Supplements and alcohol

In a US national survey cited by the Preventive Services Task Force in 2022, 52 percent of adults said they'd taken at least one dietary supplement in the past 30 days.¹ In December 2024 and January 2025, two US federal reviews of moderate drinking came out within weeks of each other and pointed in different directions.¹⁸¹⁹ In both, the evidence is sharper than the advice you'll hear: one supplement carries the Task Force's top grade, most carry no verdict at all, and the alcohol dispute turns largely on one comparison group. This lesson reads the evidence for each.

Nothing here says what to take or whether to drink; doses belong with a doctor or pharmacist.

What the Task Force graded

Lesson 1 left you with beta-carotene: cohorts in which people with high blood levels died less often, then ATBC and CARET, in which smokers given the pills got more lung cancer. A supplement trial tests the pill, not the food. So which pills pass their own test?

The US Preventive Services Task Force graded vitamins, minerals and multivitamins for preventing heart disease and cancer in 2022.¹ This course read the abstract and the Task Force's web page.

For two supplements it said no: "The USPSTF recommends against the use of beta carotene or vitamin E supplements for the prevention of cardiovascular disease or cancer. (D recommendation)" For beta-carotene "the harms of beta carotene supplementation outweigh the benefits"; for vitamin E "there is no net benefit". Both at moderate certainty.¹ The harm is lesson 1's: "an important harm of increased lung cancer incidence was reported with the use of beta carotene by persons who smoke tobacco or have occupational exposure to asbestos."¹

For multivitamins, and for every other single or paired nutrient, it said something different: "the current evidence is insufficient to assess the balance of benefits and harms". That's an I statement.

And the grades apply only to some people. The web page: "This recommendation applies to community-dwelling, nonpregnant adults. It does not apply to children, persons who are pregnant or may become pregnant, or persons who are chronically ill, are hospitalized, or have a known nutritional deficiency."¹ Most of the cases below where a supplement clearly helps sit in the group it leaves out.

Vitamin D and fish oil: one large trial, read closely

The largest test here of two popular supplements is VITAL. It randomised 25,871 US adults (men 50 and over, women 55 and over) to vitamin D, omega-3 fish oil (n-3), both, or placebos, and followed them for a median of 5.3 years.²³ This course read both abstracts.

A trial names its primary endpoints before it starts: the outcomes it was built to test. Everything else it counts is secondary. VITAL's primary endpoints were invasive cancer and "major cardiovascular events (a composite of myocardial infarction, stroke, or death from cardiovascular causes)".² A myocardial infarction is a heart attack.

Vitamin D missed both: 793 cancers against 824 on placebo, hazard ratio 0.96 (0.88 to 1.06), and 396 cardiovascular events against 409.² (A hazard ratio of 1 means no difference; brackets hold the 95 percent confidence interval.) The authors' conclusion: "Supplementation with vitamin D did not result in a lower incidence of invasive cancer or cardiovascular events than placebo."² Among its secondary endpoints, death from cancer came to 0.83 (0.67 to 1.02), an interval that just crosses 1.²

Fish oil missed both too: 386 major cardiovascular events against 419, hazard ratio 0.92 (0.80 to 1.06).³

Older people, and an older review

A Cochrane review by Bjelakovic and colleagues pooled 56 trials of vitamin D with usable death data, 95,286 people, in trials mostly of women over 70.⁴ This course read the abstract. It searched to February 2012, before VITAL.

Deaths were 12.5 percent on vitamin D and 12.7 percent without, a risk ratio of 0.97 (0.94 to 0.99). The authors: "Vitamin D3 seemed to decrease mortality in elderly people living independently or in institutional care", and "further placebo-controlled randomised trials seem warranted", because of dropouts and trials that didn't report deaths.⁴ In four of its trials, "Vitamin D3 statistically significantly decreased cancer mortality" (risk ratio 0.88, 0.78 to 0.98). The same review found a harm: "Vitamin D3 combined with calcium increased nephrolithiasis", which is kidney stones (risk ratio 1.17, 1.02 to 1.34).⁴

