DCB Newsletter #1/25: INSIDE DCB – Interview with Lisa Koch
Dear Community,
We are thrilled to present you with the next episode of our series “INSIDE DCB” – this time, with an interview with Lisa Koch. Enjoy the read!
Since April 2024, Lisa Koch has been an assistant professor at the University and University Hospital of Bern as well as the DCB (Diabetes Center Berne). Using artificial intelligence, she and her team are dedicated to developing demonstrably safe, reliable, and effective data-driven tools to shape the future of diabetes care.
Prof Dr Koch, what is your research about?
My research is concerned with the development of artificial intelligence (AI) methods for medicine. I am particularly interested in how artificial intelligence can be used safely and reliably. For example: How can we explain in an understandable way how an AI makes important decisions? How do we ensure that AI performs just as well for real people as it does in studies? And could it be that the AI is penalising certain groups? If so, how can we recognise and improve this?
What motivates you to work in the field of diabetes?
Diabetes research offers enormous potential to improve the lives of millions of people. The use of artificial intelligence in diabetes technology makes it possible to develop personalised therapies that respond to individual needs. The field of diabetes is also of scientific interest to me because there are many different applications for artificial intelligence. These range, for example, from blood glucose management using wearable devices (wearables) to the detection of secondary diseases of diabetes, such as the early detection of diabetic retinopathy based on images of the back of the eye. There are still many unanswered questions in each of these areas.
What is your greatest challenge as a professor?
My daily life as a professor is very diverse. Even the actual research itself involves a variety of different tasks: I continuously develop my short-, medium- and long-term research vision, supervise students, read and write a lot, and exchange ideas in my international environment. This also includes responsibilities in academic self-governance and teaching. In addition, there are countless other exciting tasks that are important to me: I help organise scientific conferences and am regularly invited as an expert to review the work of other professionals (peer review), participate in panel discussions, or give lectures. I would say my biggest challenge is prioritising and managing my limited time. Fortunately, since I have a great deal of independence in organising my time, the responsibility lies entirely with me: I need to learn to occasionally say no to exciting opportunities.
What goals do you aim to achieve with your team and your lab, the MLM-Lab (Machine Learning in Medicine Lab)?
I have several goals that I would like to pursue with my team. On one hand, my core research objectives are close to my heart: I aim to make artificial intelligence effective and safe for use in medicine. Ultimately, I want to contribute to improved patient care, both generally and specifically for people with diabetes. It’s important to me that, within this domain, my team and I work on projects that we find personally exciting and that we are methodologically passionate about. Apart from the research content itself, I aim to promote skills in my team and thus offer my team members good prospects for a successful career in research or the private sector. For example, we invest a lot of time in scientific communication. It is also very important to me to create a healthy and positive environment in which all team members can also prioritise their private lives.
What projects are you currently researching and how can they make life easier for people with diabetes?
We are working on various methods to automatically analyse medical images and data from wearable devices such as blood glucose meters and fitness trackers. For example, we use machine learning to tailor physiological glucose models to individuals, and investigate how we can recognise and correct inaccuracies in AI models. These technologies are designed to support people with diabetes by enabling safe, personalised treatment.
How do you work with the DCB?
Our offices are located at the DCB, which leads to regular dialogue. In particular, my team benefits from the interdisciplinary research environment that the DCB and the University of Bern have created with their endowed professorships. The DCB also offers us active access to the network in the field of diabetes technology.
Where do you hope your field of research will be in five to ten years’ time? What is your vision?
Artificial intelligence will have a fundamental impact on many areas of our society in the coming years. My vision is that this change in medicine will be positive. With my research, I want to provide trustworthy technologies to facilitate this.
About Lisa Koch
Lisa Koch is an assistant professor and research group leader at the University and Inselspital Bern. Together with her team at the DCB, she researches and develops trustworthy diabetes care supported by artificial intelligence that brings real benefits to patients and the medical profession.
