# HOMA-IR, TyG and QUICKI: interpreting insulin resistance indices

- Canonical: https://nutrifit.health/knowledge/homa-ir-tyg-index-quicki-insulin-resistance-markers
- Language: en
- Published: 2026-10-08T18:02:29.168Z
- Updated: 2026-10-08T18:02:29.168Z
- Author: NutriFit Editorial

> Insulin signaling, clamps, HOMA1/HOMA2, QUICKI, two TyG scales and Matsuda: equations, units, validation and the limits of clinical interpretation.

Insulin resistance means a reduced tissue response to a particular action of insulin. It is neither a single laboratory number nor necessarily high glucose: pancreatic β cells can compensate by secreting more insulin. When that compensation fails, glucose rises while insulin may no longer be high. The same HOMA-IR can therefore arise from different combinations of secretion, clearance and insulin action.

HOMA-IR, QUICKI and TyG are calculated surrogates useful for research and selected clinical questions. They do not directly measure receptor phosphorylation, muscle glucose uptake or liver fat. Interpretation requires the original measurements, their units, sampling conditions, the calculation model and its validation population. A diagnosis based on an isolated index overlooks those conditions.

## From the insulin receptor to glucose metabolism

The INSR receptor is an α₂β₂ complex. Extracellular α subunits contribute to insulin binding; β subunits contain the transmembrane region and tyrosine kinase domain. INSR-A and INSR-B differ in inclusion of exon 11. Hormone binding promotes conformational change and receptor autophosphorylation, including activation-loop tyrosines. Residue numbering depends on the isoform and reference sequence: Tyr960 and Tyr972 can denote corresponding sites in different variants, rather than two separate switches.

IRS-1 and IRS-2 associate with the receptor through phosphotyrosine-recognition domains. Phosphorylated YXXM motifs recruit the PI3K regulatory subunit p85; catalytic p110 converts PI(4,5)P₂ into PI(3,4,5)P₃. This membrane lipid recruits PDK1 and AKT. AKT2 is especially relevant to metabolic signaling: its activation sites are Thr309 and Ser474, whereas the frequently quoted Thr308 and Ser473 refer to AKT1. PDK1 and mTORC2 support these respective activation steps. Numbers should not be transferred between isoforms without qualification (AKT reporter study, 2017).

In muscle and adipose tissue, AKT regulates TBC1D4/AS160, a Rab GTPase-activating protein. Phosphorylation and 14-3-3 binding facilitate Rab regulation and delivery of GLUT4-containing vesicles to the cell surface; GLUT4 is encoded by SLC2A4. The SNARE machinery, including VAMP2, syntaxin-4 and SNAP23, contributes to membrane fusion. This coordinated transport system cannot be reduced to one AS160 residue explaining every form of resistance. In muscle, inhibition of GSK3β, including phosphorylation at Ser9, favors glycogen synthase activity and glycogen storage.

The liver responds differently. Insulin suppresses glucose production through several direct and indirect mechanisms. AKT–FOXO1 regulates expression of PEPCK and glucose-6-phosphatase, while SREBP-1c and mTORC1 participate in lipogenesis involving ACC, FASN and SCD1. Impaired suppression of glucose production alongside retained lipogenesis is described as selective hepatic resistance. This is a useful model, rather than a universal binary division of all hepatic signals.

In adipocytes, insulin restrains lipolysis: PDE3B lowers cAMP and PKA activity, affecting perilipin-1, hormone-sensitive lipase HSL and the ABHD5–ATGL system. Inadequate suppression increases fatty acid delivery. Visceral depots may increase portal exposure of the liver, but most adipose tissue does not drain directly into the portal vein. The fraction of glucose disposal attributable to muscle during a clamp depends on insulin dose and protocol; fixed percentages do not describe every person.

## Why signaling becomes impaired

Inflammatory and nutrient-responsive pathways involving JNK, IKKβ and mTORC1–S6K1 modify IRS phosphorylation and feedback. Serine phosphorylation is not invariably inhibitory: the effect depends on residue, timing, tissue and species. The claim that free fatty acids simply bind TLR4 directly and thereby explain resistance is too strong; the lipid environment, immune cells and accompanying stimuli matter.