The Endocrine Society's 2024 guideline draws a similar line.⁵ This course read the abstract. "The panel suggests empiric vitamin D for those aged 1 to 18 years and adults over 75 years of age, those who are pregnant, and those with high-risk prediabetes." ("Empiric", in the panel's definition, means intake above the Dietary Reference Intakes, lesson 5's DRIs, without a blood test first; "suggests" marks a conditional, weaker recommendation.) And: "The panel suggests against empiric vitamin D supplementation above the current DRI to lower the risk of disease in healthy adults younger than 75 years."⁵

The US Office of Dietary Supplements is blunt: "Overall, clinical trials find that vitamin D supplements do not reduce the risk of developing heart disease or dying from it, even if you have low blood levels of the vitamin."⁶

Where a supplement is the answer

So far the trials failed to prevent chronic disease in adults not chosen for being short of anything. Now the people the Task Force's scope left out.

Folic acid, the clean case

In 1991 the UK's Medical Research Council reported a trial at 33 centres in seven countries.¹³ This course read the abstract. It recruited 1,817 women who'd already had a pregnancy affected by a neural tube defect, a group of serious birth defects the abstract lists as anencephaly, spina bifida and encephalocele, and randomised them to folic acid, a mix of seven other vitamins, both, or neither, around the time of conception.

Of 1,195 completed pregnancies with a known outcome, 27 had a neural tube defect: 6 in the two folic acid groups and 21 in the two without. The authors called it "a 72% protective effect (relative risk 0.28, 95% confidence interval 0.12-0.71)". The other vitamins showed "no significant protective effect". And they found "no demonstrable harm from the folic acid supplementation, though the ability of the study to detect rare or slight adverse effects was limited."¹³

Why did this work where vitamin D didn't? This course's reading: the outcome is tied directly to the vitamin. The Task Force's 2023 statement says "Many of these neural tube defects are caused by low folate levels in the body."¹⁴ The pill came on top of an unchanged diet, as beta-carotene did, but here it supplied what was missing.

The women in the 1991 trial were at high risk. For everyone else the Task Force reaffirmed its recommendation in 2023, at grade A, the highest: for "all persons planning to or who could become pregnant", with "high certainty that folic acid supplementation has a substantial net benefit to prevent neural tube defects in their offspring."¹⁴ This course read the abstract and the web page. The page leaves out people at higher risk, such as those with a previous affected pregnancy, and people "taking certain medications known to block the function of folic acid".¹⁴ The NHS lists who should get a higher dose, prescribed by a GP (family doctor), including anyone with diabetes or taking medicine for epilepsy or HIV, and warns: "Do not take cod liver oil or supplements containing vitamin A (sometimes called retinol) when you're pregnant."¹⁵ The doses are on those pages and belong in a conversation with a doctor or midwife.

Three more cases, each with a condition

Vitamin B12, for vegans and many older people. "Plant foods have no vitamin B12 unless they are fortified", and "Vitamin B12 deficiency affects between 3% and 43% of older adults", many of whom absorb it poorly from food; the Office of Dietary Supplements advises people over 50 to get most of it from fortified foods or supplements. "Metformin might reduce vitamin B12 absorption"; that's one for whoever prescribes it. And the sheet is plain about the marketing: "vitamin B12 doesn't provide these benefits in people who get enough B12 from their diet", the benefits being energy, athletic performance and endurance.⁶

Iron, but after a test. Iron treats a diagnosed deficiency. The reasons not to take it just in case: "Accidental overdose of iron-containing products is a leading cause of fatal poisoning in children under 6", and people with haemochromatosis, an inherited condition in which iron builds up to toxic levels, "should avoid using iron supplements and vitamin C supplements."⁶ Together those sentences make the case the Office of Dietary Supplements' page points to: test first, and treat what the test finds, with a doctor.

Vitamin D in a northern winter. The NHS says "everyone should consider taking a daily vitamin D supplement during the autumn and winter", and all year for some groups.¹⁶ That page was last reviewed in 2020, with a review due in 2023, so check it before relying on it.

The B12 sheet has the pattern: the benefit is for people who are short. More widely, as this course's summary rather than any source's: the supplements that clearly help are aimed at a named deficiency, or at a group with a known reason to be short, and the ones that failed were given to people not chosen for a shortage, to prevent diseases not known to be caused by one.

What the label has and hasn't been checked for

This section is about the US, the regulator this course read; it didn't read the UK or EU rules.

