Berberine and the Metformin Question

Berberine gets called nature's metformin, and it turns out that is not marketing — both compounds jam the same station of the same assembly line inside your mitochondria, measured directly, with metformin run as the comparison. This episode rebuilds our earlier, more sceptical take with both pans on the scale: what berberine's forty-odd trials actually show, why the large outcomes trial does not exist and what that silence is really about, and the cost on metformin's side of the ledger — a vitamin B12 deficiency in one patient in fourteen that produces a neuropathy indistinguishable from the one everybody is already expecting, so nobody orders the test.

Show notes

Before we start, I want to tell you what changed my mind about this episode, because we have made it before and we made it differently.

The first version of this script treated berberine the way most careful people treat it: a supplement with one small trial behind it, sold on a comparison to a real drug it had not earned. That version asked berberine to prove it could prevent a heart attack, noted that it never had, and closed the case. It was rigorous. It was also, I have come to think, rigorous in one direction only — and a scale that only weighs one pan is not a scale.

Here is what the first version never did. It never asked what the drug on the other side of the comparison costs the person taking it. It granted metformin its decades of outcome data — correctly, that data is real — and then stopped, as though a medicine's harms were somebody else's topic. They are not. They are half the comparison. And in metformin's case the harm is not hypothetical, not rare, and not disputed: it is proven in a placebo-controlled trial, it is proven again across thirteen years of follow-up, and it produces a set of symptoms that look exactly like the disease being treated — which is the reason almost nobody goes looking for it.

So this is the episode rebuilt with both pans on the scale.

Both of them are plants

Start with the thing that never makes it onto the bottle, because it dissolves the argument people think they are having.

Metformin is also a plant.

Before it was a prescription it was Galega officinalis — French lilac, goat's rue — used in medieval Europe for the thirst and the constant urination we would now call diabetes. Chemists pulled a family of compounds called the guanidines out of it, refined them, and one of those refinements became the most prescribed diabetes drug on earth. The pharmaceutical everyone is trying to escape started in the same field as the supplement they are escaping to.

So the question was never natural versus synthetic. That framing is a marketing invention on both sides — the supplement aisle uses it to sell, and the clinic uses it to dismiss. Both of these came from a plant. The real difference is not what they are made of. It is that one of them had a sponsor who could afford to prove it, and the other did not.

Hold onto that sentence. We are going to come back to it, and it is going to do more work than you expect.

What we are actually trying to fix

One paragraph of groundwork, because the rest of this only makes sense if we are clear about what has gone wrong in type 2 diabetes. It is not what most people assume.

Sugar in your blood is not a contaminant. It is fuel, and your body works hard to keep the amount in circulation inside a narrow band — enough to run your brain, not so much that it starts doing damage. The hormone that does the keeping is insulin, made in the pancreas, a gland sitting behind your stomach. After a meal, insulin goes out and knocks on the door of your muscle and fat cells and tells them to open up and take the sugar in.

In type 1 diabetes the pancreas has been destroyed by the person's own immune system and there is no insulin. That is a different disease, it is not what this episode is about, and nothing here applies to it.

In type 2 diabetes there is plenty of insulin. The problem is that the cells have stopped answering the door. Insulin knocks and nothing opens. So the pancreas knocks harder — more insulin — and for years that works, which is why the disease is silent for so long. Blood sugar stays normal while insulin levels climb to two or three times where they should be. Eventually the pancreas cannot shout loudly enough, blood sugar starts to drift up, and only then does anyone measure anything and give it a name.

By the time you are diagnosed with type 2 diabetes, you have usually had the underlying disorder for ten years or more.

Two consequences follow, and both matter for what comes next. First, a treatment that lowers blood sugar by forcing more insulin into an already insulin-flooded system is treating the symptom and worsening the cause. Several older diabetes drugs do exactly this. Second, the high blood sugar is the thing that gets measured, but the high insulin is doing its own damage the whole time — to blood vessels, to fat storage, to blood pressure. That is why the interesting treatments are the ones that lower glucose and insulin together, and why a drop in insulin resistance is a better result than a drop in glucose alone.

Keep that test in mind. We are going to apply it to berberine in a few minutes, and it passes.

What berberine actually is

Let us be specific about the molecule, because "berberine" is a word people say without picturing anything.

Berberine is an alkaloid. An alkaloid is a naturally occurring compound built around nitrogen, made by a plant, and usually bitter and biologically active — caffeine is an alkaloid, so is morphine, so is nicotine. Plants make them largely to be unpleasant to eat. Quite a lot of medicine is us discovering that a plant's chemical warfare is useful at the right dose.

More precisely, berberine is an isoquinoline alkaloid. That names its skeleton: two six-sided carbon rings fused together with a nitrogen worked into one of them. Onto that skeleton berberine hangs two little oxygen-containing caps, and the whole flat structure carries a permanent positive electrical charge on its nitrogen atom.

