Skip to content

Supplements

The biggest vitamin D trial ever run missed both of its endpoints

25,871 people, five years, 2,000 IU a day. No reduction in cancer. No reduction in cardiovascular events. A separate trial of monthly high-dose vitamin D returned a mortality hazard ratio of 1.04, which is nothing. Both of them enrolled people who were mostly already replete, and that is the flaw that keeps repeating in this field: giving a nutrient to people who are not short of it is the design that has been tested most and worked least.

The observational promise here was enormous. Low vitamin D and low blood omega-3 both track with dying sooner, consistently, across enormous cohorts. Then the trials arrived and mostly came back empty. That gap is the argument of this whole section, and it points at one conclusion: a supplement is worth taking when a number on your bloodwork is out of range, and not because you are a person.

Aaron Cuha’s position

This is opinion, not a finding. Everything else on this page is the evidence.

Stop taking supplements because you are a person. Take them because a number on your bloodwork is out of range. If you cannot name the marker a bottle is moving, you are not on a stack, you are on a guess that happens to live in a bottle.

The reasoning

  • Mass supplementation is the thing the large trials keep failing to support. Vitamin D missed both primary endpoints in a 25,871-person trial and returned a mortality hazard ratio of 1.04 in another. In both, the population was already replete. Giving a nutrient to people who are not short of it is the design that has been tested most and worked least.
  • The exception is the person who is actually deficient, and you only know that from a blood test. My own panel is why I take what I take: omega-3 index at 3.9 percent against a target of 8 to 10, vitamin D at 30 against a target of 40 to 60, ferritin at 435 which is why I take no iron at all.
  • A marker gives you a stopping rule. Retest ninety days after starting anything, and if the number did not move, stop. Be loyal to the marker, not to the bottle. Without a marker there is no way to tell a working supplement from an expensive habit.
  • It is cheaper and it is safer. A targeted stack is a handful of things at doses tied to a number, not a cupboard. And every supplement is a drug with a dose-response curve that has a bad end: a single annual megadose of vitamin D increased falls by 15 percent and fractures by 26 percent in older women.
  • There is one thing I will still say is close to universally worth taking, because the gap is nearly universal and there is a marker for it: omega-3. Almost nobody eats enough oily fish to reach a reasonable omega-3 index, mine was 3.9 percent against a target of 8 to 10, and you can watch the number move. Even there, know what you are getting: the observational data are strong, the randomised outcome data outside one prescription formulation are mostly null, and there is a real dose-dependent atrial fibrillation signal.
  • I used to put CoQ10 in that sentence too. It does not belong there, and I would rather correct myself in public than keep a recommendation the evidence does not carry. The one strong randomised result is in people with moderate to severe heart failure. The most common reason people take it, statin muscle pain, has been tested seven times and pooled to nothing. So CoQ10 is a Tier 2 item tied to a diagnosis, not a floor item for everyone, and that is how it is listed below.

Where this goes beyond the evidence

Two places this runs ahead of what has been measured, and they are different sizes. The smaller one is omega-3, where the honest state is that blood levels track strongly with living longer across 17 cohorts, a general-population trial of 1 gram a day missed its cardiovascular endpoint in 25,871 people, and the one strongly positive outcome trial used 4 grams of a prescription purified EPA against a placebo that is itself disputed. I take it anyway, on the reasoning that a measurable deficiency with a cheap test and a low-risk correction is worth correcting. That is a judgement, not a trial result. The larger one is the whole test-then-treat rule, which is a policy rather than a finding. Nobody has randomised people to marker-guided supplementation against taking a multivitamin and counted what happened, so I cannot show you a trial where my method beat the alternative. What I can show you is that the alternative, giving nutrients to people who are not short of them, is the design that has been tested most and worked least. Choosing to act on a marker is an inference from those failures, and you should weigh it as one.

Who this is not for

Anyone using a supplement in place of a diagnosis. If a marker is out of range, the first question is why, and some answers to that question are conditions that need treating rather than topping up. This also does not apply to prescribed supplementation in pregnancy, malabsorption, kidney disease or documented deficiency syndromes, where the dosing is a clinical decision and not a personal optimisation.

