Status
Conditions
About
Almost every mental health disorder is more prevalent in people with diabetes compared to people without diabetes. Is has been consistently shown, that people with diabetes and mental health issues (e.g. distress, depression, anxiety, eating disorder) have worse glycaemic control and worse quality of life. Thus, mental health issues can have a substantial impact on glycaemic control but are also an outcome of diabetes therapy itself. Achieving optimal quality of life is therefore considered an important goal of diabetes therapy. Research suggests a complex network between psychosocial factors and glycaemic control that can be highly variable between persons. It is therefore assumed that subgroups of people with diabetes exist that show different trajectories of glycaemic control and mental health.
Being in one subgroup can be more strongly linked to the likelihood of developing mental illness and complications relative to others. This suggests that it may be possible to treat individuals in different subgroups in a manner that optimises their treatment and health outcomes. Accurate characterisation of the heterogeneity in people with diabetes may help individualise care and improve outcomes. This calls for a more personalised approach that considers the idiosyncrasies of different subgroups, which in in line with the ADA/EASD's call for precision medicine in diabetes.
In the last 3 years, we have established the basis of a precision mental health approach in diabetes by employing n-of-1 analyses. By using the combination of ecological momentary assessment (EMA: a methodology that allows the repeated daily sampling of psychosocial factors in people's daily life) and continuous glucose monitoring (CGM), we were able to collect intensive longitudinal data per person. With this approach, we have analysed individual associations between glycaemic parameters and psychosocial variables and identified specific sources of diabetes distress per person.
Our objective is to make use of this n-of-1 approach and identify subgroups of people who share certain characteristics in their associations between glucose and psychosocial variables. We aim to feed these identified subgroups into the development of a digital twin for precision mental health in diabetes. With the digital twin, the aim is to have a representation of a real person based on the identified subgroups that allows to make simulations and predictions of the course of glycaemic control and mental health. These predictions can inform therapy and lead to more personalised, precise treatment decisions. Ultimately, the digital twin can serve as a clinical decision support system.
In order to achieve these objectives, we have already built a longitudinal panel of 1,400 participants that continuously complete EMA and questionnaire surveys along with measuring glucose levels using CGM. To map mental health, we have already conducted >700 clinical diagnostic interviews to diagnose mental disorders (e.g. depression, anxiety, eating disorders). For identifying subgroups and developing the digital twin, we will expand the data collection and aim at including a total of 1,809 participants. Further metabolic data (e.g. HbA1c) will be collected and the incidence and remission of mental disorders will be determined by repeated clinical diagnostic interviews.
By using machine learning, the complex networks between clinical, metabolic and psychosocial data will be analysed for different subgroups, leading to new insights that have the potential to shape future guidelines. This will then be used by the digital twin approach to make predictions on future glycaemic control and mental health, thereby directly translating the new scientific evidence into actionable treatment suggestions.
Full description
People with diabetes (PWD) have a higher risk of developing mental disorders. A systematic review showed that the odds ratio for 15 mental disorders is significantly higher in PWD (OR 1.3-3.9) [de Jonge et al. Diabetologia 2014;57:699-709]. A review on mental health in precision medicine indicated that mental health disorders can drive the incidence of diabetes and lead to a negative prognosis [Kremers et al. Diabetologia 2022;65:1895-1906]. Common mental health problems in diabetes include diabetes distress, (sub)clinical depression, anxiety, and disordered eating [Snoek et al. Lancet Diabetes Endocrinol 2015;3:450-460].
The ADA/EASD's consensus statement on precision medicine highlights mental health as "highly relevant" [Chung et al. Diabetologia 2020;63:1671-1693]. Moreover, mental health data should be "combined with other data … to improve the precision of clinical decision making." In our review, we showed that precision mental health in diabetes can be achieved by monitoring and combining glucose data with behavioural and mental health data [Hermanns et al. Diabetologia 2022;65:1883-1894].
A precision medicine approach to mental health requires an in-depth exploration of mental health subgroups. This involves identifying distinct patterns and subgroups who show different courses of mental health as well as understanding the contribution of glucose in relation to physio- and psychological factors. The interplay between mental health, CGM metrics, and behavioural factors is crucial for elucidating mental health trajectories, advancing personalised treatment strategies [Hermanns et al. Diabetologia 2022;65:1883-1894, Ehrmann et al. Diabetes Spectr 2021;34:149-155, Ehrmann et al. Diabetes Care 2022;45:1522-1531].