That last detail is not trivia. It is the single fact that explains most of berberine's behaviour, and we will use it three separate times before this episode is over. A permanently charged molecule does not slip easily through the greasy membranes that wrap every cell. It is also intensely yellow — berberine is a dye as much as a drug, and for centuries it was used to colour wool and leather.

It comes from a long list of unrelated plants that all independently arrived at the same molecule: Coptis chinensis (goldthread), Berberis species (barberry, Oregon grape), Phellodendron (Amur cork tree), and Hydrastis canadensis — goldenseal, which Native American medicine used for wounds, infections and digestive trouble long before anyone had isolated the compound responsible.

Which form matters, and the label often will not say. Almost every capsule sold is berberine hydrochloride — the molecule paired with chloride to make a stable salt. You will also see berberine sulfate. And there is a third form, dihydroberberine, which is berberine with two extra hydrogen atoms added; it is absorbed several times better and is converted back into ordinary berberine inside the body.2 If a product says only "berberine complex" or "berberine extract" with no salt named and no milligrams of actual alkaloid, you do not know what you have bought. That is a real problem with this supplement and I am not going to pretend otherwise.

The machine in the middle of the cell

Now the mechanism, slowly, because everything else in this episode rests on it and because it is genuinely one of the more elegant stories in metabolism.

Inside almost every one of your cells are small structures called mitochondria. Their job is to take the food you have eaten, combine it with the oxygen you have breathed, and use the energy released to charge up a molecule called ATP — adenosine triphosphate, the rechargeable battery that powers essentially every activity in the body. When you think, lift something, or heal a cut, you are spending ATP.

The charging happens along an assembly line called the electron transport chain. Picture a bucket brigade. Electrons stripped off your food are passed from one protein complex to the next, and at each handoff a little energy is released and used to pump hydrogen ions across a membrane. That builds up pressure, like water behind a dam. The pressure then drives a molecular turbine that manufactures ATP.

The first station on that assembly line is called, without much imagination, complex I.

And here is what berberine does: it jams complex I.

That was shown directly in 2008, in the journal Diabetes, by a group who measured oxygen consumption in muscle cells and in isolated muscle mitochondria and found berberine shutting down respiration in a dose-dependent way, through a specific effect on complex I.2 A second group confirmed it in 2014 with different equipment, reporting that complex I activity was "almost fully blocked" in muscle cells by berberine.3

Your instinct on hearing "it jams the energy factory" is that this sounds bad. Stay with me, because the next step is the whole point.

The fuel gauge, and why jamming the engine helps

Cells monitor their own energy supply. The sensor they use is an enzyme called AMPK — AMP-activated protein kinase. Think of it as the cell's fuel gauge, and understand that it reads empty, not full. When ATP is plentiful, AMPK sits quiet. When ATP runs down, AMPK switches on.

And when AMPK switches on, it issues a standing order to the whole cell: stop storing, start burning. Concretely, it tells muscle to pull glucose — sugar — out of the bloodstream and use it. It tells the liver to stop manufacturing new glucose and dumping it into the blood, which is something the liver of a person with type 2 diabetes does far too enthusiastically. It tells fat cells to release stored fat for fuel rather than hoarding more.

Now put the two halves together. Berberine slows the energy assembly line a little. ATP dips a little. AMPK reads the dip and switches on. The cell responds by pulling sugar out of your blood and burning it, and by telling the liver to stop pouring more in.

The blood sugar falls because the cell thinks it is short of energy. It is a deliberate, mild, reversible fuel shortage, and it produces almost exactly the metabolic state that exercise produces — which is not a coincidence, because exercise activates AMPK too, by genuinely spending ATP.

The sentence that reframes the whole comparison

Here is the part that the phrase "nature's metformin" gets right by accident, and that almost nobody states precisely.

Metformin does the same thing.

Not something similar. Not something analogous. The 2008 Diabetes paper measured berberine's inhibition of complex I and reported it as "similar to that observed with metformin".2 The 2014 paper ran metformin as its positive control — the known comparison you include to check your assay is working — and concluded that both berberine and metformin "promote glucose metabolism by stimulating glycolysis, which probably results from inhibition of mitochondrial respiratory chain complex I".3

Two molecules. One target. Same station on the same assembly line.

I want to be careful about what this does and does not establish, because it is the hinge of the episode.

It does not establish that berberine is as good as metformin. Same target does not mean same potency, same duration, same distribution in the body, or same result in a patient. Two keys can fit one lock and turn it different distances.

What it establishes is something narrower and more useful. The usual objection to a supplement is that we do not know whether the thing it targets actually matters — that the mechanism is a laboratory curiosity that will evaporate in a real patient. That objection is unavailable here. Complex I has been validated as a target for treating type 2 diabetes by the most heavily studied oral diabetes drug in the world, across decades, in hundreds of thousands of people. We are not asking whether jamming complex I lowers blood sugar and improves outcomes in human beings. We know it does. Metformin proved it.