PTP1B is associated with the endoplasmic reticulum, with its catalytic domain facing the cytosol, and dephosphorylates signaling proteins. PTEN converts PIP₃ back to PI(4,5)P₂. SHIP2 produces PI(3,4)P₂, which itself supports signaling, so SHIP2 activity is not equivalent to abolishing all phosphoinositide signals.

For lipid-related resistance, lipid species and location matter alongside total fat content. Membrane sn-1,2-diacylglycerol can activate novel PKCs: PKCε in liver and PKCθ in muscle. Experimental studies describe hepatic PKCε-associated phosphorylation of INSR Thr1160 and muscle PKCθ-associated IRS-1 Ser1101. These are mechanistic findings; TyG cannot identify the affected receptor residue. Ceramides synthesized through pathways involving serine palmitoyltransferase can inhibit AKT through PP2A and atypical PKCs, with effects varying by molecular species and cellular context.

A normal BMI does not exclude ectopic fat. The discussion of the [TOFI phenotype and fat distribution](/knowledge/tofi-phenotype-ectopic-fat-visceral-adiposity) explains why waist circumference, body mass and organ fat describe different properties. Neither TyG nor TyG-BMI replaces organ imaging.

## Reference methods: clamps and the minimal model

In a hyperinsulinemic euglycemic clamp, investigators set an insulin infusion and adjust glucose infusion to maintain a specified glucose concentration. At steady state, the glucose requirement reflects insulin action under those experimental conditions. The approach described by DeFronzo and colleagues (1979) is an important reference, but different insulin doses and durations test different physiological states.

If GIR and endogenous glucose production EGP are both expressed in mg/min, an approximate balance under stable glucose-pool conditions is:

$$
M=\frac{\mathrm{GIR}+\mathrm{EGP}}{W}.
$$

GIR is the intravenous glucose infusion rate. This simplified balance assumes a steady glucose pool and no material urinary glucose loss; EGP is residual endogenous production, not necessarily zero.

W is body weight or fat-free mass, explicitly specified in the protocol; M is then in mg·kg⁻¹·min⁻¹. If GIR is already normalized to mass, it must not be divided again. Nonsteady glucose, urinary glucose loss or changes in the glucose pool require additional corrections. Without a tracer, residual EGP is unknown and cannot automatically be set to zero. Low- and high-insulin stages with, for example, 6,6-²H₂-glucose help distinguish suppression of production from peripheral disposal.

M/I relates disposal to achieved insulin concentration, specifying whether I is the plateau or increase above baseline and which scaling convention is used. It estimates insulin action, not metabolic insulin clearance. An M threshold cannot be transferred between body-weight and fat-free-mass normalization or between different insulin doses. A clamp requires trained staff, monitoring and hypoglycemia precautions; its cost and availability depend on the center.

A common minimal-model formulation for an intravenous glucose tolerance test represents remote insulin action X:

$$
\begin{aligned}
\frac{dG}{dt}&=-(p_1+X)G+p_1G_b,\\
\frac{dX}{dt}&=-p_2X+p_3(I-I_b),\\
S_I&=p_3/p_2,\qquad S_G=p_1.
\end{aligned}
$$

Time in this model is measured in minutes; G may be in mg/dL and I in μU/mL. p₁ and p₂ have units min⁻¹, p₃ min⁻² per unit insulin concentration, and S_I min⁻¹ per unit insulin concentration. p₁ represents glucose effectiveness, p₂ decay of remote action, and p₃ its input in response to insulin above baseline.

G and I are time-varying glucose and insulin concentrations; G_b and I_b are basal values. X has units of a rate, rather than being a measured insulin concentration. S_I estimates sensitivity and S_G glucose effectiveness at basal insulin. AIRg characterizes the early secretory response. DI = S_I × AIRg describes compensation when a hyperbolic relationship between secretion and sensitivity is justified. Sampling frequency and additional insulin administration vary across protocols; one sampling schedule is not the only standard (Bergman and colleagues, 1979).

[![Comparison of fasting indices, OGTT dynamics and a controlled clamp](https://api.nutrifit.health/api/v2/public/knowledge/files/images/nutrifit-scientific-20261002-homa-ir-tyg-index-quicki-insulin-resistance-markers-methods-en-v1.png)](https://api.nutrifit.health/api/v2/public/knowledge/files/images/nutrifit-scientific-20261002-homa-ir-tyg-index-quicki-insulin-resistance-markers-methods-en-v1.png)

The diagram compares measurement conditions. The clamp panel shows the adjustable glucose infusion; the required parallel controlled insulin infusion is not separately depicted in this schematic. Additional visible tubing does not make it a connection protocol. OGTT curves are qualitative, without a quantitative scale or patient data.

