Insulin resistance — the upstream signal everyone underrates.

What insulin resistance actually is

Insulin resistance (IR) is a state in which cells — primarily in skeletal muscle, liver, and adipose tissue — respond less efficiently to a given concentration of circulating insulin. The pancreas compensates by producing more insulin to maintain the same glucose disposal. For a while, this works. Fasting glucose stays normal. Postprandial glucose stays normal. HbA1c (glycated hemoglobin, a measure of average glucose over roughly three months) stays normal.

Underneath, the system is straining. Fasting insulin rises. The ratio of insulin to glucose at any given moment shifts. Lipid handling deteriorates — triglycerides creep up, HDL drifts down. Visceral adiposity accumulates. Hepatic de novo lipogenesis accelerates. Endothelial function changes. And then, eventually, beta-cell compensation fails, post-meal glucose spikes break through, fasting glucose climbs into the prediabetic range, and HbA1c finally registers the problem that has been building for a decade or more [Kahn 2006].

The clinical implication is straightforward: by the time HbA1c flags prediabetes, the underlying physiology has been off-track for years. If you are interested in metabolic health rather than just diabetes diagnosis, the upstream markers matter.

The three markers: HOMA-IR vs fasting insulin vs HbA1c

HbA1c (glycated hemoglobin) reflects average glucose exposure across the lifespan of circulating red blood cells, roughly 90 days. It is the gold standard for diagnosing and monitoring type 2 diabetes. It is a poor early-detection tool because it only deviates once the beta-cell compensation has started to fail. A person can have substantial insulin resistance with an HbA1c of 5.3%.

Fasting insulin is the most direct early-warning marker available on routine labs. It is rarely ordered in primary care, which is a problem. In healthy lean adults, fasting insulin typically sits in the 2–8 µIU/mL range. Values above 10 µIU/mL with normal fasting glucose suggest the pancreas is already working harder than it should. Reference ranges on lab reports are wide — often up to 25 µIU/mL — and reflect the distribution of values in a metabolically unhealthy population, not an optimal one.

HOMA-IR (homeostatic model assessment of insulin resistance) is calculated from fasting glucose and fasting insulin: (fasting glucose [mg/dL] × fasting insulin [µIU/mL]) ÷ 405. A score below 1.0 suggests strong insulin sensitivity. A score of 1.0–1.9 is generally considered normal. Scores of 2.0–2.9 suggest early insulin resistance even when individual values look "normal" in isolation, and scores of 3.0 or higher indicate clear resistance [Matthews 1985] [Tang 2025]. A 2025 evaluation in a large biobank cohort confirmed that HOMA-IR rises before HbA1c in the natural history of metabolic decline and that simultaneous consideration of HbA1c and an insulin-resistance index improves diabetes risk prediction [Tang 2025].

The practical sequence we'd ask a clinician to consider: fasting glucose, fasting insulin, HbA1c, lipid panel (with apoB if available), and a calculated HOMA-IR. None of these alone tells the full story. Together, they make the upstream picture visible.

HbA1c tells you the system has already failed. HOMA-IR tells you the system is failing. The interventions that work — diet, training, sometimes pharmacology — are the same. The window in which they work most easily is upstream.

Diet: where the RCT evidence is strongest

Three dietary patterns have repeatable RCT support for improving insulin resistance independent of weight loss, and a fourth has strong support when weight loss is the primary goal.

DASH (Dietary Approaches to Stop Hypertension). Originally designed for blood pressure, the DASH pattern — high in vegetables, fruit, whole grains, legumes, low-fat dairy, lean protein, and low in sodium and ultra-processed food — has accumulated substantial evidence for insulin resistance specifically. A 2026 meta-analysis pooling 14 RCTs found that DASH reduced HOMA-IR by approximately 0.71 points versus comparator diets, with parallel improvements in waist circumference and triglycerides [DASH meta 2026]. In gestational diabetes and PCOS populations the signal is particularly clean, with HOMA-IR reductions of 20–30% reported in individual trials.

Mediterranean. The Mediterranean pattern — emphasis on olive oil, fish, legumes, nuts, vegetables, fruit, moderate wine, minimal processed meat — has the largest cardiovascular outcome dataset of any diet, and parallel improvements in HOMA-IR and fasting insulin in shorter mechanistic trials.