The only open question about berberine is whether it reaches that target well enough, in a person, at a tolerable dose. That is a much smaller question than the one usually asked of supplements — and unlike most supplement questions, it has actually been tested.

The trial everybody cites

In 2008, a group in Shanghai did the obvious experiment. They took thirty-six adults with newly diagnosed type 2 diabetes and randomly assigned them to either berberine or metformin, half a gram three times a day, for three months.1

In the berberine group, HbA1c fell from 9.5% to 7.5%.

HbA1c is worth defining properly because it is the number this whole field lives on. Sugar in your blood sticks to haemoglobin, the protein inside red blood cells that carries oxygen. Once stuck, it stays stuck for the life of the cell, which is about three months. So measuring the percentage of your haemoglobin that has sugar attached gives you an average of your blood sugar over roughly the previous twelve weeks. It is a memory, not a snapshot, and it cannot be gamed by behaving well the morning of the test.

A drop of two full percentage points is not a subtle finding. For scale: when metformin's own extended-release formulation was tested against placebo in nearly a thousand patients, the treatment difference at the highest doses was about one percentage point.19 The authors of the berberine trial described its glucose-lowering effect as similar to metformin's, and on these numbers that is a defensible description.

Fasting blood glucose fell from 10.6 to 6.9 millimoles per litre. After-meal glucose fell from 19.8 to 11.1. Triglycerides — the main fat circulating in your blood — fell as well.1

The same paper ran a second study alongside the first: forty-eight adults whose diabetes was already poorly controlled on existing treatment, given berberine in addition. HbA1c fell from 8.1% to 7.3%. Fasting insulin dropped by 28%. A calculated measure of insulin resistance dropped by nearly 45%.1

Insulin resistance deserves its sentence. Insulin is the hormone that tells cells to take sugar out of the blood. In type 2 diabetes the cells stop listening properly, so the pancreas shouts louder — more insulin for the same effect — until it cannot keep up. Insulin resistance is the deafness. A treatment that lowers blood sugar and lowers circulating insulin at the same time is not forcing the system harder; it is making the cells listen again. That is the difference between treating the number and treating the disorder.

And now the honest part, because it is not a small one.

The authors called it a pilot study, and they were right to. Thirty-six people. Three months. One city. No placebo group in the head-to-head arm. Twenty of the fifty-eight patients across both studies — 34.5% — had transient gastrointestinal side effects. That is a real number and it belongs on the record. The trial also found no liver or kidney damage in any patient.1

On its own, this trial settles nothing. Nobody should change what they take on the strength of thirty-six people. But it does not sit on its own — and the claim that it does is how this literature usually gets dismissed.

The evidence base nobody mentions

When people say berberine rests on "one small trial", they are describing the trial that got translated into English and went viral, not the literature. The literature is considerably larger, and most of it is in Chinese-language journals that Western reviews routinely fail to search.

A 2014 systematic review pooled twenty-seven randomised controlled trials across 2,569 patients.4 A 2021 systematic review, searching eight databases including four Chinese ones, assessed forty-six trials.5 A 2018 review focused on cholesterol and triglycerides pooled sixteen trials across 2,147 participants.6

The 2021 pooled figures, against control: HbA1c down 0.73 percentage points. Fasting glucose down 0.86 millimoles per litre. After-meal glucose down 1.26. Fasting insulin down. Insulin resistance down. Body mass index down about one point. Triglycerides, total cholesterol and LDL cholesterol all down; HDL cholesterol up. Markers of inflammation improved. The authors' own summary sentence was that there is "strong evidence supporting the clinical efficacy and safety of berberine", particularly as an add-on therapy.5

Then there is a result I find genuinely striking, and it is buried in the 2014 review. Comparing berberine head-to-head against prescription lipid-lowering drugs, there was no significant difference in lowering total cholesterol or LDL — and berberine did better at lowering triglycerides and better at raising HDL.4

Those two, triglycerides and HDL, are precisely where statins are weakest. A statin is excellent at LDL and mediocre at the other two. If that finding holds up in better trials, berberine is not competing with statins; it is covering the ground statins do not.

The 2018 lipid review found total cholesterol down, LDL down, triglycerides down, HDL up — and on safety, no significant difference in adverse events between berberine and control, with no severe adverse effects reported in either group.6 Across the twenty-seven trials of the 2014 review, no serious adverse reaction was reported at all.4

Every one of those reviews also says, in its own words, that the included trials are of limited methodological quality — small, often unblinded, with poor reporting of how patients were randomised. That is true. I am not going to hide it and it is a genuine limitation.

But notice what kind of limitation it is. It is not a finding that berberine failed. It is a finding that nobody has run the good trial. And that is a different thing, with a different cause, which we now need to talk about directly.

Why the good trial does not exist

Here is the objection that ends most conversations about a natural compound: if it really worked, there would be a large trial proving it.

