The Insulin Sensitivity Stack: Berberine + Inositol + ALA + Chromium

Key Takeaways

A fasting glucose of 95 mg/dL reads as normal on every lab report. Pair it with a fasting insulin of 14 µU/mL and the HOMA-IR comes out at 3.3 — insulin resistance by most thresholds, in a person who has been told for years that their sugar is fine. That is the number this stack is built around, and it is the number most blood-sugar supplement routines never measure.

The common routine starts at the shelf. Chromium goes in first because it costs nothing and every formula contains it. Alpha-lipoic acid goes in because it is an antioxidant and antioxidants are good. Berberine arrives last, if at all, because it is the one that causes stomach upset. Nobody measures insulin, and the stack is judged by how the person feels.

The evidence-ordered way starts with a blood test and ends with an optional trace mineral. It puts the compound with the largest trial base in the load-bearing position, splits the second layer by the population the trials were actually run in, and sets the exit criterion before the first capsule. This article is that version, in full, with the grades attached. The condensed protocol card is at The Insulin Sensitivity Protocol.

Before Anything: The Sentence That Comes First

Diet, resistance training, sleep and weight loss outperform every supplement in this article, and they outperform the drug too.

In the Diabetes Prevention Program, 3,234 people with impaired glucose tolerance were randomised to placebo, metformin 850 mg twice daily, or a lifestyle programme aiming at 7% weight loss and 150 minutes of weekly activity. Over an average 2.8 years, diabetes incidence was 11.0, 7.8 and 4.8 cases per 100 person-years. Lifestyle cut incidence by 58%; metformin by 31%. The number needed to treat for three years was 6.9 for lifestyle and 13.9 for metformin.

No supplement has been tested that way. The tiers, therefore, are: lifestyle first, metformin second, and — for people who have done the first and are not candidates for or have declined the second — berberine as metformin's supplement analogue. The comparison between those two is set out in Berberine vs. Metformin, and this stack assumes you have read it.

The Layer Map

The stack is graded C as a whole — to its weakest load-bearing layer, not its strongest component. Magnesium and berberine on their own are a B.

Layer 0 — Measure Before You Buy

An insulin-sensitivity protocol judged on fasting glucose is a protocol judged on the last number to move. Glucose is defended by rising insulin for years before it drifts up; by the time a fasting glucose is abnormal, the insulin resistance behind it is old.

So the baseline is fasting glucose and fasting insulin, drawn together, from which HOMA-IR is calculated, plus HbA1c.

HOMA-IR, with a worked calculation

HOMA-IR was published in 1985 as a way to estimate insulin resistance from a single fasting sample. The approximation formula is:

HOMA-IR = fasting glucose (mg/dL) × fasting insulin (µU/mL) ÷ 405

or, in mmol/L, glucose × insulin ÷ 22.5.

Take a fasting glucose of 95 mg/dL — comfortably "normal" — and a fasting insulin of 14 µU/mL:

95 × 14 ÷ 405 = 3.3

In the original validation, HOMA-IR correlated with the euglycaemic clamp at Rs = 0.88 and with fasting insulin at Rs = 0.81. A value above roughly 2.5 is commonly read as insulin resistance in adult populations, though the threshold varies by cohort and assay and the authors of the method themselves warn against treating it as a fixed cut-off. The person in the example has a normal fasting glucose and a HOMA-IR that would be flagged in most cohorts — which is the entire argument for measuring insulin.

Now the caveat. The 1985 paper reported a coefficient of variation of 31% for the insulin-resistance estimate. A single HOMA-IR of 3.3 could be 2.5 or 4.1 on a different morning. That is why this stack is judged at 12 weeks on HbA1c, with HOMA-IR read as a trend across at least two draws, never as a verdict from one.

HbA1c: the judgement number and the exit criterion

HbA1c averages glucose over about three months and does not care what you ate yesterday. Under the American Diabetes Association's diagnostic criteria, 5.7–6.4% is prediabetes and 6.5% or above is diabetes. That second threshold is the exit criterion for this stack: an HbA1c of 6.5% at baseline means the problem has a name, and the plan is a clinician's plan with this stack as an adjunct — not the reverse.

Optional CGM

A continuous glucose monitor for the first and last fortnight shows what averages hide: post-meal excursions, overnight glucose, and variability. It is the measurement-led approach to the same problem, and it is covered in the CGM and berberine article, which is the companion to this stack rather than a duplicate of it. That article is about seeing the signal; this one is about what to take once you have.

Magnesium: repletion, not pharmacology

Magnesium is a cofactor at the insulin receptor tyrosine kinase, deficiency is common in insulin resistance, and repletion is cheap. A meta-analysis of randomised trials found supplementation lowered HOMA-IR by 0.67 versus placebo (95% CI −1.20 to −0.14, P = 0.013) but not fasting glucose or HbA1c overall — with the effect on both HOMA-IR and glucose emerging only in trials of four months or more. A 2021 pool of double-blind trials found improvements in fasting and post-load glucose in people at high risk of diabetes.

