Our scientists have developed an enhanced prediction tool to find children at high risk of type 1 diabetes years before they’re diagnosed.
Building on genetic risk
In January 2019, our research fellow Dr Richard Oram published his development of an improved ‘risk calculator’. It uses genes linked to type 1 diabetes to find those at highest risk of developing the condition. The genetic risk score, called the GRS2, was the most accurate to date in predicting a newborn baby’s risk of type 1 diabetes – information which could be used for two purposes.
First, newborns with a high risk of developing type 1 diabetes could be closely monitored to make sure the condition is picked up as soon as possible, to reduce the risk of life-threatening complications at the point of diagnosis. Second, children at high risk could be directed towards clinical trials testing new treatments aiming to delay or prevent type 1 diabetes.
To build on this work, Dr Oram and scientists from the Pacific Northwest Research Institute in Seattle have now developed a new tool to predict the likelihood of developing the condition in the first 10 years of life. They used data from the TEDDY study – involving almost 8,000 children at high risk of type 1 diabetes – to find which combination of factors known to indicate increased risk of type 1 gave the most accurate prediction.
They looked at family history of diabetes, the presence of auto-antibodies (signs that the immune attack against the pancreas has begun), weight at one year, where you live and episodes of sinusitis (a swelling of the sinuses, usually caused by an infection).
They found that the accuracy of Dr Oram’s genetic risk score was improved when combined with information about auto-antibodies and family history, helping to better predict who – from a young age – would later develop type 1 diabetes. And most importantly, it is estimated to be twice as effective at preventing potentially life-threatening diabetic ketoacidosis (DKA) at diagnosis compared to looking for auto-antibodies alone – the method scientists currently use.
Screening for type 1 diabetes
But now that we have a really accurate way of predicting who is likely to develop type 1 diabetes, how do we approach screening an entire population? Dr Oram’s team then used TEDDY’s data to test three potential strategies for population-level screening.
- Classic: selected babies with a high genetic risk and followed them closely (every 3 months until age three, every 6 months until age six, every year until age eight).
- Simple adaptive: same as above, but recalculated the combined risk score each year, removing children whose risk fell below a certain threshold.
- Advanced adaptive: same as above, but instead of removing the low-risk children, they only followed them every two years.
They compared the number of follow-up tests needed for each strategy to find three-quarters of new type 1 diabetes cases, at least four weeks before diagnosis. Using the combined risk score meant that far fewer follow-ups were needed than just calculating genetic risk alone – 25% fewer for the simple approach and 51% fewer for the advanced approach. This suggests using the combined risk score is not only a more accurate prediction tool but could deliver a more cost-effective and feasible way to screen a whole population for type 1 diabetes.
These ‘risk calculators’ aren’t currently being used and it’s important to note that this approach calculates the likelihood of developing type 1 diabetes in the future, but doesn’t give a definitive answer. Some children who are identified as high risk will not go on to develop type 1 diabetes.
Nonetheless, being able to better predict which children are at highest risk of getting type 1 diabetes years later is an exciting prospect. Dr Elizabeth Robertson, our Director of Research, said:
This research was published today in Nature Medicine.