That sentence sounds like a statement about the compound. It is actually a statement about money, and once you see that, you cannot unsee it.

A cardiovascular outcomes trial — the kind that follows thousands of people for years and counts heart attacks, strokes, amputations and deaths rather than lab values — is the gold standard, and it is the standard the first version of this episode demanded of berberine. Such a trial costs somewhere in the hundreds of millions. Nobody runs one as a public service. It is run because a company holds a patent, and the patent gives it a protected window in which to recover the cost.

Berberine cannot be patented. It is a molecule that has existed in barberry and goldthread for as long as those plants have. Anyone can extract it. Anyone can sell it. There is no protected window, so there is no mechanism by which anyone recovers a nine-figure trial, so the trial does not happen — and this would be equally true if berberine were the best metabolic agent ever discovered.

That is the part that matters. The absence of the trial carries no information about the compound. It carries information about the patent system. Those are different facts about the world and we habitually confuse them.

So when you hear that a drug has "more evidence" than a supplement, the honest translation is usually: that drug had someone with a commercial reason to generate evidence, and this one did not. A thin file is not a verdict. Sometimes it is just an unpaid invoice.

The bioavailability puzzle, and the answer hiding in your gut

There is one genuine scientific oddity in berberine's story, and the sceptical version of this episode used it as a closing argument. It deserves better, because the resolution is more interesting than the objection.

The objection: berberine's oral bioavailability is poor. Bioavailability means the fraction of a swallowed dose that actually reaches your bloodstream intact. For berberine it is famously low — a few percent at most. Remember that permanent positive charge from earlier? This is the first place it bites. Charged molecules cross greasy cell membranes badly, and the cells lining your intestine actively pump berberine back out into the gut as fast as it gets in.

So the objection runs: if almost none of it gets into your blood, how can it be doing anything? And if it is doing something anyway, we do not understand this compound well enough to take it.

The first half of that is a fair question. The second half does not follow, and the answer to the first is now reasonably well developed.

The site of action may largely be the gut itself. A growing body of work points to berberine acting on the gut microbiota — the hundred trillion or so bacteria living in your intestine, which are not passengers but an active metabolic organ, fermenting fibre, manufacturing vitamins, and producing signalling molecules that reach your liver, your fat tissue and your brain.9

Berberine is a potent antimicrobial — that is what goldenseal was used for. Sitting in the gut at high concentration, it reshapes which bacterial species thrive, and that reshaping alters bile acid handling, short-chain fatty acid production and gut hormone release, all of which feed back onto blood sugar and blood fats.910 There is a further twist: gut bacteria themselves convert berberine into dihydroberberine, the better-absorbed form, so the microbes are partly responsible for whatever does make it into your blood.

If that is right, then berberine is not a badly absorbed systemic drug. It is a gut-acting agent that happens also to have a systemic component — and the low bioavailability everyone treats as a flaw is closer to a design feature.

And here is the detail that should make you sit up, given everything we have said: metformin appears to work substantially through the gut too. The parallel between these two compounds keeps going further than anyone expects it to.

Where berberine arguably wins outright

One more piece of evidence before we turn the scale around, and it is the one I would put in front of a sceptical clinician first, because it is a head-to-head against metformin in a condition where metformin is standard.

Polycystic ovary syndrome. PCOS is a common hormonal disorder in women in which the ovaries do not release an egg reliably, male-type hormones run high, and — in most patients — insulin resistance sits underneath the whole thing driving it. Metformin is routinely prescribed for it precisely because of that insulin resistance.

In 2011, eighty-nine women with PCOS and insulin resistance were randomised to berberine, metformin, or placebo, each alongside standard hormonal treatment, for three months.7

Compared against metformin — not placebo, metformin — the berberine group showed greater reductions in waist circumference and waist-to-hip ratio, greater reductions in total cholesterol, triglycerides and LDL cholesterol, and greater increases in HDL cholesterol and in something called sex hormone-binding globulin.7

Sex hormone-binding globulin — SHBG — is a protein that circulates in your blood and grabs onto sex hormones, holding them inactive. When SHBG is low, more free testosterone floats around unbound and active, which in PCOS drives the acne, the unwanted hair growth and the disrupted cycles. Raising SHBG lowers free testosterone without touching testosterone production at all. Berberine raised it more than metformin did.

A 2021 network meta-analysis of oral insulin-sensitising agents in PCOS lists berberine among the treatments "proven safe and efficacious" for the condition, alongside metformin, the inositols and the thiazolidinediones.8 That review found other combinations beat metformin on menstrual recovery; its honest summary is that metformin alone is not the best of the available options for several outcomes that matter to patients.

Eighty-nine women is still not a large trial. But it is a randomised head-to-head against the standard drug, and berberine did not merely hold its own.

The other pan of the scale

Everything so far has been berberine answering questions. Now we turn the scale around and ask the question the first version of this episode never asked.

What does metformin cost the person taking it?