Dose 300–400 mg elemental from a well-absorbed form (citrate, glycinate; threonate works but 2 g supplies only about 144 mg elemental, so it needs topping up for the metabolic role). Take it in the evening. It runs the full 12 weeks and beyond, regardless of what else is judged.

Layer 1 — Berberine, the Load-Bearing Layer

Berberine is the layer with a trial base large enough to pool properly. A 2022 meta-analysis of 37 randomised trials in 3,048 people with type 2 diabetes found it lowered HbA1c by 0.63% (95% CI −0.72 to −0.53), fasting glucose by 0.82 mmol/L and two-hour post-load glucose by 1.16 mmol/L. The effect was larger in people with higher baseline glucose. Total adverse events were not increased (RR 0.73), and hypoglycaemia was not increased (RR 0.48, 95% CI 0.21–1.08) — berberine's insulin effect is glucose-dependent.

The original 2008 pilot that compared it directly with metformin gave 36 newly diagnosed patients either berberine or metformin at 500 mg three times daily for three months. Both arms took HbA1c from 9.5% to 7.5%. In a second cohort of 48 poorly controlled patients, berberine cut fasting insulin by 28.1% and HOMA-IR by 44.7%. A third of patients had transient gastrointestinal effects.

Dose: 500 mg three times daily with meals, or 500 mg of a liposomal preparation once daily with the largest meal. Start at one dose a day for the first week to let the gut adapt. Berberine also inhibits CYP3A4 and P-glycoprotein — it raises levels of cyclosporine, tacrolimus and several statins, and that list needs checking before it goes alongside any prescription.

Layer 2 — The Fork

This is where the stack splits, and it splits by population because that is how the trials were run.

2A — Myo-inositol, for women with PCOS

The inositol evidence is PCOS evidence, and it is good. Nine randomised trials pooled in 2017 found fasting insulin fell (SMD −1.02, 95% CI −1.79 to −0.25, P = 0.009) and the HOMA index fell (SMD −0.59, P = 0.041), with trial sequential analysis confirming the insulin result. Twenty-six trials pooled in 2023 — 1,691 women against placebo and metformin — found 1.79 times the odds of a regular cycle versus placebo (95% CI 1.13–2.85), lower total and free testosterone, higher SHBG, and non-inferiority to metformin on most endpoints.

Dose: 2 g twice daily, as powder, ideally at the 40:1 myo:D-chiro ratio that restored ovulation best in a seven-ratio comparison. Insulin markers move within 12 weeks; the SHBG and androgen effects need 24. The full case, including the null 2025 pregnancy trial and the retracted founding study, is in Myo-Inositol: The Insulin Sensitiser for PCOS and Beyond.

Outside PCOS, myo-inositol has mechanism and no outcome trials. A man with metabolic syndrome can take it; he is taking it on theory.

2B — Alpha-lipoic acid, for neuropathy symptoms or the non-PCOS route

Be clear about what ALA is being bought for. For neuropathy symptoms the evidence is strong: 600 mg once daily cut the Total Symptom Score by 51% versus 32% on placebo in five weeks, with responder rates of 62% versus 26%, and higher doses added nausea and vertigo without benefit. For glucose the evidence is a C: the most rigorous meta-analysis of 28 trials found HOMA-IR fell by 0.48 (95% CI −0.79 to −0.16) and HbA1c did not change.

Dose: 600 mg racemic ALA, 30 minutes before breakfast — food reduces absorption. The one hard stop rule: any unexplained hypoglycaemia means stop ALA and measure insulin and insulin antibodies. Insulin autoimmune syndrome — antibodies against your own insulin — appeared 30 to 120 days after starting 600 mg/day in six European patients, five carrying HLA-DRB1*04:03. The trial series and the autoimmune story are in Alpha-Lipoic Acid: Glucose Disposal, Neuropathy and the Trials That Matter.

Why not both?

Because there is no trial of the combination, because three glucose-lowering agents in a medication-free person is more than the evidence supports, and because in someone on medication it is additive hypoglycaemia risk for a C-grade gain. Pick the arm that matches the population you are in.

Layer 3 — Chromium, Honestly

Chromium picolinate is in every blood-sugar formula because it is cheap, not because it works in the people buying it.

In established type 2 diabetes, the pooled data are modestly positive: a 2014 meta-analysis of 25 trials found HbA1c fell by 0.55% (95% CI −0.88 to −0.22) and fasting glucose by 1.15 mmol/L, with the effect clearer above 200 mcg/day and in people with poor baseline control. That is real, about a third of what berberine or metformin do, and the trials disagree with each other enough that the average is soft.

In people without diabetes, the evidence is null. A 16-week double-blind trial gave 31 non-obese, normoglycaemic adults chromium picolinate 500 mcg twice daily or placebo and measured insulin sensitivity by euglycaemic clamp — the reference method. No difference (P = 0.83). Worse, the participants with the highest serum chromium had a decline in insulin sensitivity (β = −0.83, P = 0.01), which weight, fat and lipids did not explain. The authors advised caution.

