F049 AI in healthcare 3/5: Impacts in diabetes (Tadej Battelino)


“If you asked diabetologists 10 years ago about therapies available for individuals with diabetes, the reply would be considerably shorter compared to today,” says prof. Dr. Tadej Battelino, Head of Department of Pediatric endocrinology, diabetes and metabolism at the UMC - University Children’s Hospital Ljubljana, Slovenia.

Pharmacological treatments for diabetes are advancing alongside the development of new digital tools and digital therapeutics, which make diabetes management increasingly easier.

“If you are lucky enough that you don't need insulin for your diabetes, therapy is significantly easier. At the moment we have probably groups of drugs that really know well maybe 10 years,” says prof. dr. Battelino, who among other things serves on the editorial board of the European Journal of Endocrinology and Pediatric Diabetes, he is co-organizer of Advanced Technologies and Treatment of Diabetes Conference (ATTD), and also Chief Clinical at Dreamed Diabetes - Israeli based developer of personalized diabetes management solutions. DreamedDiabetes Advisor Pro decision support system received an FDA approval in 2018. 

Most widely know technical innovation for diabetes treatment is the artificial pancreas, aimed at those who are insulin dependent - people with diabetes type 1, and some with diabetes type 2. Those that still can produce some amount of insulin, take other medications.

These are:

According to WHO, the number of people with diabetes has risen from 108 million in 1980 to 422 million in 2014.  In 2016, diabetes was the direct cause of 1.6 million deaths and in 2012 high blood glucose was the cause of another 2.2 million deaths.

  • GLP-1 receptor agonists - a successful group of drugs, which, apart from regulating glucose through enhancement of glucose-dependent insulin secretion, also reduce weight.

  • SGLT2 Inhibitors are used when individual’s sugar is low than desired. They inhibit the reabsorption of glucose in the kidney. Excessive glucose is excreted, weight loss occurs and glucose is regulated. SGLT2 inhibitors have important effects on the kidney and heart in addition to the glucose regulation.This is much more successful for people with type 2 diabetes who do not need insulin. Long term outcomes are improved. An oral version of the drug recently became available.

  • More traditional drugs include metformin for which the exact mechanisms on the body are still unknown, but metformin has many beneficial effects on different organs. The drug is even popular in Silicon Valley as a longevity drug

  • Sulfonylureas are a very old very inexpensive drug directly stimulating release of insulin from pancreatic beta cells, and are useful for patients with some beta cell function. 

While diabetes in an increasing global problem, technological and medical solutions are in constant development, not only to ease chronic disease management but to also decrease comorbidities and complications. 

The impact of AI in diabetes

Various companies are designing decision support systems for patients and doctors. After years of research and development, in 2016 FDA approved MiniMed 670G system (Medtronic) a hybrid closed-looped system for glucose measurement and insulin delivery (so-called artificial pancreas). This is still a hybrid closed-loop system that was FDA and European CE Mark approved. “A part of the insulin adjustments are done by the algorithm by AI, particularly during the night to moderate basal insulin. Since a night is one third of the day this has long term implications and the burden of management is considerably decreased. We are now testing the next generation which also automates bolusing - additional doses of insulin needed based on an individual’s specific needs,” explains dr. Battelino. The devices in testing seem to improve so-called time in range, which is the new clinical outcome target. “Individuals are in range up to 80% of time. Never before were people with type 1 diabetes managed to keep time in range that high. This is important because the longer an individual is near normal glucose range, the better is the long-term outcome.”

Four research projects about closed loop systems are at the moment financed by the NIH - the American public agency for medical research.

All projects are researching different algorithms, explains dr. Battelino who is in the research group for one of the algorithms. He believes the final results, known in 2020, will go from research to practice soon. “Americans are really really wise because the best system will be available to their population quickly. Once these projects are finished it we will know the current best option. And in the US, when ever the National Institute of Health funds something, FDA is right behind it. They will know immediately what is safe and has the potential for FDA approval.”