## HOMA: basal feedback rather than a tissue measurement

HOMA models the interaction between basal β-cell secretion and insulin action using fasting concentrations. Endogenous glucose production, mainly hepatic with a renal contribution, is important in the fasting state. Nevertheless, HOMA is not an exclusively hepatic test and does not separate organ-specific resistance. Matthews and colleagues (1985) used a computer-solved nonlinear model; the familiar HOMA1 equation is an approximation, rather than the entire original model.

$$
\begin{aligned}
\mathrm{HOMA1\!\!\!-IR}&=\frac{G_{\mathrm{mmol/L}}I_{\mathrm{\mu U/mL}}}{22.5},\\
\mathrm{HOMA1\!\!\!-IR}&\approx\frac{G_{\mathrm{mg/dL}}I_{\mathrm{\mu U/mL}}}{405}.
\end{aligned}
$$

G is glucose and I insulin from comparable fasting samples. Insulin in μU/mL and mIU/L has the same numerical value; conversion to pmol/L requires the convention appropriate to the assay. Multiply glucose in mmol/L by approximately 18.018 to obtain mg/dL. The conventional denominator 405 uses the rounded conversion 22.5 × 18; it is not a different physiological model. At G = 5.0 mmol/L and I = 10 μU/mL, HOMA1-IR is 2.22. This illustrates calculation, not a diagnostic threshold.

The approximate β-cell function equation is:

$$
\mathrm{HOMA1\!\!\!-\%B}=\frac{20I}{G_{\mathrm{mmol/L}}-3.5}.
$$

It does not measure β-cell mass or total secretory reserve. At G ≤ 3.5, the denominator makes the equation unsuitable. A high %B may indicate compensation rather than an exceptionally healthy pancreas; interpretation depends on the insulin secretion needed at the prevailing sensitivity.

HOMA2 is an updated computer model incorporating nonlinearities. Use the official calculator with the correct insulin-assay or C-peptide input and its valid input ranges, rather than inserting values into HOMA1. HOMA2-%S relates to HOMA2-IR as 100/IR. C-peptide can help assess endogenous secretion, but renal clearance limits its interpretation. HOMA1 and HOMA2 do not have interchangeable numerical cutoffs (Wallace and colleagues, 2004; Oxford HOMA2).

Pulsatile secretion, sampling variability, clearance, proinsulin cross-reactivity and immunoassay differences affect estimates. Historical precision in one validation study is not a universal error bound for modern laboratories. Comparable timing, metabolic conditions and assay platform help longitudinal interpretation, without making the surrogate a direct sensitivity measurement.

## QUICKI: a different scale for the same two measurements

QUICKI uses base-10 logarithms of numerical insulin values in μU/mL and glucose values in mg/dL:

$$
\mathrm{QUICKI}=\frac{1}{\log_{10} I+\log_{10} G_{\mathrm{mg/dL}}}.
$$

Both inputs must be positive. Units are fixed because logarithms act on numerical values relative to the stipulated units; glucose in mmol/L must first be converted. Within the usual positive-denominator domain, higher QUICKI represents greater sensitivity. Its relationship to conventional HOMA1 calculated with 405 is exact:

$$
\mathrm{QUICKI}=\frac{1}{\log_{10}(\mathrm{HOMA1\!\!\!-IR})+\log_{10}(405)}.
$$

QUICKI therefore adds no independent laboratory information. Within the same sample, HOMA1 and QUICKI rank participants in opposite directions, producing the same ROC-AUC when classification direction is handled consistently. Their Pearson correlations with clamp measures can differ because the scale is transformed. Logging values does not remove systematic insulin-assay bias.

Katz and colleagues (2000) reported r = 0.78 between QUICKI and the clamp in the original 56-person group. This validates a particular application, rather than proving universal superiority over HOMA. Proposed values such as 0.339 or 0.382 require their population, reference method and assay context; they are not general health boundaries.

