A diagnosis of diabetes brings a lifetime of medication, lifestyle alterations, careful management of blood glucose levels, and the long-term risk of life-threatening complications. But for many, this disease is preventable and, importantly, predictable. Pre-diabetics can use lifestyle interventions to alter the course of their progression to avoid developing diabetes. Armed with better predictions on who to target, health authorities and clinicians can focus their limited resources on those in the greatest need. 

A life-long disorder

Diabetes is a complex, lifelong disease resulting in an inability to maintain a normal blood glucose level. This is characterized by a propensity for hyperglycemia (high blood sugar) or hypoglycemia (low blood sugar). The disease is categorized into two main types. Type 1 diabetes, mediated by autoimmunity against the insulin-producing beta cells, normally develops in childhood, whereas type 2 diabetes develops in adulthood. 

Type 2 diabetes is much more common than type 1, comprising around 90% of diabetes cases worldwide. While there are many risk factors for type 2 diabetes, including genetics, the disease is heavily associated with obesity and physical inactivity. Maintaining a healthy body weight, regular exercise and good nutrition are ways to avert or delay type 2 diabetes. With 422 million people suffering from diabetes in 2014, type 2 diabetes constitutes a large-scale public health concern.

Researchers are now looking into ways to categorize and predict groups of people who are on a trajectory to develop diabetes. They are using an arsenal of technological advances, such as machine learning, to carry out investigations to identify prodromes of the disease.

Doing more with less

One research project, published in Scientific Reports, aims to develop a public health tool for Qataris to gauge their personal risk of diabetes¹. Laith Abu-Raddad of Weill Cornell Medicine Qatar, is leading an international research collaboration that uses mathematical modelling to simulate a cohort of Qataris and builds on this to develop a public health risk factor score that can predict individuals’ diabetes risk. 

Abu-Raddad says that the increase in computational power afforded by technological advances means that more countries can avail themselves of the benefits of data. In wealthy countries such as the United States, he says, primary data is more easily accessible. The US conducts epidemiological surveys on diabetes once every two years, providing a wealth of data to fuel the development of a national diabetes risk score tool that takes into account how predictive variables change over time.

This isn’t an option for many countries. Less wealthy nations may have more infrequent surveys or use different methodologies that make the collection of empirical data difficult. In this sense, the rise of mathematical modelling has made good science more accessible. “Technology provides us a very inexpensive tool that negates the need for surveys,” says Abu-Raddad.

And the technology works. When Abu-Raddad’s team compared the scores of their simulated data to those of the primary data collected from Qatari subjects, the scores aligned “very well,” he says.

Abu-Raddad goes on to say that a primary goal of predicting diabetes risk is to build public awareness around what individuals can do to reduce their chances of developing diabetes. “It’s about building self-awareness and the need to change behavior,” he says. In countries like the United States and the United Kingdom, online tools exist where anyone can plug in very simple metrics, and instantly retrieve their personal diabetes risk score. “This is what we plan to do in Qatar,” he says. “It empowers people at an individual level to know their risks, which will hopefully lead to a change in behavior and create more awareness about diabetes.”

More than just glucose

Another international team, led by Robert Wagner of the University of Tübingen, Germany, says that we need to fundamentally change the definition of what we classify as pre-diabetes to better predict the risk of diabetes. “The current definition of pre-diabetes is only glucose-based,” explains Wagner. In his team’s paper, published in Nature Medicine, they describe how pre-diabetes’ current definition doesn’t reflect future metabolic trajectories, nor many of the variables that truly predict how an individual’s health might progress from healthy to diabetic².

Wagner says that in order to understand the pathophysiology of diabetes, we must investigate before the disease takes hold. “By the time diabetes manifests, there are often lots of metabolic derangements, which makes the analysis of the picture biased,” he says. Wagner adds that the development of diabetes is a long process, and without looking at pre-diabetes it can be difficult to know whether high blood sugar is cause or consequence of some of the metabolic derangements associated with diabetes.

Wagner and his team classified a cohort of people at increased diabetes risk into groups based on clusters of health metrics associated with the disease. Out of six groups, only three were found to have increased blood glucose levels, and only two had imminent diabetes risk. One group had moderate diabetes risk, but notably increased risk of kidney disease and all-cause mortality. “We know that glucose is not the whole story,” says Wagner.

Methods such as machine learning and advanced modelling enable researchers to make sense of the masses of data available to them, says Wagner. Technology such as continuous glucose monitoring devices, which sit under the skin and continuously measure interstitial fluid glucose levels as a proxy for blood glucose, has the potential to deliver so much data that insights would be lost without state-of-the-art analysis techniques.

With so many people living with diabetes worldwide and such a high disease burden on individuals and health authorities, there is a wealth of research into recognizing the warning signs as early as possible. By using the latest technologies, scientists can make sense of the available data and develop systems to categorize those at the highest risk. In addition to indicating when an individual needs to be tested for diabetes, this enables health authorities and clinicians to focus costly and resource-intense interventions on those who need them.

References