The law is the Dietary Supplement Health and Education Act of 1994, DSHEA. The FDA's own questions page, current as of 27 August 2026, says: "Under DSHEA, FDA does not have the authority to approve dietary supplements before they are marketed. Generally, a firm does not have to provide FDA with the evidence it relies on to substantiate safety before or after it markets its products".⁷ (There's an exception for a new ingredient not already in the food supply.) The same page adds:

• Dose. "there are no laws or regulations that limit the serving size of a dietary supplement or the amount of a dietary ingredient that can be in a serving". The manufacturer decides.⁷
• The claim. Some label claims must carry a disclaimer "informing consumers that FDA has not evaluated the claim", and saying the product isn't intended to "diagnose, treat, cure, or prevent any disease". Certain kinds of claim, health claims among them, do need review first.⁷

The Office of Dietary Supplements adds two things people assume. On testing organisations' seals: "These seals do not guarantee that a product is safe or effective." And on "natural": "Some all-natural botanical products, for example comfrey and kava, can harm the liver."⁸

What the gap lets through

Tucker and colleagues went through the FDA's database of tainted supplements from 2007 to 2016.⁹ This course read the abstract. The FDA had found 776 products with undeclared, unapproved drug ingredients. Nearly all were sold for sexual enhancement, weight loss or muscle building, and of 28 products warned about more than once, 19 had new unapproved ingredients at the later warning.⁹ That counts what the FDA caught. It isn't a rate for the shelf, and it says nothing about an ordinary vitamin.

Liver injury is the other documented harm. A 2017 summary of a research symposium, by Navarro and colleagues: "HDS-induced liver injury now accounts for 20% of cases of hepatotoxicity in the United States based on research data" (HDS is herbal and dietary supplements).¹⁰ That's a share of recognised liver-injury cases, not a risk to any one user. "The major implicated agents include anabolic steroids, green tea extract, and multi-ingredient nutritional supplements".¹⁰ A 2024 analysis of a US national survey run from 2017 to 2020 estimated that 15.6 million US adults had taken one of six botanicals linked to liver injury, turmeric and green tea extract among them, in the past month.¹¹ That's how many were exposed, not how many were harmed. Both papers read at abstract level.

Alcohol: what both sides agree on

In January 2025 the US Surgeon General, Vivek Murthy, issued an advisory on alcohol and cancer.¹⁷ This course read its summary, first section, absolute-risk section and recommendations. It finds "a causal relationship between alcohol use and increased risk for at least seven different types of cancer, including breast (in women), colorectum, esophagus, liver, mouth (oral cavity), throat (pharynx), and voice box (larynx)", contributing to "nearly 100,000 cancer cases and about 20,000 cancer deaths each year" in the US.¹⁷

The review more favourable to moderate drinking agrees on breast cancer. The National Academies of Sciences, Engineering, and Medicine (NASEM) found that "Compared with never consuming alcohol, consuming a moderate amount of alcohol was associated with a higher risk of breast cancer".¹⁸ It also says "Alcohol has been identified as a carcinogen in humans, and associations of alcohol with cancer are likely linear", though at moderate intake it could draw no conclusion for colorectal, mouth, throat, oesophageal or laryngeal cancer against never drinking.¹⁸ The other federal review, from the Interagency Coordinating Committee on the Prevention of Underage Drinking (ICCPUD), lists the same seven cancers as the Surgeon General.¹⁹

These sources measure a drink differently. NASEM's US drink is 14 grams of alcohol; the 2018 Global Burden of Disease paper's is 10 grams; Wood's paper, below, equates 100 grams a week with 12.5 units, so a UK unit is 8 grams (this course's arithmetic, as are the drink conversions that follow).¹⁸²⁵²⁶ This course didn't check the size of a Canadian standard drink.

And heavy drinking is harmful on every reading: in Zhao and colleagues' 2023 meta-analysis, 65 grams of alcohol a day or more (about four and a half US drinks) carried a relative risk of death of 1.35 against lifetime non-drinkers.²¹

How big, in people

Logic and Argument lesson 9 turned a relative risk into cases per thousand. The advisory gives absolute lifetime risks of any alcohol-related cancer, from an Australian cohort.¹⁷ For women: "approximately 16.5%" for those drinking less than one drink a week, "19.0%" at one drink a day on average, and "approximately 21.8%" at two.¹⁷

From 16.5 to 19.0 is 2.5 percentage points, about 2 or 3 more women in every 100 over a lifetime. As a relative change, 19.0 divided by 16.5 is about 1.15: 15 percent higher. From 16.5 to 21.8 is 5.3 points, about 32 percent higher. Both sums are this course's.