[![Glucose and insulin for HOMA-IR and QUICKI, glucose and triglycerides for TyG, with opposite HOMA-IR and QUICKI directions](https://api.nutrifit.health/api/v2/public/knowledge/files/images/nutrifit-scientific-20261002-homa-ir-tyg-index-quicki-insulin-resistance-markers-inputs-en-v1.png)](https://api.nutrifit.health/api/v2/public/knowledge/files/images/nutrifit-scientific-20261002-homa-ir-tyg-index-quicki-insulin-resistance-markers-inputs-en-v1.png)

Arrows indicate interpretation rather than a causal sequence: higher HOMA-IR corresponds to greater resistance and higher QUICKI to greater sensitivity within the applicable model domain. Molecular icons are illustrative; the text defines equations and units.

## TyG: check the parentheses first

TyG combines fasting triglycerides and glucose, usually available without an insulin assay. Two scales occur in the literature and should be distinguished explicitly:

$$
\begin{aligned}
\mathrm{TyG}_A&=\ln\!\left(\frac{TG_{\mathrm{mg/dL}}G_{\mathrm{mg/dL}}}{2}\right),\\
\mathrm{TyG}_B&=\frac{\ln(TG_{\mathrm{mg/dL}}G_{\mathrm{mg/dL}})}{2},\\
\mathrm{TyG}_A&=2\mathrm{TyG}_B-\ln 2.
\end{aligned}
$$

ln denotes the natural logarithm. TG = 100 mg/dL and G = 90 mg/dL give TyG_A = 8.412 and TyG_B = 4.552. The historical B-scale value 4.68 maps mathematically to approximately 8.667 on scale A. This is a scale conversion, not a universal cutoff. If both measurements are initially in mmol/L, scale A requires:

$$
\mathrm{TyG}_A=\ln\!\left(\frac{(88.57TG_{\mathrm{mmol/L}})(18.018G_{\mathrm{mmol/L}})}{2}\right).
$$

The combined multiplier inside the logarithm is approximately 797.93. Check both the written equation and the reported numerical range in each study. An ambiguous notation such as “Ln[...] /2” does not justify mixing thresholds near 4.5 and 8.5.

TyG reflects combined disturbances of glucose and lipid metabolism, but does not directly measure DAG, ceramides, steatosis or insulin action. Triglycerides vary with recent meals, alcohol, lipoprotein-lipase genetics, illness and medication. Fibrates and other lipid treatments can change TG, and glucose-lowering treatments G, without proportional changes in every tissue. Accessible enzymatic assays are a practical advantage, not freedom from analytical or preanalytical error: ongoing glycolysis during delayed sample processing can lower measured glucose.

## TyG-BMI, waist circumference and other composite scores

$$
\begin{aligned}
\mathrm{TyG\!\!\!-BMI}&=\mathrm{TyG}\times\mathrm{BMI},\\
\mathrm{TyG\!\!\!-WC}&=\mathrm{TyG}\times WC,\\
\mathrm{TyG\!\!\!-WHtR}&=\mathrm{TyG}\times(WC/H).
\end{aligned}
$$

BMI is in kg/m²; WC is usually in centimeters. WC and height H must use matching units in WHtR. Specify which TyG scale was used. These products have distributions and cutoffs different from TyG alone. BMI does not directly measure body fat and waist is not a measurement of visceral-depot volume. LAP and VAI are separate equations with different inputs. Adding anthropometry may improve prediction in a selected population without guaranteeing a universally better measure of resistance.

## The OGTT and Matsuda index

The index developed by Masafumi Matsuda and Ralph DeFronzo (1999) combines basal and mean concentrations during an oral glucose tolerance test:

$$
\mathrm{ISI}_{\mathrm{Matsuda}}=\frac{10000}{\sqrt{G_0I_0\overline{G}\,\overline{I}}}.
$$

Glucose is in mg/dL and insulin in μU/mL. Mean concentrations are calculated from the chosen protocol points, commonly 0, 30, 60, 90 and 120 minutes. An area under the curve is not substituted directly for a mean. Shortened variants require their own validation. Higher values generally indicate greater sensitivity. The original publication reported r = 0.73 with a clamp measure, which is not numerical agreement one to one.

During an oral challenge, absorption, incretin effects, insulin secretion and clearance, glucose production and disposal act together. Matsuda represents mixed oral-test physiology and does not separate liver and muscle as a multistage tracer clamp can. Gutt 0,120, Cederholm and Stumvoll indices use different assumptions and sampling points; design determines their suitability rather than an assertion that Matsuda is always best.