Lower-carbohydrate patterns. Carbohydrate restriction (broadly defined: less than 130 g/day or "low" by RCT convention) consistently improves fasting insulin, HOMA-IR, and triglycerides, and reduces postprandial glucose excursions. The question is durability. Adherence to strict low-carb patterns at 12 and 24 months tends to drift. The clean read of the data is that carbohydrate restriction works mechanistically but that long-term outcomes track adherence more than macro composition.

Any pattern producing 5–10% weight loss. When intentional weight loss occurs — through any sustainable pattern — HOMA-IR improvements of 30–50% are routine. The biggest single lever on insulin resistance in most adults is reducing visceral adipose tissue, and the dietary pattern that produces it is, within reason, the right one.

Exercise: resistance training and Zone 2 versus HIIT

Skeletal muscle is the largest site of insulin-mediated glucose disposal in the body. Anything that increases the quantity or quality of skeletal muscle improves whole-body insulin sensitivity. That is the entire mechanistic story behind why exercise works on insulin resistance.

Resistance training increases muscle mass and GLUT4 (glucose transporter type 4) content in trained tissue. The published RCT signal on HOMA-IR with structured resistance training (two to three sessions per week, compound lifts, progressive overload) is consistent and dose-dependent. The intervention is cheap, the side-effect profile is favorable, and the adaptation compounds over years.

Zone 2 endurance training — defined operationally as a sustained intensity at which blood lactate stays roughly at 2 mmol/L — is the popular biohacking prescription for mitochondrial density. The mitochondrial-biogenesis case is real. The question is whether Zone 2 specifically is uniquely effective, or whether any sustained aerobic work in the moderate-to-vigorous range produces similar adaptations. A 2025 narrative review concluded that the evidence for Zone 2 as the optimal intensity for mitochondrial capacity is weaker than the popular framing implies, and that higher-intensity work also drives mitochondrial adaptation [McNair 2025 Zone 2].

HIIT (high-intensity interval training) improves insulin sensitivity in a time-efficient package. In type 2 diabetes populations, an umbrella review of systematic reviews concluded that HIIT produces glycemic control improvements at least equivalent to moderate-intensity continuous training, often in less weekly training time [Poon 2025 HIIT umbrella]. The practical translation: if the schedule allows for two sessions of resistance training plus longer zone 2 work, do that. If the schedule allows for two sessions of resistance training plus 20 minutes of HIIT twice a week, that also works. The dose matters more than the format.

Where metformin and GLP-1s fit

Metformin. The Diabetes Prevention Program (DPP) remains the largest and longest randomized trial of metformin in prediabetes. After 2.8 years, metformin reduced progression to type 2 diabetes by 31% versus placebo, and intensive lifestyle intervention reduced it by 58% [DPP 2002]. At 21 years of follow-up in the DPP Outcomes Study, cumulative diabetes incidence remained 17% lower in the metformin arm and 24% lower in the lifestyle arm versus placebo [DPPOS 2025]. Metformin worked best in the participants with the highest baseline BMI, the highest fasting glucose, and a history of gestational diabetes. In younger, leaner participants, lifestyle outperformed metformin.

For someone with HOMA-IR in the 2.5–4.0 range, normal HbA1c, and a strong family history of type 2 diabetes, metformin is a defensible adjunct to lifestyle work — with the appropriate clinician relationship and B12 monitoring. It is not a substitute for the upstream interventions; it is an additional lever.

GLP-1 receptor agonists. Semaglutide and tirzepatide improve insulin resistance dramatically, but the mechanism is largely downstream of weight loss and reduced caloric intake — fat-mass reduction is doing most of the work, with secondary improvements in beta-cell function and direct effects on hepatic glucose handling. The trade-offs are real: meaningful lean-mass loss if protein and resistance training aren't preserved (we wrote about this in GLP-1 weight regain and the broader GLP-1 era analysis), cost, and the question of duration. For a patient with clear metabolic disease and lifestyle interventions that have plateaued, the case is strong. For early insulin resistance in an otherwise healthy adult, the case is much weaker.

A note on CGMs in this conversation

Continuous glucose monitors (CGMs) have entered the non-diabetic-wellness market aggressively. They produce a real-time feedback signal that some users find behavior-changing. They also produce noisy data that is easy to over-interpret. The interstitial glucose lag of roughly 10–15 minutes, sensor-to-sensor variance, and the absence of robust outcome data in healthy adults all complicate the story. We wrote a dedicated piece on this — see CGMs for non-diabetics — and the short version is: CGM is one useful input, not the metabolic oracle the marketing suggests.

A tiered framework

Frameworks, not protocols. Take this to a clinician.