Based on our previous projects DIA-LINK1, DIA-LINK2 and PRO-MENTAL, we set the basis of such a precision mental health approach. By combining CGM with ecological momentary assessment (EMA), a methodology allowing the repeated daily assessment of mental health in daily life, we established an innovative approach that can be used for precision mental health [Ehrmann et al. Diabetes Care 2022;45:1522-1531]. In a recently accepted publication, we used this approach to identify individual drivers of diabetes distress with n-of-1 analyses [Ehrmann et al. Diabetologia 2024;67:2433-2445]. The study showed the existence of subgroups in which distress is differentially influenced either by the mental aspects (perception of glucose) or the actual CGM metrics. We also demonstrated that glycaemic control and psychosocial well-being at follow-up were differentially influenced depending on the individual drivers of distress.
To further analyse such subgroups and n-of-1 associations to provide personalised treatment, digital twins are a valuable tool for identifying relevant subgroups of PWD who share similar courses of glucose and mental health. In diabetes, a digital twin represents a virtual model of an individual, integrating data from various sources, including CGM, electronic health records, EMA, behaviour, and sensors. A digital twin is ideal to study highly dynamic networks between metabolic, behavioural and psychological factors [Cappon et al. J Diab Sci Technol 2024:19322968241262112, Mosquera-Lopez t al. Trends Endocrinol Metab 2024;35:549-557, Chu et al. Frontiers Medicine 2023;10:1178912]. However, a recent systematic review highlights that although psychosocial factors play a role in glucose control, none of the reviewed digital twins incorporated such factors. The authors conclude that "addressing this deficiency opens new perspectives and opportunities for improving the holistic management of T1D through more comprehensive and inclusive modeling approaches" [Cappon et al. J Diab Sci Technol 2024:19322968241262112].
With a digital twin, various treatment scenarios can be simulated. This helps diabetologists to develop precise treatment plans addressing both metabolic and psychological needs, ensuring more personalised interventions. In this project, we aim to develop a digital twin for precision mental health in diabetes.
Over the past years, we have established a theoretical framework [Hermanns et al. Diabetologia 2022;65:1883-1894, Ehrmann et al. Diabetes Spectr 2021;34:149-155] and laid the groundwork for a precision mental health approach in diabetes [Ehrmann et al. Diabetes Spectr 2021;34:149-155, Ehrmann et al. Diabetes Care 2022;45:1522-1531, Schmitt et al. Die Diabetologie 2024;20:861-872]. In the PRO-MENTAL study, we have recruited a longitudinal panel of 1,300 PWD who regularly complete a 14-day EMA period and questionnaires every 6 months, in addition to wearing a CGM.
The objective of the new project is to capitalise on the achievements of the PRO-MENTAL study, and to continue the data collection but expand the already rich data base by including the following data:
With the digital twin, we aim to simulate different courses of mental health (e.g. depression) and glucose control (e.g. time in range) as our two primary outcomes. Simulation will be based on different EMA and glucose profiles, PROs, inflammation status, medication, dietary behaviour and physical activity.
We also aim at simulating how mental health/glucose control will develop depending on changes in the aforementioned variables. For example, by categorising a person according to their EMA and CGM profiles (following the approach established by Ehrmann et al. [Diabetologia 2024;67:2433-2445]), we aim to predict whether this person is more or less likely to develop a mental disorder and how glucose control develops. The digital twin also allows the simulation of how different interventional measures may affect the overall course of mental health and glucose. This allows the identification and selection of the most promising treatment strategy to positively affect both mental health and glucose. The ultimate goal is to inform clinical decision-making with help of the digital twin, thereby translating previous research into an actionable approach towards precision medicine in diabetes.
The digital twin can be used to increase scientific understanding of the complex interplay between glucose and psychosocial factors. Previous research on this interplay has mainly focused on specific parts (see [Skinner et al. Diabet Med 2020;37:393-400, Speight et al. Diabet Med 2020;37: 483-492, Nefs et al. Diabet Med 2020;37:418-426]), but are lacking a holistic approach. A digital twin enables a more comprehensive analyses of subtypes in which psychosocial factors (e.g. stress, mood, diabetes distress, sleep, mental disorders) and different parameters of metabolic control (e.g. hypoglycaemia, time in range) influence each other over time. Thus, the use of machine learning/artificial intelligence has the potential to reveal complex networks by considering a multitude of influencing and confounding factors simultaneously.