Fasting indices also do not characterize every postmeal excursion. The article on [glycemic variability and CGM](/knowledge/glycemic-variability-cgm-postprandial-endothelial-damage) explains why an average, a fasting measurement and a time profile provide different information. CGM itself is not a direct measurement of insulin sensitivity.

## Reading validation, AUC and thresholds

| Method | Inputs | Useful perspective | Main limitation |
|---|---|---|---|
| HOMA1/HOMA2 | Basal G and I; eligible C-peptide inputs for HOMA2 | Basal feedback | Secretion, clearance, model and assay |
| QUICKI | The same basal G and I | Sensitivity on a transformed scale | Opposite ranks to HOMA1 |
| TyG | Basal G and TG | Combined metabolic phenotype | Parentheses, lipid influences, population validation |
| Matsuda | OGTT G and I | Response during an oral challenge | Protocol and mixed physiology |
| Clamp | Infusions, G, I; tracer when required | Action at a specified exposure | Dose, residual EGP and normalization |

| Primary study | Finding being evaluated | Conclusion it does not support |
|---|---|---|
| Matthews, 1985 | Initial validation of the basal computer model | One HOMA cutoff for every laboratory |
| Katz, 2000 | QUICKI–clamp r = 0.78 in the original group | Independent information beyond G and I |
| Matsuda, 1999 | Oral-index–clamp r = 0.73 | Complete interchangeability with clamps |
| Guerrero-Romero, 2010 | TyG compared with a euglycemic clamp | Transfer of an historical scale and cutoff unchecked |
| Vasques, 2011 | TyG compared with a hyperglycemic clamp | Equivalence of all clamp designs |

An ROC curve plots sensitivity against 1 − specificity. AUC measures discrimination, not probability calibration, agreement with a clamp or causality. Positive predictive value depends on prevalence. Using HOMA as the reference tests agreement with another surrogate rather than tissue action directly. Glucose and TG are already metabolic-syndrome criteria, so shared inputs account for part of apparent discrimination. Similar overlap occurs in scores incorporating waist or BMI.

Age, sex, puberty, body composition, ancestry, treatment and assay platform alter distributions. A table of ethnic “normal values” without identifiable validated samples creates false precision. Correlations from different populations and reference standards cannot fairly be ranked as a competition among indices.

## Associations with disease are not diagnoses

Higher indices have been investigated in metabolic syndrome, MASLD, cardiovascular risk, arterial stiffness, coronary calcification, renal disease and microvascular outcomes. Each question needs its own design, confounder adjustment and validation. A relationship between TyG and brachial-ankle pulse-wave velocity, as in Won and colleagues (2018), does not mean carotid-femoral velocity was measured or prove that lowering TyG prevents major cardiovascular events. The index cannot diagnose steatosis, MASH or fibrosis, or automatically determine a need for elastography.

In PCOS, resistance can coexist with a normal BMI; prevalence depends on definitions and samples rather than a universal percentage for lean women. The international 2023 guideline describes limited clinical relevance of routine insulin assays. An OGTT for glycemic status is a separate question: recommending glucose tolerance assessment does not mean every patient should undergo a multitimepoint insulin-based Matsuda assessment.

## Situations in which the number is particularly misleading

With G = 13.5 mmol/L and I = 3 μU/mL, arithmetic HOMA1-IR is 1.8. This cannot rule out serious metabolic dysfunction: deficient secretion is possible and basal-equilibrium assumptions are compromised. The pair alone cannot establish severe peripheral resistance, but it cannot certify health either.

Chronic kidney disease changes insulin and C-peptide clearance while genuine tissue resistance may also coexist. There is no universal eGFR boundary below which every high HOMA is an artifact. Injected insulin, variable immunoreactivity of analogs and insulin antibodies complicate interpretation. Insulin autoimmune syndrome is a particular clinical entity, not an explanation for every high insulin result.

Hemolysis, specimen handling and immunoassay interference can affect measurements. Biotin interferes with some assay designs; the direction, magnitude and appropriate interruption depend on platform and dose. A universal half-hour pause or fixed percentage error is unjustified. When a result conflicts with the clinical picture, the laboratory can assess method-specific interference.

First define the clinical question: dysglycemia diagnosis, cardiovascular risk, liver assessment or insulin-action research. Glucose, HbA1c, an OGTT when indicated, lipids, blood pressure and anthropometry answer different questions. A low HOMA or TyG does not negate abnormal glycemia, and a high value alone does not establish an indication for supplements.