Aaron Cuha, who built this archive, reads the trials behind every page on it, and runs these protocols on himself against quarterly bloodwork. He is a coach, not a doctor, and no clinical credential is claimed for him anywhere on this site. More about who writes this

Read the trials this rests on

What the large trials actually found

  1. The biggest vitamin D trial ever run missed both of its primary endpoints

    Measured in

    25,871 adults in VITAL, men aged 50 and over and women 55 and over, randomised to 2,000 IU of vitamin D3 daily or placebo and followed a median 5.3 years

    Neither primary endpoint was met. Vitamin D did not reduce invasive cancer and did not reduce major cardiovascular events. Mean baseline vitamin D was 30.8 ng/mL, which is replete, and only 12.7% of participants were below 20.

    What this does not show

    That vitamin D is useless, or that repletion does not matter in someone actually deficient. It shows that giving 2,000 IU to a largely replete population does not prevent cancer or heart disease. The widely quoted 22% reduction in autoimmune disease is a separate ancillary result that came in at p = 0.05 and then weakened to a confidence interval crossing 1.0 once two more years of case adjudication were applied, with no persistence after stopping.

    HumanVitamin D supplements and prevention of cancer and cardiovascular disease (2019)
  2. And the biggest vitamin D mortality trial found a hazard ratio of 1.04

    Measured in

    The D-Health trial, a randomised placebo-controlled trial of monthly high-dose vitamin D in older Australian adults, with all-cause mortality as the endpoint

    No mortality benefit. The placebo group already averaged about 31 ng/mL, which is the same problem VITAL had: you cannot demonstrate the benefit of repletion in a population that is already replete.

    What this does not show

    Anything about people who are genuinely deficient. That question was tested separately, in 8,851 Mongolian children who were 95.6% deficient. Supplementation tripled their blood levels and the primary endpoint was still null, which is the most uncomfortable result in this literature.

    HumanThe D-Health trial: a randomised controlled trial of the effect of vitamin D on mortality (2022)
  3. High intermittent vitamin D doses increased falls and fractures

    Measured in

    2,256 community-dwelling women aged 70 and over, randomised to a single annual oral dose of 500,000 IU of vitamin D or placebo

    The supplement group fell more, not less. Falls incidence rate ratio 1.15 and fractures 1.26. More is not safer, and in this trial more was worse.

    What this does not show

    That daily physiological dosing carries the same risk. This was a single enormous annual bolus, and the harm signal is replicated across other high intermittent-dose designs rather than resting on one trial. It is the clearest evidence in the field that a supplement is a drug with a dose-response curve, including a bad end.

    HumanAnnual high-dose oral vitamin D and falls and fractures in older women: a randomized controlled trial (2010)
  4. Two omega-3 outcome trials, opposite results, and the difference may be the placebo

    Measured in

    REDUCE-IT randomised 8,179 statin-treated patients with raised triglycerides to 4 g of icosapent ethyl daily against a mineral oil placebo. STRENGTH randomised a comparable population to a different high-dose omega-3 against a corn oil placebo.

    REDUCE-IT was strongly positive on major cardiovascular events. STRENGTH was null. The two are the most-argued pair in supplement evidence, and the leading explanation is not the drug: the mineral oil comparator in REDUCE-IT was associated with rises in inflammatory markers and LDL in the placebo arm, which would widen the gap between arms without the active drug doing more.

    What this does not show

    That REDUCE-IT is wrong, or that the mineral oil objection is settled. The defence of that placebo was published in a manufacturer-funded supplement with a manufacturer co-author and the lead author chairing the manufacturer's data monitoring committee, which is worth knowing when weighing it. What the pair does show is that a prescription high-dose purified EPA is not the same thing as a fish oil capsule, and neither result transfers to the other.

    HumanEffect of high-dose omega-3 fatty acids vs corn oil on major adverse cardiovascular events in patients at high cardiovascular risk (STRENGTH) (2020)
  5. Omega-3 raises atrial fibrillation risk, and the risk scales with dose

    Measured in

    VITAL-RHYTHM, the prespecified atrial fibrillation analysis within the 25,871-participant VITAL trial

    Omega-3 supplementation was associated with increased incident atrial fibrillation. Across the wider randomised literature the signal is dose-dependent, rising with each additional gram per day.

