime and risetime. For the MW8, MotionWare Software (version 1.4.20; CamNtech Ltd) provided TST, and we calculated sleep duration from the sleep onset and sleep offset provided by the software. We used a similar process with the iAide2, using iAide2 software (version 7.0). We analyzed 64 nights and evaluated the agreement between the iAide2 and the MW8 for sleep duration and TST based on intraclass correlation coefficients (ICCs). RESULTS The absolute ICCs (2-way mixed effects, absolute agreement, single measurement) for sleep duration (0.69, 95% CI –0.07 to 0.91) and TST (0.56, 95% CI –0.07 to 0.82) were moderate. The consistency ICC (2-way mixed effects, consistency, single measurement) was excellent for sleep duration (0.91, 95% CI 0.86-0.95) and moderate for TST (0.78, 95% CI 0.67-0.86). We determined a simple calibration approach. After calibration, the ICCs improved to 0.96 (95% CI 0.94-0.98) for sleep duration and 0.82 (95% CI 0.71-0.88) for TST. The results were not sensitive to the specific participants included, with an ICC range of 0.96-0.97 for sleep duration and 0.79-0.87 for TST when applying our calibration equation to data removing one participant at a time and 0.96-0.97 for sleep duration and 0.79-0.86 for TST when recalibrating while removing one participant at a time. CONCLUSIONS The measurement errors of the uncalibrated iAide2 for both sleep duration and TST seem too large for them to be useful as absolute measurements, though they could be useful as relative measurements. The measurement errors after calibration are low, and the calibration approach is general and robust, validating the use of iAide2’s sleep measurement functions alongside its other features in physical activity research.
Year:
2024
Background Type 2 diabetes (T2D) is a significant global health challenge. Physicians need to assess whether future glycemic control will be poor on the current trajectory of usual care and usual-care treatment intensifications so that they can consider taking extra treatment measures to prevent poor outcomes. Predicting poor glycemic control from trends in hemoglobin A1c (HbA1c) levels is difficult due to the influence of seasonal fluctuations and other factors. Objective We sought to develop a model that accurately predicts poor glycemic control among patients with T2D receiving usual care. Methods Our machine learning model predicts poor glycemic control (HbA1c≥8%) using the transformer architecture, incorporating an attention mechanism to process irregularly spaced HbA1c time series and quantify temporal relationships of past HbA1c levels at each time point. We assessed the model using HbA1c levels from 7787 patients with T2D seeing specialist physicians at the University of Tokyo Hospital. The training data include instances of poor glycemic control occurring during usual care with usual-care treatment intensifications. We compared prediction accuracy, assessed with the area under the receiver operating characteristic curve, the area under the precision-recall curve, and the accuracy rate, to that of LightGBM. Results The area under the receiver operating characteristic curve, the area under the precision-recall curve, and the accuracy rate (95% confidence limits) of the proposed model were 0.925 (95% CI 0.923-0.928), 0.864 (95% CI 0.852-0.875), and 0.864 (95% CI 0.86-0.869), respectively. The proposed model achieved high prediction accuracy comparable to or surpassing LightGBM’s performance. The model prioritized the most recent HbA1c levels for predictions. Older HbA1c levels in patients with poor glycemic control were slightly more influential in predictions compared to patients with good glycemic control. Conclusions The proposed model accurately predicts poor glycemic control for patients with T2D receiving usual care, including patients receiving usual-care treatment intensifications, allowing physicians to identify cases warranting extraordinary treatment intensifications. If used by a nonspecialist, the model’s indication of likely future poor glycemic control may warrant a referral to a specialist. Future efforts could incorporate diverse and large-scale clinical data for improved accuracy.
Year:
2024
BACKGROUND Type 2 diabetes (T2D) is a significant global health challenge. Physicians need to assess whether future glycemic control will be poor on the current trajectory of usual care and usual-care treatment intensifications so that they can consider taking extra treatment measures to prevent poor outcomes. Predicting poor glycemic control from trends in hemoglobin A 1c (HbA 1c ) levels is difficult due to the influence of seasonal fluctuations and other factors. OBJECTIVE We sought to develop a model that accurately predicts poor glycemic control among patients with T2D receiving usual care. METHODS Our machine learning model predicts poor glycemic control (HbA 1c ≥8%) using the transformer architecture, incorporating an attention mechanism to process irregularly spaced HbA 1c time series and quantify temporal relationships of past HbA 1c levels at each time point. We assessed the model using HbA 1c levels from 7787 patients with T2D seeing specialist physicians at the University of Tokyo Hospital. The training data include instances of poor glycemic control occurring during usual care with usual-care treatment intensifications. We compared prediction accuracy, assessed with the area under the receiver operating characteristic curve, the area under the precision-recall curve, and the accuracy rate, to that of LightGBM. RESULTS The area under the receiver operating characteristic curve, the area under the precision-recall curve, and the accuracy rate (95% confidence limits) of the proposed model were 0.925 (95% CI 0.923-0.928), 0.864 (95% CI 0.852-0.875), and 0.864 (95% CI 0.86-0.869), respectively. The proposed model achieved high prediction accuracy comparable to or surpassing LightGBM’s performance. The model prioritized the most recent HbA 1c levels for predictions. Older HbA 1c levels in patients with poor glycemic control were slightly more influential in predictions compared to patients with good glycemic control. CONCLUSIONS The proposed model accurately predicts poor glycemic control for patients with T2D receiving usual care, including patients receiving usual-care treatment intensifications, allowing physicians to identify cases warranting extraordinary treatment intensifications. If used by a nonspecialist, the model’s indication of likely future poor glycemic control may warrant a referral to a specialist. Future efforts could incorporate diverse and large-scale clinical data for improved accuracy.
