Prediabetes on 3 tests tracked with 15-fold diabetes risk in a 20-year study of 7,186 Koreans

The word usually arrives on a lab report: prediabetes, a blood sugar result above normal but below the line for diabetes. What the report rarely says is which test put it there, and whether that matters. A blood sample taken before breakfast, a sugary drink followed by a second sample two hours later, and an HbA1c, a measure of average blood sugar over recent months, can each flag the condition, and they often disagree about who has it.

A Korean study that put the same people through all three tests every two years, for up to 20 years, suggests the disagreement carries information. Among 7,186 adults, each kind of prediabetes went with several times the risk of later type 2 diabetes, and risk rose in steps with each test failed, to about 15 times the risk of normal results for people who failed all three. Two measures of how the body makes and responds to insulin added a further layer: people low on either one were more likely to develop diabetes, even when their blood sugar results were normal. The team, led by Hun Jee Choe of Hallym University with colleagues at Seoul National University Bundang Hospital, Ajou University and Emory University in Atlanta, published the study in Diabetologia on October 17.

What the three prediabetes tests measure

A 2012 review in The Lancet, led by Adam Tabák, describes prediabetes as a high-risk state for diabetes, with blood sugar above normal but below the diabetes threshold. Each of the three usual definitions catches a different kind of trouble.

Impaired fasting glucose means a raised blood sugar after a night without food. Impaired glucose tolerance shows up only after the body is challenged: the person drinks a measured dose of sugar, and blood taken two hours later is higher than it should be. The third definition uses HbA1c, which reflects how much sugar has stuck to red blood cells, with HbA1c levels between 5.7% and 6.4% counted as prediabetes.

Experts have argued for years about which definition to use. A review by Justin Echouffo-Tcheugui and Elizabeth Selvin of Johns Hopkins counted five different definitions of prediabetes in current use, and little agreement on which one to prefer. The Tabák review also notes that insulin resistance and beta-cell dysfunction, the failing insulin-making cells of the pancreas, start before glucose changes are detectable. The Korean team added two measures aimed at that machinery: the insulinogenic index, a measure of insulin output, and an insulin sensitivity index.

Why the Korean cohort could answer the question

Much of what is known about how fast each type of prediabetes turns into diabetes comes from comparing separate studies. A 2013 meta-analysis from the University of Leicester pooled 70 of them and found that progression rates differ by prediabetes definition. People with both impaired fasting glucose and impaired glucose tolerance developed diabetes at about twice the rate of people with raised fasting glucose under the US definition, around 70 cases a year per 1,000 people against 36. But those numbers came from different populations measured in different ways.

The Korean study tested all three definitions in the same people. Its participants were adults aged 40 to 69 from the community-based Korean Genome and Epidemiology Study. They did not have diabetes at the start, and they returned every two years for both the glucose drink test and an HbA1c, for as long as two decades, which meant every person could be classified by all three definitions at every visit.

The team also stratified participants by two measures of insulin physiology. One captures insulin secretion, how much insulin the pancreas releases when sugar arrives; the other captures insulin sensitivity, how readily the body’s tissues respond to that insulin. Each was split into low and high. The researchers then estimated each group’s risk of diabetes over time, and used a second type of model to work out yearly chances of moving between normal blood sugar, prediabetes and diabetes, and the time spent in each state.

How much did each type of prediabetes raise diabetes risk?

Each of the subtypes defined by impaired fasting glucose, impaired glucose tolerance or HbA1c went with a large rise in risk. Compared with people whose results were normal, those with impaired fasting glucose had nearly six times the rate of new diabetes, and those with impaired glucose tolerance or a raised HbA1c a little under five times, after the researchers adjusted for other differences between groups.

The more striking pattern was the ladder. Each extra test failed pushed the risk higher, stepwise with the number of criteria met: about three times the normal rate after failing one test, a little over seven times for two, and about 15 times for all three. The range around that last estimate is wider, from about 12 to 19 times, but even its low end sits well above the other groups. The ladder and the per-test figures are separate analyses: the ladder sorts people by how many tests they failed, so its first step describes people who failed exactly one, while the higher per-test figures sort people by which test flagged them.