Metformin's benefits are real and I am going to state them clearly before I state anything else, because an episode that only prosecutes is not an investigation. Metformin lowers blood sugar effectively. It has decades of cardiovascular outcome data behind it. It is inexpensive. In a retrospective study of more than ten thousand patients with diabetic kidney disease — a group who are usually told to avoid it — metformin use was associated with lower all-cause mortality and slower progression to end-stage kidney failure, with exactly one case of metformin-associated lactic acidosis recorded in the entire cohort.20 The lactic acidosis fear that dominates prescribing is, on that evidence, substantially overstated.

So: a good drug. Genuinely. Now the cost.

Metformin depletes vitamin B12, and this is not in dispute

In 2010 the BMJ published a randomised, placebo-controlled trial that ran for four and a third years in 390 patients with type 2 diabetes. Metformin or placebo, three times a day.11

I want to emphasise the design. This is not a survey. It is not a database trawl where sicker people happened to be taking more drugs. It is the same instrument — randomised, placebo-controlled — that is demanded of supplements and almost never demanded of a drug's harms.

The result: vitamin B12 concentrations fell by 19% relative to placebo. The absolute risk of outright B12 deficiency was 7.2 percentage points higher in the metformin group. The absolute risk of low-but-not-yet-deficient B12 was 11.2 points higher. Homocysteine — a compound that accumulates when B12 runs short, and that is itself associated with cardiovascular disease — rose.11

The authors expressed it as a number needed to harm of 13.8.

That phrase is worth unpacking, because it is the cleanest single number in this episode. Number needed to harm is how many people you have to treat before you cause one additional case of the harm. For every fourteen people on metformin for four years, one of them gets a vitamin B12 deficiency they would not otherwise have had.

Fourteen. Not fourteen hundred. That is not a rare idiosyncratic reaction. It is a predictable pharmacological consequence, and it happens to roughly one patient in every crowded waiting room.

The mechanism is understood. B12 is absorbed at the very end of the small intestine, in a stretch called the terminal ileum, by a process that requires calcium. Metformin interferes with that calcium-dependent step. The vitamin arrives and is not let through.

And this was confirmed over a much longer horizon. The Diabetes Prevention Program Outcomes Study followed participants for thirteen years. Low B12 was significantly more common on metformin at five years. Combined low and borderline-low B12 was more common on metformin at five years and still more common at thirteen. Each additional year of metformin use raised the odds of B12 deficiency by about 13%.12

Each year. It compounds.

The part that should be better known than it is

Now the sentence that, for me, changes the character of this whole comparison.

When vitamin B12 runs low, one of the first things to suffer is your peripheral nerves — the wiring running out to your hands and feet. B12 is required to maintain myelin, the fatty insulation wrapped around nerve fibres that lets a signal travel fast and cleanly. Insulation degrades, signals scramble.

What that feels like: numbness and tingling in the feet, usually starting at the toes and creeping upward symmetrically on both sides. Burning. Pins and needles. Loss of position sense, so balance gets worse in the dark. Eventually pain, or loss of sensation altogether.

Now read that list again, because it is also, word for word, the description of diabetic peripheral neuropathy — the nerve damage caused by years of high blood sugar, which is one of the most feared complications of the disease metformin is prescribed to treat.

They are clinically indistinguishable at the bedside. Same distribution, same symmetry, same progression, same description from the patient. A doctor cannot tell them apart by looking, by asking, or by testing sensation with a monofilament.

And here is the consequence. A person with type 2 diabetes who has been on metformin for six years walks in and says their feet are numb. There is an explanation sitting right there that fits perfectly, that everyone expects, and that requires no further investigation: diabetic neuropathy. The chart gets that word written in it. Tight glycaemic control is reinforced. Perhaps gabapentin is offered for the pain.

Nobody orders a B12 level. Why would they? The diagnosis is obvious.

Except that in some of those patients it is not diabetic neuropathy at all. It is a drug side effect that is treatable — B12 deficiency caught early can be corrected with an injection or a sublingual tablet, and the nerve damage can improve. Diabetic neuropathy, by contrast, is largely irreversible. One of these is a repairable problem being filed under the name of an unrepairable one.

This is not speculation about what might happen. The thirteen-year follow-up study found exactly the fingerprint you would predict: neuropathy prevalence was higher in the metformin group among those with low B12 levels.12 And a separate randomised trial of B vitamins in diabetic neuropathy found that symptom improvement was greatest in patients with the biochemical markers of B12 deficiency — and was inversely related to metformin use.14

The reviews have been saying to check for years. A 2016 clinical review recommends annual B12 testing for anyone on long-term metformin.13 The authors of the 2010 BMJ trial wrote that regular measurement "should be strongly considered".11 The thirteen-year study concluded that routine testing "should be considered".12

Three separate research groups, across fifteen years, telling prescribers to run a cheap blood test. It is still not standard practice in most places. A harm that nobody checks for does not become a smaller harm. It becomes an invisible one — and an invisible harm is worth nothing at all in a comparison, because it never gets counted.