Year:
2024
We evaluated the effectiveness of a mobile health (mHealth) intervention for diabetic kidney disease patients by conducting a 12-month randomized controlled trial among 126 type 2 diabetes mellitus patients with moderately increased albuminuria (urinary albumin-to-creatinine ratio (UACR): 30-299 mg/g creatinine) recruited from eight clinical sites in Japan. Using a Theory of Planned Behavior (TPB) behavior change theory framework, the intervention provides patients detailed information in order to improve patient control over exercise and dietary behaviors. In addition to standard care, the intervention group received DialBetesPlus, a self-management support system allowing patients to monitor exercise, blood glucose, diet, blood pressure, and body weight via a smartphone application. The primary outcome, change in UACR after 12 months (used as a surrogate measure of renal function), was 28.8% better than the control group’s change ( P = 0.029). Secondary outcomes also improved in the intervention group, including a 0.32-point better change in HbA1c percentage ( P = 0.041). These improvements persisted when models were adjusted to account for the impacts of coadministration of drugs targeting albuminuria (GLP-1 receptor agonists, SGLT-2 inhibitors, ACE inhibitors, and ARBs) (UACR: −32.3% [95% CI: −49.2%, −9.8%] between-group difference in change, P = 0.008). Exploratory multivariate regression analysis suggests that the improvements were primarily due to levels of exercise. This is the first trial to show that a lifestyle intervention via mHealth achieved a clinically-significant improvement in moderately increased albuminuria.
Year:
2024
Background Increasing physical activity improves glycemic control in patients with type 2 diabetes (T2D). Mobile health (mHealth) interventions have been proven to increase exercise, but engagement often fades with time. As the use of health behavior theory in mHealth design can increase effectiveness, we developed StepAdd, an mHealth intervention based on the constructs of social cognitive theory (SCT). StepAdd improves exercise behavior self-efficacy and self-regulation through the use of goal-setting, barrier-identifying, and barrier-coping strategies, as well as automatic feedback functions. A single-arm pilot study of StepAdd among 33 patients with T2D showed a large increase in step count (mean change of 4714, SD 3638 daily steps or +86.7%), along with strong improvements in BMI (mean change of −0.3 kg/m2) and hemoglobin A1c level (mean change of −0.79 percentage points). Objective In this study, we aim to investigate the efficacy and safety of StepAdd, an mHealth exercise support system for patients with T2D, via a large, long, and controlled follow-up to the pilot study. Methods This is a randomized, open-label, multicenter study targeting 160 patients with T2D from 5 institutions in Japan with a 24-week intervention. The intervention group will record daily step counts, body weight, and blood pressure using the SCT-based mobile app, StepAdd, and receive feedback about these measurements. In addition, they will set weekly step count goals, identify personal barriers to walking, and define strategies to overcome these barriers. The control group will record daily step counts, body weight, and blood pressure using a non–SCT-based placebo app. Both groups will receive monthly consultations with a physician who will advise patients regarding lifestyle modifications and use of the app. The 24-week intervention period will be followed by a 12-week observational period to investigate the sustainability of the intervention’s effects. The primary outcome is between-group difference in the change in hemoglobin A1c values at 24 weeks. The secondary outcomes include other health measures, measurements of steps, measurements of other behavior changes, and assessments of app use. The trial began in January 2023 and is intended to be completed in December 2025. Results As of September 5, 2023, we had recruited 44 patients. We expect the trial to be completed by October 8, 2025, with the follow-up observation period being completed by December 31, 2025. Conclusions This trial will provide important evidence about the efficacy of an SCT-based mHealth intervention in improving physical activities and glycemic control in patients with T2D. If this study proves the intervention to be effective and safe, it could be a key step toward the integration of mHealth as part of the standard treatment received by patients with T2D in Japan. Trial Registration Japan Registry of Clinical Trials (JRCT) jRCT2032220603; https://rctportal.niph.go.jp/en/detail?trial_id=jRCT2032220603 International Registered Report Identifier (IRRID) DERR1-10.2196/53514
Year:
2024
Collaborators (12)
Seiji Nishikage
Kobe University
Ryo Suzuki
Professor & Chairperson
Tokyo Medical and Dental University
Masaomi Nangaku
Graduate School of Medicine, The University of Tokyo
Nobuhito Hirawa
Director / Associate Professor
Yokohama City University Medical Center
Yushi Hirota
Kobe University Graduate School of Medicine
Kazuhiko Ohe
Specially Appointed Professor
Juntendo University
Masao Toyoda
Tokai University School of Medicine
Kouichi Tamura
Yokohama City University Graduate School of Medicine
Masaki Igarashi
assistant professor
University of Tokyo Hospital
Hironori Waki
Professor
Akita University
Toshimasa Yamauchi
Professor
Graduate School of Medicine, The University of Tokyo
Koji Oba
University of Tokyo
Make your next move.
Prepare your application
Review drafts before you send them.