After completing a bachelor’s degree in electrical engineering (BSc) and a master’s degree in biomedical engineering (MSc) at ETH Zurich, Lisa Koch completed a PhD in machine learning for medical image analysis at Imperial College London. After a postdoctoral position at ETH Zurich, she joined the Swiss wearable medical device start-up Ava, where she eventually became head of the data science team. In this position, she recognised the need for demonstrably secure machine learning in healthcare. In 2021, she returned to academic research to conduct research on this topic as a group leader for machine learning in medical diagnostics in the Berens Lab at the Hertie Institute for AI in Brain Research at the University of Tübingen, Germany.
Thanks so much for reading and we’ll provide you with the next episode of this series soon!
This edition was authored by Sunjoy Mathieu , Communication Manager at DCB, and originally published on https://www.d-journal.ch in German and French.
This post was previously published in Linkedin. Click here to see the original publication.
More recent news
DCB Newsletter #5/26: DID YOU KNOW… CGM data could reveal more than glucose levels?
Dear DCB Community, Continuous glucose monitoring (CGM) has changed the way people manage diabetes. Glucose readings,...
DCB Newsletter #4/26: DID YOU KNOW… that the menstrual cycle can affect diabetes management every month?
Dear DCB Community, We often talk about diabetes management in terms of glucose levels, insulin dosing, carbohydrates,...
DCB Newsletter #3/26: DID YOU KNOW… why type 1 diabetes is rising and what projections to 2050 tell us?
Dear Community, We are happy to provide you with a new episode of our “DID YOU KNOW…” series – this time with a focus...
But could the same data tell us much more about a person's health than we currently realise?
A new study published in npj Digital Medicine suggests that looking beyond individual glucose readings could reveal patterns associated with diabetes-related health risks. The findings point towards an interesting possibility: CGM technology may one day help us understand glucose levels, and also identify individual risk profiles and support more personalised diabetes care.
What did the researchers discover?
The researchers analysed CGM data from 90,757 participants in the iCanNet cohort, including people with normoglycemia, prediabetes, type 1 diabetes and type 2 diabetes. Rather than looking at individual glucose values alone, they examined 48 different CGM-derived metrics, including glucose variability and patterns over time. Their analysis revealed distinct glucose patterns across the different groups. Combining several metrics also improved the researchers' ability to distinguish between them.
Particularly interesting were the findings among people living with type 2 diabetes.
The researchers identified four distinct glycemic phenotypes – different patterns of glucose levels and fluctuations – associated with varying risks of diabetes-related complications.
For example:
- Sustained high glucose levels were associated with a 59% higher relative risk of acute complications.
- High glucose variability was associated with a 75% higher relative risk of acute complications and a 41% higher relative risk of chronic complications.
The researchers also developed a Digital Glycemic Risk Score (DGRS), combining multiple CGM metrics to identify people with potentially elevated risks.
Among participants with type 2 diabetes who had already achieved a Time in Range of at least 70%, the model reclassified 25% into a high-risk category for acute complications and 22.3% into a high-risk category for chronic complications. These were classifications generated by the model, not confirmed predictions of complications occurring in these individuals.
The findings suggest that combining different glucose patterns may reveal information about health risks that commonly used CGM metrics alone do not fully capture.
From CGM data to new insights. The study analysed 48 CGM-derived metrics from 90,757 participants and identified distinct glucose patterns across the diabetes spectrum. Among people with type 2 diabetes, four glycemic phenotypes were associated with
What could this mean for diabetes care?
Today, CGM systems primarily help people understand what is happening with their glucose levels and respond accordingly. But the enormous amount of data these devices generate may hold additional information that we are only beginning to explore.
By analysing glucose patterns over time, researchers may be able to develop digital biomarkers – measurable indicators derived from digital health data – that could help characterise individual metabolic profiles and identify potential health risks.
In the future, this could support healthcare professionals in recognising people who may benefit from closer monitoring or more personalised treatment approaches.
The Nordic region remains the global hotspot for childhood‑onset T1D. Finland has historically reported the highest incidence worldwide – around 60 per 100,000 children per year – although this rate appears to have stabilised or slightly declined since the early 2000s. Sweden, Norway and Denmark follow closely, and the IDF Diabetes Atlas 11 still places Finland, Sweden and Norway among the global top ten for published incidence.
