Simple scores of insulin resistance were linked to death risk in 18,513 US adults

In a US national health survey followed for a median of almost ten years, five of twelve blood-test scores for insulin resistance were consistently linked to the risk of dying, including in people who already had heart disease. It is an observational study, so it cannot show cause, and the scores are not ready for clinical use.

Insulin is the hormone that tells muscle, fat and liver cells to take up sugar from the blood. In insulin resistance, cells respond less, so the body needs more insulin to keep blood sugar normal. It becomes more common with age and weight gain and is an early step towards type 2 diabetes, heart disease and kidney disease.

Measuring insulin resistance directly takes a lengthy clinic test with infusions of insulin and sugar. So researchers have invented many shortcut scores, calculated from routine measurements: fasting blood sugar and insulin, blood fats (triglycerides and "good" HDL cholesterol), the long-term blood sugar marker HbA1c, waist size and blood pressure. Nobody agrees which is best. A team led from Shandong Xiehe University in China compared twelve of them in one large US data set. Five were consistently linked to the risk of dying. As this is an observational study, these are links, not proof of cause.

What they did

The researchers used the National Health and Nutrition Examination Survey (NHANES), which examines a sample of the US population every two years. They included 18,513 adults from the 1999–2018 surveys who had fasting blood tests, with an average age of 49, and linked them to national death records up to the end of 2019.

For each person they calculated twelve scores, among them HOMA-IR (from fasting sugar and insulin), the TyG index (from triglycerides and sugar), the "atherogenic index of plasma" (the ratio of triglycerides to HDL), the stress hyperglycemia ratio (fasting sugar compared with HbA1c), and eGDR, which combines waist size, high blood pressure and HbA1c (for eGDR, higher means more sensitive to insulin). People were also sorted into five stages of what the American Heart Association calls the "cardiovascular-kidney-metabolic syndrome", from stage 0 (no risk factors) to stage 4 (heart disease or stroke). The analyses took into account age, sex, ethnicity, income, smoking, alcohol and cancer.

What they found

  • Over a median of 9.7 years, 2,735 people (14.8%) died, from 2.8% in stage 0 to 41.6% in stage 4.
  • Five scores held up in every check: a higher eGDR went with a lower risk of death (about 7% lower per point), while higher HOMA-IR, TyG index, atherogenic index and stress hyperglycemia ratio went with a higher risk. These links remained after taking medicines for blood pressure, diabetes and cholesterol into account and after leaving out people who died in the first two years.
  • Six other scores were also linked to death at first, but the links weakened or disappeared in at least one of these checks.
  • In people with heart disease or stroke (stage 4), HOMA-IR, eGDR, the TyG index, the stress hyperglycemia ratio and one more score were still linked to the risk of dying.
  • Many links were not straight lines. For HOMA-IR, for example, risk was lowest near the middle of the range and rose at both ends.
  • Results looked different in different stages and blood-sugar groups, but none of these differences survived correction for the many tests, so the authors treat them as exploratory.

What to keep in mind

  • Association, not cause. The authors note that the observational design "precludes causal inference" and that hidden factors, including how intensively people were treated, may play a part.
  • Each score was measured once.
  • eGDR shares ingredients (waist size, high blood pressure) with the staging system, which may blur the comparison.
  • The heart-kidney stages could only be approximated with the available data; stage 3 in particular was defined narrowly and was very small (161 people).
  • The survey's sampling design was not fully used, so the results describe the people studied, not risks for the whole US population.
  • Lab methods for insulin and some questionnaire items changed over the twenty years of surveys.

Why it matters

The scores in this study can be calculated from tests that many people already have, and some, like eGDR and the TyG index, don't even need an insulin measurement. The study suggests that a handful of them carry useful information about long-term risk at every stage, from healthy people to those with heart disease. But the authors stress that their predictive value still needs to be formally tested in other groups of people before doctors could use them.