    What this does not show

    That omega-3 is net harmful, or that a low dietary dose carries the same risk. It does mean the common assumption that fish oil has no downside is wrong, and that a person already prone to arrhythmia has a specific reason to discuss dose rather than escalate it.

    HumanEffect of marine omega-3 fatty acid and vitamin D supplementation on incident atrial fibrillation (2021)
  6. The observational promise was enormous, which is exactly why the trials matter

    Measured in

    17 prospective cohorts pooled for blood omega-3 levels and mortality, and a separate individual-participant meta-analysis of vitamin D and mortality across a large consortium

    People with the highest blood omega-3 levels had roughly 15 to 18% lower mortality than those with the lowest. People in the bottom quarter of vitamin D had substantially higher mortality than the top. Both associations are large, consistent and replicated.

    What this does not show

    Cause. Low vitamin D is a marker of being indoors, being unwell, being sedentary and carrying more weight, all of which independently raise mortality. That is the whole lesson of this pillar: the observational signal for these two nutrients was as strong as observational signals get, and the randomised trials still mostly came back null.

    HumanBlood n-3 fatty acid levels and total and cause-specific mortality from 17 prospective studies (2021)
  7. CoQ10 cut events in heart failure and did nothing for statin muscle pain

    Measured in

    Q-SYMBIO randomised 420 patients with moderate to severe chronic heart failure to CoQ10 100 mg three times daily or placebo for two years. Separately, 7 randomised trials in 321 patients tested CoQ10 for statin-associated muscle symptoms.

    In heart failure the result was strong: major adverse cardiovascular events in 15% on CoQ10 against 26% on placebo, hazard ratio 0.50 (95% CI 0.32 to 0.80, p = 0.003), with cardiovascular mortality 9% against 16% and all-cause mortality 10% against 18%. For statin muscle pain, the pooled result was null, weighted mean difference minus 0.42 (95% CI minus 1.47 to 0.62), with no improvement in staying on the statin either.

    What this does not show

    That CoQ10 helps healthy people. Q-SYMBIO was 420 people with diagnosed moderate to severe heart failure on standard therapy, and its short-term endpoints at 16 weeks were null before the two-year outcome result appeared. There is no outcome trial of CoQ10 in healthy adults taking it for longevity, and the single most common reason people take it, statin muscle pain, is the thing it has been shown not to fix.

    HumanThe effect of coenzyme Q10 on morbidity and mortality in chronic heart failure: results from Q-SYMBIO, a randomized double-blind trial (2014)
  8. The statin muscle pain result, stated on its own because it is the common use case

    Measured in

    A systematic review and meta-analysis of 7 randomised placebo-controlled trials, 321 patients with statin-associated myalgia

    No benefit on muscle symptoms, weighted mean difference minus 0.42 with a confidence interval from minus 1.47 to 0.62. No improvement in the proportion of patients who stayed on their statin either. Only 2 of the 8 reviewed studies showed a positive effect individually.

    What this does not show

    That nobody responds. Pooled nulls can hide responders, and the trials were small, ranging from 37 to 76 participants. What it does mean is that if you are taking CoQ10 specifically for statin muscle pain, you are acting on a hypothesis that has been tested seven times and has not held up.

    ReviewEffect of coenzyme Q10 on statin-associated myalgia and adherence to statin therapy: a systematic review and meta-analysis (2020)

The method, in five steps

The ordering is the part people get wrong. They buy first and test never, which makes it impossible to know whether anything is doing anything.

  1. 01

    Get the panel before you buy anything

    A baseline that includes a full lipid panel with particle size, HbA1c and fasting insulin, high-sensitivity CRP, ferritin and a full iron panel, 25-hydroxy vitamin D, an omega-3 index, homocysteine, liver enzymes, and a thyroid panel. Without this you cannot tell a working supplement from an expensive habit, and you cannot tell a helpful one from a harmful one, which the iron entry above makes concrete.

