- Prediabetes is not one uniform condition – researchers have identified distinct risk clusters with very different chances of progression and complications
- A machine-learning approach using blood-based DNA methylation markers classified high-risk groups with around 90 per cent accuracy in the study
- If refined into a practical test, this could support earlier, more targeted prevention without time-intensive clinical investigations
Prediabetes is often treated as a single label, but biologically it can look very different from person to person.
Some people remain stable for years with relatively low risk.
Others progress quickly towards type 2 diabetes or develop complications.
That variation matters because prevention is not one-size-fits-all – and the right intensity of support depends on who is actually at high risk.
Researchers from partner institutes within the German Center for Diabetes Research set out to make risk assessment easier and more precise.
Previous research has suggested that prediabetes can be grouped into at least six clusters, each with a distinct metabolic profile and a different likelihood of progression and complications.
The problem is that assigning someone to those clusters typically requires detailed testing – oral glucose tolerance tests, insulin measurements and imaging.
That is too slow and resource-heavy for routine use at scale.
In this study, scientists explored whether blood-based biomarkers could offer a shortcut.
They combined DNA methylation analysis – a way of measuring chemical tags on DNA that influence gene activity – with machine learning.
The goal was to identify an epigenetic signature that reliably flags high-risk clusters.
Their model used 1,557 epigenetic markers measured in blood.
Using these markers, the researchers reported that they could assign people to high-risk prediabetes clusters with around 90 per cent accuracy, including in an independent validation group.
Many markers were cluster-specific and appeared to map onto different biological pathways, which supports the idea that the clusters reflect genuinely different underlying biology rather than arbitrary categories.
Many of the markers had already been linked in other epigenetic studies with type 2 diabetes, chronic inflammation and cardiovascular and kidney disease.
That matters because it suggests the test is not only reflecting current blood sugar status, but also capturing wider risk biology that could shape future outcomes.
If this approach can be translated into a practical test, the implications are significant.
- Getting blood sugar back to normal may cut heart risk in prediabetes
- Study finds prediabetes is more life-threatening among people aged 20 to 54
- Gut microbiome among people with prediabetes reshaped by pistachio snacking at night
Instead of relying on time-intensive assessments, clinicians could use a standardised blood test to stratify risk, then match prevention intensity accordingly – more assertive interventions for those at highest risk and lighter-touch support where risk is lower.
The researchers are not claiming a finished diagnostic tool.
Their next step is to narrow the marker set so it becomes cheaper and more workable in routine diagnostics, with the longer-term aim of developing a dedicated analysis chip.
The direction is clear – move from broad labels to targeted prevention based on measurable biology, delivered through a test that is simple enough to use widely.
