Prediabetes is not one risk: meeting all three criteria raised diabetes odds 15-fold

Prediabetes is usually treated as a single yellow light on the road to type 2 diabetes. A new long-term study argues it is really several lights of very different brightness, and that counting how many are lit sharpens the warning considerably [s1]. Published in Diabetologia, the analysis followed more than 7,000 adults for up to two decades and found the risk of progressing to diabetes ranged across subtypes and multiplied as more definitions applied at once [s1].

The researchers analysed 7,186 Korean adults aged 40 to 69 from the community-based Korean Genome and Epidemiology Study, followed for up to 20 years with an oral glucose tolerance test and HbA1c every two years [s1]. Prediabetes can be flagged three ways — by impaired fasting glucose, by impaired glucose tolerance on the oral test, or by a raised HbA1c — and the team compared how each subtype, alone and in combination, predicted progression [s1]. The bibliographic record for the study is dated 17 October 2026 [s2].

The subtypes carry different risk

Compared with people whose glucose was normal, each prediabetes subtype raised the risk of developing diabetes, but not equally [s1]. The adjusted hazard ratio was 5.84 (95% confidence interval 5.03 to 6.79) for impaired fasting glucose, 4.60 (4.17 to 5.07) for impaired glucose tolerance, and 4.68 (4.22 to 5.19) for HbA1c-defined prediabetes [s1]. A hazard ratio of roughly five to six means that, over the follow-up, people in these groups developed diabetes at five to six times the rate of those with normal glucose, with impaired fasting glucose carrying the highest single-criterion risk.

The clearer signal came from counting criteria. Relative to normal glucose, the adjusted hazard ratio was 3.10 (95% CI 2.81 to 3.41) for meeting one criterion, 7.33 (6.50 to 8.28) for two, and 15.3 (12.2 to 19.3) for all three [s1]. Markov modelling — a method that estimates how people move between health states over time — put the shortest time spent in prediabetes, 1.7 years, and the lowest chance of returning to normal glucose, 5.4%, in the group meeting all three criteria [s1]. These are the people furthest along and least likely to turn back.

Adding the physiology

The study then went beyond blood-sugar readings to the mechanics underneath them: how much insulin the body secretes, captured by the insulinogenic index, and how well it responds to that insulin, the insulin sensitivity index [s1]. Low values on either predicted higher diabetes risk, and having both impaired carried the greatest risk, with an adjusted hazard ratio of 4.06 (95% CI 3.37 to 4.90) [s1].

Notably, even people whose glucose readings were still normal, but who had low insulin secretion or sensitivity, were at raised risk compared with those whose physiology was intact [s1]. That is the study's more provocative claim: a conventionally "normal" result can hide a body already struggling, and the standard glucose cut-offs miss it.

What this is, and is not

This is an observational cohort study, and the usual cautions apply [s1]. It can establish that subtypes and combinations predict diabetes; it cannot by itself prove that acting on those subtypes changes outcomes, because no one was assigned to an intervention. The cohort is Korean adults aged 40 to 69, so the precise hazard ratios need not transfer to other ancestries or ages — glucose-tolerance patterns are known to differ across populations. And the insulin indices, while informative, are research measures not routinely calculated in primary care.

What the study adds is a case that the single label "prediabetes" bundles together people on quite different trajectories, from a modest threefold risk to a 15-fold one [s1]. The authors argue that integrating the physiological subtypes with the conventional glucose definitions "enhances risk stratification" and could help identify the highest-risk individuals for earlier, targeted prevention [s1]. In practice that would mean treating a borderline result not as a uniform nudge toward diet and exercise, but as a spectrum where some people warrant far closer attention than others.

For now, the practical reading is narrower and worth stating plainly: a prediabetes result is not one thing, the number of ways you meet it matters, and someone ticking all three boxes is in a genuinely different situation from someone who ticks one. Whether reshaping screening around that improves health is a question for the trials this kind of study is designed to motivate, not settle.

This article is informational and does not constitute medical advice. Decisions about diabetes risk and prevention belong with a patient and their clinician.