  2. 02

    Start the floor, then wait

    Tier 1 goes in first and on its own. Adding eleven things in one week guarantees that if something helps or something upsets your stomach, you will never know which thing it was.

  3. 03

    Add only against a marker that is out of range

    One item per marker, and write the marker down next to it. If you cannot finish the sentence 'I am taking this because my ______ is ______', the bottle does not go on the list.

  4. 04

    Retest at ninety days and be loyal to the marker, not the bottle

    If the number moved, keep going. If it did not, stop, and do not assume the dose was too low. A supplement that has had ninety days and has not moved the one number it was chosen for has told you what it does.

  5. 05

    Re-examine the whole list twice a year

    Markers change, and the reason you started something can go away. A stack should get smaller over time at least as often as it gets bigger.

The owner’s own stack, with the marker beside each item

This is one person’s protocol, built from one person’s panel, published so you can see what the rule looks like when it is applied rather than described. It is not a recommendation for you, and the doses are tied to his numbers. The column that matters is the marker: if an item cannot name one, that is stated too.

Tier 1

The floor, for almost everyone

Four things, roughly sixty dollars a month at professional-grade brands. These are the ones the owner takes without waiting for a panel, on the reasoning that almost nobody eats enough oily fish, almost nobody gets enough sun, and creatine is the most-studied supplement in existence. Even here, get the baseline panel, because three of the four have a marker you can read and the number is what tells you whether the dose is right.

Omega-3, EPA and DHA2,000 mg a day with food
The marker
Omega-3 index, target 8 to 10 percent
What the evidence is
Strong observational data on blood levels and mortality. Randomised outcome data mostly null except for a prescription high-dose formulation, and a real dose-dependent atrial fibrillation signal.
Vitamin D3 with K22,000 to 4,000 IU of D3, with 100 to 200 mcg of MK-7
The marker
25-hydroxy vitamin D, target 40 to 60 ng/mL
What the evidence is
Two very large randomised trials found no benefit for cancer, cardiovascular events or mortality in already-replete populations. Repletion of genuine deficiency is a different question and is the only one this dose is aimed at.
Magnesium glycinate200 to 300 mg before bed
The marker
Red blood cell magnesium, not serum magnesium
What the evidence is
Serum magnesium is a poor test because the body defends it at the expense of tissue stores. The sleep and cognition claims are thin, and the most-quoted brain magnesium paper was retracted by its own authors.
Creatine monohydrate3 to 5 grams a day, every day, no loading phase needed
The marker
None. This one is not marker-driven.
What the evidence is
The most studied supplement there is, with a large randomised literature on strength and lean mass and an unusually clean safety record. Run it whether or not you lift.
Tier 2

Tied to a number on your panel

Nothing in this tier goes in until a specific marker says so, and each one comes back out if the marker has not moved in ninety days. This is where most people's cupboards should shrink. Every item below names its trigger.

Bergamot extract1,000 mg a day
The marker
Added when LDL particle pattern is small and dense, pattern B
What the evidence is
Randomised trials in humans exist and show lipid changes. They are small, largely from one research group in Italy, and measure lipid panels rather than events.
Berberine500 mg with each main meal, two or three times a day
The marker
Added when HbA1c or fasting glucose is elevated
What the evidence is
Multiple randomised trials against placebo and against metformin show glucose and lipid effects. Trial quality is mixed, most trials are small and from one region, and there is no outcome data.
Curcumin phytosome500 mg twice a day with food
The marker
Added when high-sensitivity CRP is up
What the evidence is
Absorption is the whole problem, and plain turmeric powder is not the same intervention as a phytosome or piperine formulation. Trials use specific formulations and the evidence does not transfer across them.
N-acetylcysteine600 mg once or twice a day
The marker
Added when liver enzymes are stressed
What the evidence is
Established as a prescription drug for paracetamol overdose and as a mucolytic. The longevity and general liver-support uses are extrapolation from that, not the same evidence.
Methylated B complexL-5-MTHF 400 mcg, methylcobalamin 1,000 mcg, P5P 20 to 25 mg, two or three times a week
The marker
Added only with an MTHFR variant, and adjusted on homocysteine at retest
What the evidence is
Lowering homocysteine with B vitamins is reliably achievable. Whether lowering it changes outcomes is the part large trials have repeatedly failed to show.
CoQ10100 mg three times a day with food, which is the trial dose and higher than most products supply in total
The marker
Added when there is a diagnosis of moderate to severe heart failure, under the care of the clinician managing it. Not added for statin muscle pain.
What the evidence is
The one strong randomised outcome result in this category, and it is narrow. Q-SYMBIO randomised 420 patients with moderate to severe heart failure on standard therapy and cut major cardiovascular events from 26 percent to 15 percent over two years. Its own sixteen-week endpoints were null first. For statin muscle pain, seven randomised trials in 321 patients pooled to no benefit and no improvement in staying on the statin.
No iron at allZero, including inside a multivitamin or a greens powder
The marker
Excluded because ferritin is high. The owner's is 435.
What the evidence is
This is the most important entry in the tier, because it is the one where the default behaviour of taking a general multivitamin actively causes harm. High ferritin does not need help. Get the lab before you ever take iron.
Tier 3