## Diet, meal timing and physical activity

A Mediterranean dietary pattern, sufficient fiber and appropriate energy-balance changes can improve metabolic health. Viscous fiber affects absorption; fermentable fiber influences microbial metabolites and pathways including FFAR2/FFAR3–GLP-1. These mechanisms do not prove direct elimination of DAG in a particular patient. PREDIMED evaluated cardiovascular outcomes rather than a TyG-defined cure; its 2018 republication accounted for identified departures from randomization procedures.

Sutton and colleagues (2018) tested an early six-hour eating period with dinner before 15:00 in a small crossover study of men with prediabetes, with five weeks per schedule. This supports further investigation, not an identical 25–35% HOMA reduction for everybody or a mandatory 08:00–16:00 window. Meal timing must be considered with medication, tolerability and nutritional adequacy.

Muscle contraction stimulates glucose transport independently of the usual insulin signal through AMPK, Rac1 and TBC1D1/TBC1D4 regulation. Training changes capillary supply, mitochondrial function and body composition. Aerobic and resistance training offer complementary adaptations; one universal zone-2 lactate range is not a diagnostic or treatment standard for everybody.

## Supplements: a mechanism does not establish clinical benefit

Berberine and dihydroberberine are investigated as metabolic agents. Cellular and animal studies explore complex I, energetic stress and AMPK–ACC–CPT1, alongside lipoprotein regulation involving LDLR and PCSK9. These pathways do not guarantee complete TG suppression, equal efficacy in humans or absence of adverse effects. Small pharmacokinetic dihydroberberine studies establish exposure rather than long-term clinical outcomes; greater bioavailability alone does not define a therapeutic dose.

Myo-inositol, D-chiro-inositol and inositol phosphoglycans relate to investigated insulin-action pathways. Tissue epimerase models and ovarian features of PCOS do not justify universal treatment at a 40:1 ratio. The 2023 PCOS guideline does not recommend specific inositol types, doses or combinations because high-quality evidence remains limited.

A chromodulin model has been proposed for chromium, but is not an established obligatory receptor-enhancement mechanism in every person. Protein-bound α-lipoic acid as an enzyme cofactor differs from an oral supplement. The R form and racemate require their own studies rather than an assumption that the S form cancels half the benefit. Improvement of neuropathy symptoms does not prove normalization of insulin sensitivity. Magnesium contributes to Mg-ATP-dependent reactions; treating deficiency differs from prescribing a particular chelate to everybody. Population, outcomes, tolerability and interactions matter for each product, beyond a fall in a calculated index.

## Using indices responsibly

Keep original G, I and TG, units, date and fasting conditions. State HOMA1 versus HOMA2 and the exact TyG parentheses. Interpret values against an appropriate validation population and use comparable methods for follow-up. Consider secretion and clearance alongside sensitivity, and index changes alongside glycemia, lipids and meaningful clinical outcomes.

This article is an editorial examination of primary studies, mathematical definitions and methodological limitations. It is not a systematic review, independent clinical validation or a universal diagnostic and treatment protocol.

## FAQ

### Which index is most accurate?

There is no universal winner. A clamp specifies exposure, HOMA and QUICKI represent the basal state, Matsuda oral dynamics and TyG combined glucose and lipid metabolism. Comparison requires a common reference and comparable participants.

### Does QUICKI add information beyond HOMA1?

With the same G and I, it is a monotonic transformation in the opposite direction. Correlation can change with scale, while rankings and consistently oriented ROC-AUC agree.

### Why is TyG sometimes near 4.5 and sometimes 8.5?

Half the logarithm of a product differs from the logarithm of half the product. State the exact parentheses and mg/dL inputs; their numerical cutoffs cannot be mixed.

### Does low HOMA rule out metabolic dysfunction?

No. With inadequate secretion and high glucose, the product may be modest while model assumptions fail. Glycemia needs separate assessment.

### Can high TyG determine supplements?

An index alone establishes neither an indication nor a product or dose. Changing TG or glucose does not prove improvement in every tissue or clinical outcome.