So: 200–400 mcg with breakfast if your HbA1c is already in the diabetic range and you are working with a clinician. Otherwise skip it. It will not hurt, and it will not help.

The Twelve-Week Schedule

Week 0 — Baseline. Fasting glucose and insulin (HOMA-IR), HbA1c, lipids including triglycerides, serum or red-cell magnesium, ALT/AST, creatinine. Optional CGM for two weeks. Start magnesium.

Weeks 0–2 — Magnesium only. Establish the habit. If a CGM is on, this is the unmedicated baseline.

Week 2 — Add berberine at 500 mg once daily with the largest meal. If tolerated after a week, go to 500 mg three times daily with meals.

Week 4 — Add the fork. Myo-inositol 2 g twice daily if PCOS; ALA 600 mg before breakfast otherwise. Start weekly home fasting glucose if on the ALA arm alongside berberine. Add chromium here only if the baseline HbA1c was ≥ 6.5% and a clinician is involved.

Weeks 4–12 — Run it. Weekly fasting glucose. A symptom diary if on ALA for neuropathy — symptom relief in the trials appeared by week 5. Any hypoglycaemic reading below 70 mg/dL (3.9 mmol/L) or symptoms of shakiness, sweating or confusion: stop berberine and ALA that day.

Week 12 — The judgement point. Repeat HbA1c, fasting glucose and insulin, lipids, ALT/AST, creatinine. Optional second CGM fortnight (weeks 10–12).

Exit Criteria — Decide These at Week 0

• Baseline or 12-week HbA1c ≥ 6.5%. Diabetes is a diagnosis under ADA criteria. The stack does not treat it; a clinician's plan does, with this as an adjunct if they agree.
• Any hypoglycaemic episode. Stop berberine and ALA. If on ALA, get insulin and insulin antibodies measured.
• You are on insulin, a sulfonylurea or a glinide and have not involved the prescriber. Do not start Layer 1 or 2 until you have.
• Neuropathy symptoms — burning, numbness, stabbing pain — in someone not known to have diabetes. That is an HbA1c and a foot examination first, ALA second.
• No HbA1c movement at 12 weeks with confirmed adherence. The signal to escalate to the drug tier, not to add chromium.

What This Stack Deliberately Excludes

• Cinnamon, bitter melon, gymnema and the rest of the "glucose formula" shelf. The trial bases are small, heterogeneous and mostly not placebo-controlled at the doses sold.
• D-chiro-inositol alone. In the ratio comparison it was the worst arm for ovulation; it is not the same product as 40:1.
• ALA above 600 mg. Higher doses added nausea, vomiting and vertigo in both ALADIN and SYDNEY 2 without adding benefit.
• Chromium for anyone without diabetes. See above.
• GLP-1 agonists, SGLT2 inhibitors and acarbose. These are the prescription tier and belong to a different conversation — The Metabolic Longevity Protocol covers the advanced pharmacological stack, and it is an alternative to this one for people already under prescription, not an extension.

Frequently Asked Questions

Can I take this stack with metformin?

Berberine plus metformin is additive on the same pathway and both trials and the substance profile flag it; it needs a prescriber's agreement and glucose monitoring. Magnesium alongside metformin is uncontroversial. Inositol has been trialled with metformin in PCOS without a hypoglycaemia signal.

Why measure insulin if my glucose is normal?

Because glucose is the last number to move. In the worked example, a fasting glucose of 95 mg/dL with an insulin of 14 µU/mL gives a HOMA-IR of 3.3 — flagged in most adult cohorts despite a normal glucose. Insulin resistance is years old by the time glucose shows it.

How do I choose between inositol and ALA?

By population. PCOS: inositol, because the trials were run in PCOS and show cycle regularity, insulin and androgen effects with non-inferiority to metformin. Neuropathy symptoms, or no PCOS: ALA, with the understanding that its glucose effect is a HOMA-IR change without an HbA1c change.

Is chromium worth adding just in case?

Only if your HbA1c is already diabetic. In people without diabetes, the reference-method trial found no effect and a signal in the wrong direction at high serum levels. In diabetes, it adds about half a point of HbA1c. That is the whole case.

What if my HOMA-IR goes up at 12 weeks?

Check the draw was genuinely fasting, and remember the 31% coefficient of variation — a single value can move a third either way. Read HbA1c first, and HOMA-IR as a trend across draws. If HbA1c is flat or down, one higher HOMA-IR is noise.

Can I run this without berberine?

You can, but then it is a magnesium-and-fork stack graded C throughout, with no layer that has demonstrated an HbA1c effect. Berberine is the reason the stack has a B-grade layer at all.

Related Research

• The Insulin Sensitivity Protocol
• Berberine vs. Metformin
• Continuous Glucose Monitors for Biohackers
• Myo-Inositol: The Insulin Sensitiser for PCOS and Beyond
• Alpha-Lipoic Acid: Glucose Disposal, Neuropathy and the Trials That Matter