Optional, and honestly optional

These are the ones the owner runs and would drop first. They are here labelled as what they are rather than promoted, because a protocol that cannot say which parts are speculative is not a protocol.

Ashwagandha600 mg a day of a standardised root extract
The marker
None. Taken for stress and sleep.
What the evidence is
The product matters more than the compound here. Trials use standardised root extract, and in one analysis of 25 commercial products only two were confirmed root-derived.
Alpha-lipoic acid300 to 600 mg a day
The marker
Alongside glucose markers
What the evidence is
Best evidence is in diabetic neuropathy, at doses and in a population most users do not share.
Plant sterols and stanols2 grams a day with main meals
The marker
Alongside LDL
What the evidence is
Reliably lower LDL by a modest amount. No outcome trial has shown that this particular route of lowering it changes events.
Lycopene10 to 20 mg a day
The marker
None. Taken for prostate health.
What the evidence is
Observational association with prostate outcomes. Randomised data is thin and inconsistent.
Cinnamon extract500 to 1,000 mg a day of Ceylon, water-based
The marker
Alongside glucose markers
What the evidence is
Small glucose effects in small trials, with meta-analyses disagreeing. Genuinely optional.

The rest of this pillar

open now

What is actually in the bottle

The most-quoted statistic in this field was retracted for data fabrication. Two products sold as NMN contained no NMN. Only two of twenty-five ashwagandha products were made from the plant part the trials used. And what each certification seal does and does not verify, which is not what people think.

Open

Questions people ask about this

Does vitamin D actually do anything?
For preventing cancer, heart disease or death in people who are not deficient, the randomised answer is no. A 25,871-person trial missed both of its primary endpoints and a separate mortality trial returned a hazard ratio of 1.04. Both enrolled populations that were already replete, which is the design flaw that keeps repeating. Whether repletion helps a genuinely deficient person is a separate question with much weaker trial support than most people assume.
Which supplements are worth taking without a blood test?
Creatine has the largest and cleanest randomised literature of anything in the category and is not marker-driven. Omega-3 and vitamin D are on the owner's list because almost nobody eats enough oily fish or gets enough sun, but both have a marker you can read, so testing turns a guess into a measurement. Everything beyond those should wait for the panel.
How do I know if a supplement is working?
Name the marker before you start, then retest it at ninety days. If the number did not move, stop. This is the single practical rule on this page, and it is what separates a stack from a cupboard. Be loyal to the marker, not the bottle.
Is 59 percent of herbal supplements mislabelled?
No. That figure comes from a 2013 paper that was retracted in July 2024 after an institutional investigation found evidence of data fabrication. Label problems are real and are documented by other groups with other methods, but no market-wide mislabelling rate exists or can exist, because there is no mandatory product listing and therefore no denominator.
Does CoQ10 help with statin muscle pain?
Seven randomised trials in 321 patients say no. The pooled weighted mean difference was minus 0.42 with a confidence interval crossing zero, and it did not help people stay on their statin either. CoQ10 does have one strong randomised outcome result, but it is in moderate to severe heart failure, not in healthy people and not in statin myalgia.