Now hold berberine to the same standard

The obvious next move is the one the supplement industry makes constantly: berberine works by the same mechanism and causes no nutrient deficiencies, therefore berberine is strictly better.

The first half of that is well supported. We spent a third of this episode establishing it.

The second half is where I have to stop and be straight with you, because this is the point at which it would be easy to sell you something, and we said earlier that the rule cuts both ways.

Nobody has looked.

Searching the published literature for any study measuring vitamin B12 or micronutrient status in people taking berberine returns nothing. Not a negative result — no result. The trials exist, the patients existed, and that blood test was never run.

So the honest statement is not "berberine does not deplete nutrients". The honest statement is: metformin's B12 depletion has been demonstrated in a placebo-controlled trial and confirmed over thirteen years; berberine's effect on nutrient status has never been measured either way.

That is a weaker claim than the supplement aisle makes. It is also, I think, still a decisive one, for three reasons — and I want to lay them out so you can judge them rather than take my word.

One: the mechanism does not transfer. The two compounds share a target inside the cell — complex I — but B12 depletion has nothing to do with complex I. It comes from metformin interfering with a specific calcium-dependent absorption step in the terminal ileum. That is a separate property of the metformin molecule, not a consequence of the shared mechanism. There is no reason from first principles to expect berberine to do it, because that is not the part they have in common.

Two: the trials would likely have caught something large. Across forty-six trials in one review and twenty-seven in another, with more than two thousand patients in each, no serious adverse reactions were reported.45 Anaemia and overt neurological symptoms are the kind of thing that shows up even when you are not specifically hunting for it. A dramatic depletion effect would probably have surfaced. A subtle one absolutely would not have, and could be sitting there unrecorded right now.

Three: an unmeasured risk and a measured one are not symmetrical — but they are also not the same as no risk. Metformin's B12 problem is a known quantity with a known size and a known fix. Berberine's is an open file. In a decision about what to take, a known harm you can screen for is in some ways safer than an unknown one you cannot. I am not going to pretend the open file is a clean bill of health.

Two claims that do not survive checking

Two further things are said constantly in this comparison, and both are weaker than advertised. We check the claims we want to be true with the same instrument as the rest.

Magnesium. You will very often hear that metformin depletes magnesium alongside B12. The evidence is genuinely mixed. A 2019 study of 929 patients in Dutch primary care found metformin use negatively associated with magnesium concentrations.16 But an earlier study measuring magnesium inside cells rather than in serum found the opposite — metformin independently associated with higher intracellular magnesium, which the authors suggested might be protective.17 Those two do not agree. Until they are reconciled, magnesium is a hypothesis, not a finding, and we will not use it as evidence.

Vitamin D. Sometimes added to the list of things metformin depletes. It was tested directly, in a post-hoc analysis of that same randomised placebo-controlled trial, and the answer was clean: metformin had no effect on vitamin D levels. The authors state flatly that their results show metformin does not lead to vitamin D deficiency.15

So the nutrient case against metformin is not a long list. It is one item, and that item is extremely well established. One well-proven harm is worth more in an argument than five vague ones, and padding the list with the vague ones is how a good case gets discredited by the first specialist who checks.

The side-effect comparison, run fairly

One more correction to the first version of this episode. It made much of berberine's 34.5% rate of transient gastrointestinal side effects — nausea, diarrhoea, cramping, bloating — and treated that as a mark against it, since berberine is often sold on the promise of "no side effects".

It is a fair correction of the marketing. Berberine plainly does cause gut upset in a substantial minority of people, and anyone starting it should expect that possibility.

But 34.5% was never compared to anything, and a number without a comparator is not evidence, it is an impression.

So: in a randomised trial comparing metformin formulations, gastrointestinal side effects — diarrhoea, dyspepsia and flatulence — were 40% with standard immediate-release metformin.18

Forty per cent for the drug. Thirty-four and a half for the supplement. The supplement is, if anything, slightly better tolerated — and it is the one with the reputation for being hard on the stomach, because it is the one whose side effects got reported as a scandal rather than as a footnote.

Both numbers, incidentally, come down substantially with slow dose escalation and taking the dose with food. That is true for both compounds and for the same reason.

The weakest link is not the molecule

I have spent this episode making a case for a compound. Now I want to separate that from a case for a product, because they are not the same thing and this is where berberine most often fails people in practice.

When a doctor writes a prescription for metformin, an enormous amount is settled by that act. The tablet contains the stated number of milligrams. It contains the compound named and not a related one. It has been made to a manufacturing standard, tested for contamination, and assigned a batch number that can be traced and recalled. None of that is glamorous and all of it is doing work.

A berberine capsule bought online carries none of those guarantees automatically. In most countries supplements are regulated as food, not as medicine, which means the manufacturer is responsible for their own quality control and nobody checks the label against the contents before it ships.