In a broader care management picture, the future will bring an ever more digital approach to diabetes management, according to prof. dr. Battelino. “What we now believe is that we need digital clinics. In February 2020 in Madrid there will be the first consensus on the digital clinic, to set standards how AI could be used with a person,” he mentions. The system will be an advisor, tested in a double blind study comparing decisions for patients done by the system of a doctor. The study called Advice For You will compare results. “If the computer is as good as the doctor, as good is enough, then all of a sudden we have an additional diabetes specialist in our team. Obviously what the support system advises is just a suggestion, the decision still needs to be taken by a doctor or in the next stage by an individual with diabetes,” explains dr. Battelino.  

Glucose regulation improving due to new drugs and technology, dr. Battelino doubts we will ever come to a point where individuals wouldn’t need to think of their condition and heath.

“Everyone still needs to take health-related decisions every day. Which is a challenge today with a booming industry in nutrition.”

says dr. Battelino, who is sceptical about all the fashionable nutrition approaches and diets. As he warns, there is a lot of very poor science or even known science in nutrition, simply because a lot of money is involved. Food is not controlled the same way as medicines are. Medicines go through extreme stringency of clinical trials in their development. “Several big food producers promote their products claiming they were clinically validated, when in fact there was actually no research done, at least not in the way we understand clinical research. Whatever you decide to do and who to listen to, moderation is the key component to long term health. Food didn't change much in the last half a million years. The only real troubling trend is the abundance of simple carbs, which were non-existing several decades ago and are now very abundant.”

Eric Topol wrote an opinion piece in New York Times in March after participating in a nutrition study. The healthy study participants used glucose monitors to enable researchers to later analyse billions of data points to see what drove the glucose response to specific foods for each individual. Dr. Topol’s conclusion was, that coming up with a truly personalized diet today is virtually impossible. The clinical research he took part in, did spark an interest in healthy population to start measuring their glucose levels even in the absence of a clinical need  to do so. Again, dr. Battelino believes it would be an exaggeration in self-care to do so.

Health is described as wellbeing of your body and your mind. If you are preoccupied with health or this preoccupation becomes overwhelming, you lose your mind and soul part of your health. I would certainly not recommend that every healthy person would use a continuous glucose meter to ascertain basically what food is or is not appropriate. That would be a serious exaggeration in my view. What I believe is that generally if you discuss food with an individual you can always describe what food choices are advisable and what food choices actually are less recommended or should be for this or that reason. General recommendations still leave every individual a considerable choice in what she or he likes prefers,” he says.

Learn more by listening to the full discussion. 

Some questions addressed:

  • In your advice, how are people today supposed to orient around nutrition guidelines, given how science is changing?

  • Eric Topol wrote an opinion piece in New York TImes in March after participating in a nutrition, supported by glucose monitors and later machine learning, to analyse billions of data points to see what drove the glucose response to specific foods for each individual. His conclusion is, that coming up with a truly personalized diet today is virtually impossible. Will every individual concerned for her health use a blood glucose monitor?  

  • Diabetes is a complex disease. And the general knowledge is that patients with diabetes need to take insulin. It’s much more complicated than that - Just quickly: how many different drugs do people with diabetes need to take? 

  • How does technology complement the development of therapies? 

  • After years of research and development, in 2016 FDA approved MiniMed 670G system (Medtronic) a hybrid closed-looped system for glucose measurement and insulin delivery (so-called artificial pancreas). This is still a hybrid system, which means patients have to partially manage it. To start: for which patients is this technology appropriate for? 

  • What do individuals using the artificial pancreas still need to do manually? 

  • Four research projects about closed loop systems, financed by the NIH - the American pulbic agency for medical research are currently running.  Can you explain the complexity of insulin levels adjustment complexity and the role of technology in its management? What exactly is being researched in these four projects?  

  • How has the whole disease management system changed with the help of technology?

  • July this year Wired broke a story about two researchers that discovered disturbing vulnerabilities in Medtronic's popular MiniMed and MiniMed Paradigm insulin pump lines, that could be remotely targeted by attackers. Because the company did not react to fix the problem, the researchers developed a sort of universal remote for every one of these insulin pumps in the world, to show how dangerous this problem is. This brings up the question of trust in new devices and the problem of potential hacks. How did you receive this news? 

  • How do you see the problem of liability in technology and medical decision-making?