A 2024 systematic review of European paediatric data (1994–2022) confirmed that the Nordic countries retain the highest absolute incidence in Europe, even as their rate of increase has slowed relative to the rapid rise in previously low‑incidence Central and Eastern European countries. A Swedish nationwide cohort study of 2.9 million children found that the heritability of childhood‑onset T1D has remained high and stable at around 0.83 over 30 years, implying that known environmental factors explain less than 10% of the observed rise.
In France, Santé publique France reports an annual T1D incidence increase of approximately 4%. Paediatric incidence rose from 15.4 to 19.1 per 100,000 person‑years between 2010 and 2015, with an overall rate of 19.5 per 100,000 in 2015–2017 and substantial regional variation. Sardinia (Italy) now reports the world’s highest published paediatric incidence at 76.3 per 100,000 (2022), having overtaken Finland, while the United Kingdom and Ireland also remain in the global top ten.
Projections to 2050
Two major modelling efforts suggest the same direction of travel. The GBD 2021 study projects that age‑standardised global type 1 diabetes prevalence will rise by about 23.9% between 2021 and 2050, a smaller relative increase than for type 2 diabetes, but a substantial absolute rise given the lifelong management burden of T1D. The IDF Diabetes Atlas 11 projects that overall adult diabetes prevalence (type 1 and type 2) will grow from 589 million in 2024 to around 853 million by 2050, a 46% absolute increase. While most of this growth is driven by type 2 diabetes, T1D contributes meaningfully through sustained or rising incidence in high‑income countries and improved survival, which increases prevalence even if incidence stabilises.
For the U.S., a modelling exercise based on SEARCH data suggests that, under a linear‑increase scenario, the number of young people with T1D could nearly triple by 2050, with the steepest relative growth among Hispanic and Black children.
Why is incidence rising?
T1D is a polygenic autoimmune disease in which environmental triggers act on a susceptible genetic background. Several, often overlapping, hypotheses are under investigation:
- Reduced early‑life exposure to common infections (“hygiene hypothesis”), with mixed evidence.
- Rising childhood adiposity and insulin resistance, which may accelerate the clinical presentation of existing islet autoimmunity (“accelerator hypothesis”).
- Viral triggers such as enteroviruses and, more recently, SARS‑CoV‑2, where meta‑analyses report a 14–27% increase in paediatric T1D incidence during the pandemic.
- Dietary and microbiome‑related factors, including early cow’s‑milk exposure, gluten timing and shifts in the infant gut microbiome.
- Vitamin D status and latitude, endocrine‑disrupting chemicals, pollution and perinatal factors such as higher maternal age, higher birth weight and Caesarean delivery.
The Swedish cohort study mentioned above is a stark reminder of how much remains unexplained: even with population‑wide data on several environmental exposures, less than 10% of the rise in T1D incidence in Sweden could be attributed to known factors. This underscores the need for long‑term birth cohorts, population‑level islet‑autoantibody screening and mechanistic immunology.
Implications for the next decade
Three practical implications stand out for health systems, innovators and the diabetes community:
- Health systems need to prepare for more people living with type 1 diabetes. Even under conservative assumptions, the number of people living with type 1 diabetes will continue to rise. This has practical consequences for insulin and sensor supply, reimbursement for diabetes technology, specialist care and clinical-trial infrastructure.
- Early detection is becoming more relevant. Type 1 diabetes can often be detected before symptoms appear, through markers of early autoimmune activity. With the first therapy now approved to delay clinical onset in people at an early disease stage, screening is no longer only a research question. Programmes such as Fr1da in Germany and the D1Ce trial in Italy show how early identification could inform future intervention strategies.
- We still need to understand why incidence is rising. Many environmental factors are being studied, but no single explanation is sufficient. Until we better understand the drivers behind the increase, true primary prevention will remain out of reach. This makes high-quality registries, long-term cohort studies and coordinated research more important than ever.