## Sources

- [Matthews DR et al. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia. 1985;28:412–419.](https://doi.org/10.1007/BF00280883)
- [Wallace TM, Levy JC, Matthews DR. Use and Abuse of HOMA Modeling. Diabetes Care. 2004;27:1487–1495.](https://doi.org/10.2337/diacare.27.6.1487)
- [University of Oxford. HOMA2 Calculator: model description, input types and interpretation. Official software information; accessed 8 October 2026.](https://process.innovation.ox.ac.uk/software/p/2112/homa2-calculator/1)
- [Katz A et al. Quantitative insulin sensitivity check index: a simple, accurate method for assessing insulin sensitivity in humans. J Clin Endocrinol Metab. 2000;85:2402–2410.](https://doi.org/10.1210/jcem.85.7.6661)
- [Guerrero-Romero F et al. The product of triglycerides and glucose, a simple measure of insulin sensitivity. Comparison with the euglycemic-hyperinsulinemic clamp. J Clin Endocrinol Metab. 2010;95:3347–3351.](https://doi.org/10.1210/jc.2010-0288)
- [Vasques ACJ et al. TyG index performs better than HOMA in a Brazilian population: a hyperglycemic clamp validated study. Diabetes Res Clin Pract. 2011;93:e98–e100.](https://doi.org/10.1016/j.diabres.2011.05.030)
- [Matsuda M, DeFronzo RA. Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care. 1999;22:1462–1470.](https://doi.org/10.2337/diacare.22.9.1462)
- [DeFronzo RA, Tobin JD, Andres R. Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol. 1979;237:E214–E223.](https://pubmed.ncbi.nlm.nih.gov/382871/)
- [Bergman RN, Ider YZ, Bowden CR, Cobelli C. Quantitative estimation of insulin sensitivity. Am J Physiol. 1979;236:E667–E677.](https://doi.org/10.1152/ajpendo.1979.236.6.E667)
- [Norris DM et al. An improved Akt reporter reveals intra- and inter-cellular heterogeneity and oscillations in signal transduction. J Cell Sci. 2017;130:2757–2766.](https://doi.org/10.1242/jcs.205369)
- [Petersen MC et al. Insulin receptor Thr1160 phosphorylation mediates lipid-induced hepatic insulin resistance. J Clin Invest. 2016;126:4361–4371.](https://doi.org/10.1172/JCI86013)
- [Li Y et al. Protein kinase C Theta inhibits insulin signaling by phosphorylating IRS1 at Ser(1101). J Biol Chem. 2004;279:45304–45307.](https://doi.org/10.1074/jbc.C400186200)
- [Won KB et al. Relationship of insulin resistance estimated by triglyceride glucose index to arterial stiffness. Lipids Health Dis. 2018;17:268.](https://doi.org/10.1186/s12944-018-0914-2)
- [Teede HJ et al. Recommendations from the 2023 International Evidence-based Guideline for the Assessment and Management of Polycystic Ovary Syndrome. Hum Reprod. 2023;38:1655–1679.](https://doi.org/10.1093/humrep/dead156)
- [Sutton EF et al. Early time-restricted feeding improves insulin sensitivity, blood pressure, and oxidative stress even without weight loss in men with prediabetes. Cell Metab. 2018;27:1212–1221.e3.](https://doi.org/10.1016/j.cmet.2018.04.010)
- [Estruch R et al. Primary prevention of cardiovascular disease with a Mediterranean diet supplemented with extra-virgin olive oil or nuts. N Engl J Med. 2018;378:e34.](https://doi.org/10.1056/NEJMoa1800389)
- [Estruch R et al. Retraction and Republication: Primary Prevention of Cardiovascular Disease with a Mediterranean Diet. N Engl J Med. 2018;378:2441–2442.](https://doi.org/10.1056/NEJMc1806491)
- [Turner N et al. Berberine and its more biologically available derivative, dihydroberberine, inhibit mitochondrial respiratory complex I. Diabetes. 2008;57:1414–1418.](https://doi.org/10.2337/db07-1552)
- [Moon JM et al. Absorption kinetics of berberine and dihydroberberine and their impact on glycemia: a randomized, controlled, crossover pilot trial. Nutrients. 2021 Dec 28;14(1):124.](https://doi.org/10.3390/nu14010124)
- [Sacks DB et al. Guidelines and recommendations for laboratory analysis in the diagnosis and management of diabetes mellitus. Clin Chem. 2023;69:808–868.](https://doi.org/10.1093/clinchem/hvad080)