Current work on precision mental health in diabetes is entirely preliminary scientific work with no translation to clinical practice. The development of a digital twin would be the first step of a precision mental health approach in diabetes. It could be used by diabetologists in clinical practice as a clinical decision support system:
The TwinPeaks project will be a prospective, longitudinal, observational, non-interventional study for which ethical approval will be obtained. The project will include 1,809 participants overall and at the end, approx. 6,000 person-years will be surveyed.
Sample size considerations:
With our clinical interviews, we focus on three key mental disorders: major depression, general anxiety disorder and any type of eating disorder. In order to have a sufficient sample size for the identification of subgroups, we aim at having at least 100 participants for each of the three mental disorders.
Based on current data from PRO-MENTAL, we estimate the 12-month prevalence of having any eating disorder in people with diabetes to be approx. 6.5%, the prevalence of Major Depression at 11.8% and the prevalence of general anxiety disorders (GAD) at 7.2%.
Thus, for achieving at least 100 participants for each of the three mental disorders, we have used the lowest prevalence of the three to calculate the sample size. Based on the prevalence of any eating disorder, we need 100/0.065=1,538 participants. Anticipating a 15% drop-out rate, we need to recruit 1,538/0.85 = 1,809 participants. This sample size will allow us to have a sufficient number of persons for specific disorders (e.g. dysthymia, bulimia, panic disorder).
Work packages:
The current project will build directly on the PRO-MENTAL study but will expand the previous work with five work packages.
Work package 1 - Data collection for the development of the digital twin: In the PRO-MENTAL study, we have already recruited 1,300 participants with type 1 and type 2 diabetes. The majority of participants have already completed a baseline clinical diagnostic interview and are completing an online survey with questionnaires as well as a 14-day EMA period every 6 months. With the TwinPeaks project, we will continue with the data collection but will recruit more participants and include new variables.
In order to develop the digital twin for mental health, additional data are needed. For the TwinPeaks project, we will expand the data collection and focus on four data sources:
Work package 2 - Data management and data preparation: The first step is to clean and prepare the plethora of data from different sources and in different formats for further statistical analysis. A key aspect will be the formatting of the data to account for the different time frames: minutes for CGM data, hours and days for EMA data, months for questionnaire data, and years (12 month and lifetime) for clinical diagnostic interviews. The data from several different sources will be made available through a parameterised interface for data collection and storage. Data-centric solutions will be used as the backend, such as employing a NoSQL database. The NoSQL database allows for the storage of unstructured data from various sources and can be horizontally scaled to handle a growing volume of data. A possible implementation is based on Apache Cassandra, known for its high availability and scalability. Suitable metadata and data standards are defined to ensure high comparability. The infrastructure for data collection and storage follows the recommendations of the German Ethics Council for the use of Big Data in medicine.
When the data infrastructure is established, different approaches of machine learning methods will be tested and compared. Special attention will be given to the use of the resulting models in a Federated Learning Setting. A Federated Learning Setting is a machine learning concept where models are trained on distributed data sources without the need for central data storage or exchange. This approach preserves data privacy and security by keeping the data local on respective devices or servers.This approach allows for innovative use of machine learning methods in medical practice while following the principles of the General Data Protection Regulation (GDPR) and the German Ethics Council.
Work package 3 - Development of the digital twin: Digital twins for subgroups are developed for identifying critical events and starting points for intervention from multimodal data using state-of-the-art machine learning and data science techniques. Ultimately, optimised approaches will be provided for the automatic identification of patterns, similarities in the data, and outliers.