The specific failure modes, so you know what you are looking at:

• The label says an extract, not an alkaloid. "500 mg barberry root extract" tells you nothing about how much berberine is in it. "500 mg berberine HCl" does. If the actual alkaloid content is not stated in milligrams, you cannot dose it and you cannot compare it to a trial.
• The plant source is unstated. Berberine-bearing plants contain other alkaloids alongside it, and different species carry different ones in different proportions. Goldenseal in particular contains hydrastine as well as berberine, and the two behave differently.
• The trials used a standardised preparation and yours may not be. Every meta-analysis in this episode says the same thing in its limitations: that better trials using standardised preparations are needed. That limitation applies just as much to the bottle in your hand.

This is a real disadvantage and it is not one the mechanism can argue away. Berberine the molecule has a strong case. Berberine the supplement aisle is a lottery, and the difference between those two sentences is the single most practical thing in this episode.

What to do about it: buy a product that names the salt and the milligrams of alkaloid, look for third-party testing, and prefer a manufacturer that publishes a certificate of analysis for the batch. That is not a guarantee. It is the best available approximation of the thing a prescription gives you for free.

And it is worth saying what this costs. Berberine is inexpensive — the 2014 review explicitly noted that its low cost relative to first-line medicines makes it a plausible option for patients of limited means over long periods.4 That matters, and it is a genuine argument in its favour. But cheap and unverified is a different offer from cheap and verified, and you should know which one you are accepting.

Who should not take this

This section is mandatory on every episode of this show and it is never a footnote. A favourable verdict that skips it is an advertisement.

Not in pregnancy. Not while breastfeeding. Not in newborns. This is the firmest contraindication berberine has. Berberine displaces bilirubin — the yellow pigment that causes jaundice — from the albumin protein that normally carries it safely through the blood. In a newborn, freed bilirubin can cross into the brain and cause kernicterus, a permanent form of brain injury. Two berberine-containing herbs were banned in Singapore for three decades over exactly this concern, particularly in infants with G6PD deficiency, an inherited condition affecting red blood cells.23 A later clinical study found no organ toxicity in twenty adult patients and described the underlying evidence as conflicting — but "conflicting" is not "refuted", and the population at risk here is newborn infants. There is no version of this where the risk is worth taking.

Not with G6PD deficiency, at any age, for the same reason — berberine has been implicated in aggravating jaundice in people with this condition.23

Caution with existing liver disease. Reviews of goldenseal and its alkaloids flag possible liver toxicity, along with nerve and light-sensitivity effects, as areas where the safety data are genuinely thin.24 The clinical trials have not shown liver damage — the head-to-head trial specifically looked and found none1 — but an already-damaged liver is not the place to find out.

What it interacts with

This is the part most people taking berberine do not know, and it is the most likely way berberine actually hurts somebody.

Remember that permanent positive charge, and the pumps in your intestinal wall that push berberine back out? The main pump is called P-glycoprotein. It is a general-purpose bouncer that your gut uses to eject a wide range of foreign compounds before they can be absorbed. Berberine inhibits it.

Which means berberine does not just fail to get past the bouncer — it puts the bouncer out of action for everything else you swallowed. Drugs that are normally partly ejected are now fully absorbed, and their blood levels rise.

Measured effects: berberine raised the blood exposure of digoxin — a heart drug with a notoriously narrow margin between a working dose and a toxic one — to between 123% and 170% of control, dose-dependently. It raised exposure to cyclosporine A, an immunosuppressant used after organ transplants, by 62% to 96%.21 Modelling work predicts meaningful interactions with certain cancer drugs as well.22

A 70% rise in digoxin level is not a theoretical concern. That is the difference between a therapeutic dose and a dangerous one.

So: if you take digoxin, cyclosporine, tacrolimus, a transplant medication, a chemotherapy tablet, or any drug your doctor monitors with blood levels, do not start berberine without telling them. Not because berberine is dangerous in itself, but because it changes the dose of everything else.

And the obvious one: berberine lowers blood sugar. Stacked on top of insulin, a sulfonylurea such as gliclazide, or metformin itself, it can push blood sugar too low. That is hypoglycaemia — shakiness, sweating, confusion, hunger, and at the extreme, loss of consciousness. If you are on glucose-lowering medication and you add berberine, you and your prescriber need to be watching for it.

How you would know it was hurting you

In plain words, and what to do that day.

• Sweating, shaking, sudden hunger, confusion, a racing heart. That is low blood sugar. Take fast-acting sugar now, then tell your prescriber before the next dose. Do not simply take berberine again tomorrow and hope.
• Yellowing of the whites of your eyes or of your skin, dark urine, pale stools, pain under the right ribs. That is the liver. Stop, and get a liver blood test this week.
• Gut upset that does not settle within two to three weeks, or any diarrhoea severe enough to dehydrate you. Stop and reassess — this is the common one, and it should fade, not build.
• If you take digoxin: nausea, visual disturbance (classically yellow-green haloes), palpitations, a slow pulse. That is digoxin toxicity. Stop the berberine and get a level checked urgently.
• New numbness or tingling in the feet. On either compound. See the next section, because this is the one that gets misfiled.

