HOMA-IR and the Kraft insulin test: seeing insulin resistance 10 years before diabetes

A normal fasting glucose is preserved until about 60% of pancreatic beta-cell function is already lost. If you want to see a metabolic problem before HbA1c climbs above 5.7%, there are two tools your physician should put next to your blood work: HOMA-IR and the Kraft insulin test.

Why fasting glucose alone hides the problem

The Whitehall II study (Lancet 2009) tracked over 6,000 people for 13 years before their type 2 diabetes diagnosis. Fasting glucose only starts rising 4–5 years before diagnosis, but insulin sensitivity begins declining 13 years earlier. That means for most of the "silent phase", your glucose looks normal while insulin has to keep climbing to hold that norm in place.

This is exactly where HOMA-IR comes in.

HOMA-IR: two numbers, one equation

HOMA-IR (Homeostasis Model Assessment of Insulin Resistance), developed by Matthews et al. (Diabetologia 1985), combines fasting insulin and fasting glucose into a single index:

"HOMA-IR = (fasting insulin mU/L × fasting glucose mmol/L) / 22.5"

Interpretation from the published literature:

• ≤ 1.0 — optimal insulin sensitivity
• 1.0–1.9 — upper end of normal, not metabolically healthy
• 2.0–2.9 — early insulin resistance, usually without symptoms
• ≥ 3.0 — clinically significant resistance, strong risk factor

HOMA-IR needs only one fasting draw, is inexpensive, and you can compute it yourself. Its limit: once the pancreas is exhausted (beta-cell function falling, insulin deficient), the formula can look falsely normal. For those cases we have a more precise, more demanding test.

The Kraft insulin test: a slow 5-hour walk

Dr Joseph Kraft published "Detection of Diabetes Mellitus In Situ (Occult Diabetes)" in 1975, based on glucose and insulin curves in over 14,000 patients after a 100 g glucose load. A standard oral glucose tolerance test (OGTT) tracks the glucose response; the Kraft test tracks the insulin response in parallel. What emerges is an insulin curve whose shape says more than the glucose curve.

Kraft classified five patterns. Only pattern I is truly normal — insulin peaks at 0.5–1 h and returns to baseline by 2 h. Patterns II–V represent progressive dysfunction: delayed peak, exaggerated peak, prolonged decay. The clinically most relevant in the Kraft database were patterns III–V in patients whose glucose curves were perfectly normal — occult hyperinsulinaemia (Crofts et al., BMJ Open Diabetes Res Care 2016).

"Kraft called it "diabetes in situ" — the disease is already there, but standard testing does not see it."

The full Kraft protocol is rarely done in Estonian clinics because it requires 5 hours and 4–5 insulin draws. In practice we use a shorter modification: OGTT plus insulin at 0, 60 and 120 min. That captures most of the meaningful patterns.

Which numbers actually signal risk

The insulin reference ranges most labs report (2–24 mU/L) are statistically derived, not metabolically optimal. Kraft, Crofts and others have proposed stricter thresholds that separate a genuinely healthy phenotype:

• Fasting insulin < 6 mU/L — optimal
• Fasting insulin 6–10 mU/L — borderline, worth watching
• Fasting insulin > 10 mU/L — early hyperinsulinaemia regardless of glucose
• 2-hour insulin > 50 mU/L after OGTT — Kraft's strong risk marker

Which patient gets which test

In practice I always start with HOMA-IR — cheap, fast, catches most cases. I add OGTT with insulin when HOMA-IR is borderline (1.7–2.5) but the clinical picture is loud: visceral fat gain, PCOS, acanthosis, post-meal fatigue, low HDL, high triglyceride/HDL ratio. That's when OGTT+insulin frequently uncovers an occult Kraft pattern III that HOMA-IR missed.

What next when the number is high

A high HOMA-IR is not a diagnosis — it is a roadmap. Insulin resistance is largely a lifestyle variable, and it is one of the few metabolic markers that genuinely moves in 8–12 weeks with real intervention: body composition (visceral fat down), 150+ min a week of strength and interval training, carbohydrate quality and quantity, 7–8 hours of sleep, deliberate work on chronic stress. The DiRECT trial (Lancet 2018) showed 46% of patients entering type 2 diabetes remission over 12 months with such an intervention — and HOMA-IR normalized before weight stabilized.

That is why I re-check HOMA-IR in every metabolic patient 12 weeks into the program: it is the compass that tells us the direction is correct long before weight, cholesterol or HbA1c confirm it.

HOMA-IR cut-offs — why 'normal' is not 'optimal'

Labs typically report a HOMA-IR reference of < 2.5–2.7, but the metabolically optimal range is 0.5–1.4 (Gayoso-Diz et al., BMC Endocr Disord 2013). A value of 2.0 'within range' often means the pancreas is already compensating — 5–10 years before fasting glucose starts to rise.

When HOMA-IR misleads and Kraft is more accurate

• Nocturnal hypoglycemia or shift work — distorts fasting insulin
• Hypothyroidism — reduces basal insulin secretion without any improvement in sensitivity
• Pregnancy, PCOS, late-phase compensation — HOMA-IR stays 'normal' but 2h OGTT insulin > 60 µU/ml

A practical algorithm: how to order and interpret

Combine fasting glucose + insulin + HbA1c + TG/HDL ratio. If HOMA-IR > 1.8 or TG/HDL > 1.5, consider a 2h OGTT with insulin. The Kraft test (5-point OGTT) uncovers hidden hyperinsulinism in ~60% of patients with 'normal' HOMA-IR and HbA1c (Crofts, Diabesity 2015).