To develop the digital twin, a hybrid modelling approach is used as the mathematical model. This hybrid model leverages two approaches: (1) Mechanistic models that are based on mechanistic physiological models of glucose regulation. These models are using ordinary differential equations (ODEs) to simulate the regulation of glucose in people with diabetes. (2) Long-Short Term Memory (LSTM) models that are exclusively data-driven and are more useful when mechanistic physiological models are not available: e.g. there is no physiological model for the impact of mood, diabetes distress or depression on glucose. Both approaches allow the modelling of dynamic systems, but LSTMs are powerful in data-driven, complex and non-linear scenarios, while ODEs perform well in accurately modelling physiological states. The LSTM can be used to model psychosocial factors that are difficult to formalise, whereas ODEs are used to model glucose control, which is well understood and described by physiological laws. Therefore, the Digital Twin uses ODEs to model glucose trajectories and LSTM to capture psychosocial influences. By combining these two methodologies, a digital twin can be constructed that accurately captures complex relationships from a large data set, while adhering to physiological constraints. For the twinning procedure, an appropriate technique will be chosen based on the data and data fit (e.g., Markov Chain Monte Carlo [MCMC]). Taken together, the combination of ODE and LSTM can be used to learn complex, non-linear relationships in the data and make predictions. MCMC can then be used to quantify the uncertainties of these predictions and calibrate the parameters of the ODE/LSTM model. By combining ODE/LSTM networks for modelling temporal dynamics and MCMC methods for uncertainty quantification and parameter calibration, the accuracy and robustness of digital twins can be improved.
With this approach, anomaly detection will be possible, for example, to identify specific behavioural patterns or characteristics associated with an increased risk for mental disorders or critical metabolic events.
Work package 4 - Validation process: Validation of the digital twin is a crucial element that will determine the future clinical usefulness. Thus, we will collect new data that serves as the ground truth for the predictions and simulations made by the digital twin.
Firstly, it is important to note the kind of predictions we are aiming for with the digital twin. For glucose, we aim at predicting individual glucose control on an aggregate level such as time in range, time below range, coefficient of variation (CV) or HbA1c within a 6-month timeframe. We do not aim at predicting the exact course of glucose over e.g. the next hours. For mental health, we also aim at predicting individual mental health within a 6-month timeframe on a more general level such the course of (sub)clinical mental disorders, questionnaire scores of e.g. well-being, time with distress or overall stress level.
For validation of the prediction of glycaemic control, CGM and HbA1c data will be continuously assessed every 6-months for the whole sample. Similarly, for the validation of the prediction of mental health, EMA will be continuously assessed every 6-months for the whole sample. For feasibility reasons, clinical diagnostic interviews will only be collected in a subset of the sample.
Sample size considerations for this subset are as follows: Usually, for machine learning models, the ratio between training and test data sets is 80:20 [Gholamy et al. Int J Intell Technol Appl Stat 2018;11:105-111]. Therefore, for validation purposes, we will randomly select 20% of our sample, 0.2 x 1,809 = 362 participants, for follow-up clinical interviews. For these 362 participants, clinical diagnostic interviews will be conducted.
Based on the validation data, the accuracy (e.g. sensitivity, specificity, positive/negative predictive value) and precision (e.g. difference between simulated time in range and actual time in range) of the predictions from the digital twin are assessed.
Work package 5 - Preliminary interface for clinical use: To further increase the translational aspect of the TwinPeaks project and to stimulate clinical application, it is planned to build a preliminary interface of the digital twin. In this interface, data of an individual patient can be entered who should be digitally twinned. The actual twinning procedure will then happen in the backend. The results of the simulations and predictions will be displayed in the interface. With this preliminary interface, a pilot version of a clinical decision support system will be developed to get first insights of the transferability of our approach to clinical practice.
Timelines + milestones:
Month 1-18: Work package 1 - Data collection for development of the digital twin Month 19-22: Work package 2 - Data management and data preparation Month 22-26: Work package 3 - Development of the digital twin Month 19-32: Work package 4 - Validation process Month 31-36: Work package 5 - Preliminary interface for clinical use Milestone 1: Completion of initial data collection (N=1,809) --> M18 Milestone 2: Finalisation of the digital twin --> M26 Milestone 3: Completion of collection of validation data --> M30 Milestone 4: Validation of the digital twin --> M32 Milestone 5: Finalisation of a preliminary interface --> M36
Enrollment
Sex
Ages
Volunteers
Inclusion criteria
Exclusion criteria
1,809 participants in 1 patient group
Loading...
Data sourced from clinicaltrials.gov
Clinical trials
Research sites
Resources
Legal