What test to ask for, before and after

This is the practical heart of the episode, and it applies whichever of the two you end up taking.

Get a vitamin B12 level. Before you start, so you have a baseline, and once a year after. If you are already on metformin and have been for more than two or three years, ask for one at your next appointment. It is a cheap, routine blood test.

Two refinements worth knowing, because serum B12 is an imperfect test. If your level comes back in the low-normal range and you have symptoms, ask about methylmalonic acid — a compound that accumulates specifically when B12 is functionally short, and which often reveals a real deficiency that the plain B12 number missed. Homocysteine is the other, and it rose measurably in the placebo-controlled trial.11

Also reasonable at baseline: liver function tests, kidney function, and HbA1c so you can tell in three months whether the thing is working at all.

And if you have numbness or tingling in your feet and you are on metformin: ask specifically whether your B12 has been checked. Not whether you have neuropathy — whether your B12 has been checked. You may be the one person in fourteen.

A stop date

Nothing on this show is a thing you take forever without a reason to keep taking it.

Three months, then measure. That is the horizon on which berberine's effect on HbA1c has been demonstrated, and it is the length of time HbA1c itself averages over. If after three months your HbA1c, fasting glucose and lipids have not moved, berberine is not working for you and there is no case for continuing it.

If it has worked, the next review is at twelve months, with B12 included. Beyond that we are past where the trial evidence reaches, and you should know that you are past it.

A typical trial dose is 500 mg two or three times a day, taken with meals. Starting at once daily and building up over two to three weeks is how most people avoid the gut upset.

So where does this leave the comparison

Let me put both columns on the table and then say what I think.

They share a mechanism. Not metaphorically — both inhibit mitochondrial complex I, measured directly, with metformin run as the comparison in the same experiments.23 The target is already validated by metformin's own outcome data.

On the numbers that have been measured, they perform comparably. Head-to-head at three months, similar glucose lowering.1 Across dozens of trials and thousands of patients, meaningful falls in HbA1c, glucose, insulin resistance and lipids.456 On triglycerides and HDL, berberine beat the prescription lipid drugs.4 In PCOS it beat metformin on body composition, on lipids and on SHBG.7

Metformin has outcome data that berberine does not. Real, decades deep, counting deaths and heart attacks rather than lab values. This remains the single strongest argument for the drug and I am not diminishing it.

But that gap is not a measurement of berberine. It is a measurement of who could afford the trial. Berberine cannot be patented, so no sponsor can recover a nine-figure outcomes study, so it has not been run — and it would not have been run if berberine were twice as good as metformin. The silence is about the patent system.

And metformin's column has a cost entry that was simply missing before. One in fourteen people over four years acquires a vitamin B12 deficiency,11 compounding at roughly 13% more risk per year,12 which produces a neuropathy indistinguishable from the diabetic neuropathy everyone is already expecting — so it is attributed to the disease, and the test that would catch it is not ordered. A treatable problem filed under an untreatable name.

Against which berberine's honest entry is: we do not know, because nobody measured. Better than a proven harm, and not the same as none.

What I would actually say

Berberine is not a lesser substitute for a real drug. It is a compound acting on the same validated target, with a respectable and consistently positive trial record across more than four thousand patients, a safety profile that in head-to-head terms is no worse than metformin's and in gut tolerability is slightly better, and advantages metformin does not have on triglycerides, on HDL and in PCOS.

What it lacks is an outcomes trial, and it lacks it for a reason that has nothing to do with whether it works.

If you are choosing, the honest summary is this. Berberine is a reasonable first option for insulin resistance, prediabetes, PCOS and dyslipidaemia — the situations where you are trying to change a trajectory rather than rescue a crisis, and where the harm from being wrong is small and recoverable. It is a reasonable adjunct in established type 2 diabetes, which is where the pooled evidence is strongest.

If you have advanced type 2 diabetes with established cardiovascular disease, metformin's outcome data is a real asset and it would be irresponsible of me to wave it away. Take the drug — and then get your B12 checked every year, because the people who prescribed it to you very likely will not.

And whichever you take: nobody should be swapping one for the other without telling the person who prescribed it, because berberine changes the blood levels of other medicines and because two glucose-lowering agents stacked together can put you on the floor.

The first version of this episode said you would be swapping a known answer for an open question. I no longer think that is right. You would be choosing between a known answer that comes with a known and unmonitored cost, and a well-supported mechanism whose biggest trial nobody had a commercial reason to fund.

Those are both real options. Only one of them has been allowed to look like one.

Sources

Every paper referenced on air, in the order it comes up. Links go to the publisher via DOI.