At DCB, we see these trends as both a warning and a call to action.
The rising incidence of type 1 diabetes will require sustained collaboration between clinicians, researchers, industry, policymakers and crucially people living with diabetes and their families.
Our work in innovation, clinical research infrastructure and community engagement aims to contribute to that collective effort.
Selected sources
GBD 2021 Diabetes Collaborators. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050. The Lancet, 2023. https://www.thelancet.com/article/S0140-6736(23)01301-6/fulltext
International Diabetes Federation. IDF Diabetes Atlas, 11th edition. 2025. https://diabetesatlas.org/
Ogle GD, et al. Global type 1 diabetes prevalence, incidence, and mortality estimates 2025: Results from the IDF Atlas 11 and the T1D Index v3.0. Diabetes Research and Clinical Practice, 2025. https://www.sciencedirect.com/science/article/pii/S0168822725002918
Stahl-Pehe A, et al. Wave in Pediatric Type 1 Diabetes Incidence After the Emergence of COVID-19: Peak and Trough Patterns in German Youth — DPV Registry. Diabetes Care, 2025. https://diabetesjournals.org/care/article/48/4/e47/157803/
Buchmann M, et al. Type 1 diabetes incidence curves differ by age for girls and boys between 1996 and 2022: Results from the North Rhine-Westphalia Diabetes Registry. Diabetes Research and Clinical Practice, 2025. https://www.sciencedirect.com/science/article/pii/S0168822725000105
Nagl K, et al. Ongoing Increase in Incidence of Diabetes in Austrian Children and Adolescents 1989–2021. Pediatric Diabetes, 2023. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12017070/
Schoenle EJ, et al. Epidemiology of Type I diabetes mellitus in Switzerland: Steep rise in incidence in under-5 year old children in the past decade. Diabetologia, 2001. https://www.researchgate.net/publication/12017578
Santé publique France. Diabetes — Data. Updated 2024. https://www.santepubliquefrance.fr/en/diabete/data
Piffaretti C, et al. Trends in childhood type 1 diabetes incidence in France, 2010–2015. Diabetes Research and Clinical Practice, 2018. https://www.sciencedirect.com/science/article/abs/pii/S0168822718304662
Ruiz-Grao MC, et al. Trends in the Incidence of Type 1 Diabetes in European Children and Adolescents from 1994 to 2022: A Systematic Review and Meta-Analysis. Pediatric Diabetes, 2024. https://onlinelibrary.wiley.com/doi/10.1155/2024/2338922
Ji J, et al. Stable heritability of type 1 diabetes in a Swedish Nationwide Cohort Study. Nature Communications, 2024. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12174315/
D’Souza D, et al. Incidence of Diabetes in Children and Adolescents During the COVID-19 Pandemic: A Systematic Review and Meta-Analysis. JAMA Network Open, 2023. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10314307/
Quinn LM, et al. Environmental Determinants of Type 1 Diabetes: From Association to Proving Causality. Frontiers in Immunology, 2021. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8518604/
Esposito S, et al. Changes in the Global Epidemiology of Type 1 Diabetes in an Evolving Landscape of Environmental Factors. International Journal of Molecular Sciences, 2023. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10141720/
Rogers MAM, et al. Fluctuations in the incidence of type 1 diabetes in the United States from 2001 to 2015: a longitudinal study. BMC Medicine, 2017. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5688827/
Imperatore G, et al. Projections of Type 1 and Type 2 Diabetes Burden in the U.S. Population Aged <20 Years Through 2050. Diabetes Care, 2012. https://pmc.ncbi.nlm.nih.gov/articles/PMC3507562/
We hope this overview is helpful for your own work in diabetes care, research or advocacy. As always, we are grateful for your interest in and support of our activities and we look forward to continuing this conversation with you in future editions of our “DID YOU KNOW…” series.
An earlier version of this article was originally published in the private newsletter of our CEO, Derek Brandt here: Diabetes Technology News.
This post was previously published in Linkedin. Click here to see the original publication.
DCB Research AG
Freiburgstrasse 3
3010 Bern
Schweiz