The transition models show what that risk looks like over time. People who met all three criteria had the shortest prediabetes duration, an average of 1.7 years in that state, and the lowest chance of drifting back to normal blood sugar, 5.4%.

Those figures compare groups with one another. Absolute rates from other studies give a sense of scale. The 2012 Lancet review estimated that each year 5% to 10% of people with prediabetes will progress to diabetes, while a similar share return to normal.

Insulin output and sensitivity flagged diabetes risk on their own

Blood sugar tests describe the result; the two insulin measures describe the machinery behind it. Participants with low insulin secretion or low insulin sensitivity were more likely to develop diabetes, and those low on both measures fared worst; in the authors’ words, dual impairment conferred greatest risk.

The insulin measures mattered even where the glucose tests saw nothing. Among people whose blood sugar was normal, those with low insulin output or sensitivity were more likely to develop diabetes than people who scored high on both. In the authors’ words, “Diabetes progression differs based on both glycaemic subtype and physiological subtype.” In other words, a normal lab result did not mean the same thing for everyone.

The authors conclude that adding the two insulin measures to the usual definitions “enhances risk stratification and may help to identify high-risk individuals for earlier targeted intervention.”

How the Korean results fit with other prediabetes research

The idea that prediabetes hides several different conditions has been building for a few years. In 2021 a group in Tübingen, Germany, led by Robert Wagner, sorted adults at high risk of diabetes into six clusters using glucose tests, MRI scans of body and liver fat, and genetic risk. They argued that the current definition of prediabetes neither reflects subphenotypes of the disease nor predicts what happens next, and found one cluster with only moderate diabetes risk but more kidney disease and higher death rates. The Korean study arrives at a related conclusion with simpler measurements and a much longer follow-up, though it tracked only diabetes, not complications.

The stepwise risk also fits the Leicester meta-analysis, where those who met two definitions progressed faster than those who met one. And the high stakes of prediabetes are not in doubt. In the Rotterdam Study in the Netherlands, among individuals aged 45 years with prediabetes, the lifetime chance of developing diabetes was about 74%.

What changes the outlook is the subject of a separate body of work. The US Diabetes Prevention Program trial, published in 2002, enrolled adults with raised fasting and after-meal blood sugar. Those who joined a lifestyle program aiming for 7% weight loss and 150 minutes of activity a week had 58% fewer new cases of diabetes, and those on metformin 31% fewer, as compared with placebo. The Korean study does not test whether matching such programs to a person’s subtype works better.

Limits of a single Korean prediabetes cohort

This is an observational study, so it describes who went on to develop diabetes; it cannot show that any one test result caused the outcome or that intervening on a subtype would change it. All participants were Korean adults aged 40 to 69 from a single community study, so the size of each risk may differ in other populations and age groups.

The results compare rates between groups, describing how much faster one group moved to diabetes than another. A fifteenfold risk means something different for a 45-year-old than for a 69-year-old, whose starting risk is higher.

Over 20 years, people’s weight, habits and treatment change too, and some participants may have received advice or medicines along the way, so the starting label is only part of what shaped each person’s outcome.

What the Korean results mean for people told they have prediabetes

The study suggests that a prediabetes label carries very different odds depending on how it was reached. In this Korean cohort, failing all three tests went with a risk of diabetes about 15 times that of normal results and a short stay in prediabetes, while failing one went with about three times. Whether any one person needs more testing, or what kind, is decided with their doctor. Do not stop or change a prescribed medicine without talking to your doctor.

The authors frame the insulin measures as a way to sharpen risk estimates, and other studies would need to confirm their value before they could change practice. For now, the study’s clearest contribution is a 20-year record, in the same people, showing that “prediabetes” covers a wide spread of risk.