Now the men's figures, from the same advisory: about 10.0% below one drink a week, 11.4% at one drink a day, and 13.1% at two.¹⁷

Notice one more detail. The advisory's comparison group "included very light drinkers, but not non-drinkers to avoid the sick quitter effect."¹⁷ Hold that phrase; the whole dispute below turns on it.

The J-curve, and the case that it's real

In many cohorts light drinkers die less often than people who don't drink, and heavier drinkers die more often. Plotted against the amount drunk, the risk dips and then climbs: a J-curve.

Stockwell and colleagues pooled 87 cohort studies in 2016 and, before adjusting for anything, "replicated the classic J-shaped curve": light drinkers' relative risk of death was 0.86 (0.83 to 0.90).²⁰ In half a million Chinese adults, Millwood and colleagues' conventional analysis found the same shape for strokes and heart attacks in men: "men who reported drinking about 100 g of alcohol per week (one to two drinks per day) had lower risks of all three diseases than non-drinkers or heavier drinkers."²² Both read at abstract level.

The strongest recent statement of this side's case is NASEM's Review of Evidence on Alcohol and Health, released in December 2024.¹⁸ This course read its five-page Highlights in full and the findings, conclusions and summary of its chapter on all-cause mortality. Its design took on the best-known objection directly: "An important requirement for this review, unlike other reviews, was to have a comparison group that separated never drinkers from former drinkers to avoid “abstainer bias.”"¹⁸ It counted only studies from 2010 on. Its "moderate" is the old US limit: "two drinks or 28 grams of alcohol in a day for men and one drink or 14 grams of alcohol in a day for women".¹⁸

Its main finding, in its own words: "On the basis of a meta-analysis of eight eligible studies, there was a 16 percent lower risk of all-cause mortality among those who consumed moderate levels of alcohol compared with those who never consumed alcohol (RR = 0.84, 95%CI [0.81, 0.87])." And its conclusion: "compared with never consuming alcohol, moderate alcohol consumption is associated with lower all-cause mortality (moderate certainty)."¹⁸ For heart disease: "consuming moderate amounts of alcohol is associated with a lower risk of CVD mortality in both females and males", also at moderate certainty.¹⁸

It states its limits. "No conclusions in the report were issued with high certainty." The comparison "mainly reflected alcohol consumption toward the lower end of the range defined as moderate consumption versus never consuming alcohol." And the report "does not offer dietary recommendations or advice."¹⁸

Two other bodies see a benefit for some people. The UK's Chief Medical Officers, in 2016, found "That the net benefits from small amounts of alcohol are less than previously thought (with substantial uncertainties around the level of protection) and are significant in only a limited part of the population. That is women over the age of 55".²⁹ And the Global Burden of Disease group, in its 2022 paper, found that for people aged 40 and over "the burden-weighted relative risk curve was J-shaped for all regions".²⁷ This course read that abstract.

One outcome keeps coming back. Wood and colleagues' 2018 analysis of 599,912 current drinkers found alcohol "roughly linearly associated with a higher risk" of stroke, heart failure and fatal hypertensive disease, but "By contrast, increased alcohol consumption was log-linearly associated with a lower risk of myocardial infarction" (falling steadily as intake rose).²⁵ This course read the abstract. On this course's reading, much of the protective case on the heart rests on that heart attack line; NASEM's rests on all-cause and cardiovascular mortality against never-drinkers. This course didn't read the proponents' own account of a mechanism, so it reports only the finding.

This is the shape the VITAL box warned about: one outcome going the other way from the rest. And because Wood compared drinkers only with other drinkers, the sick-quitter problem below can't touch it, though for the same reason it can't say whether drinking beats not drinking (this course's reading of the design). The same paper found that, against drinking up to 100 grams a week, those drinking more than 350 (about 25 US drinks) had "lower life expectancy at age 40 years of approximately" 4 to 5 years, and concluded that "These data support limits for alcohol consumption that are lower than those recommended in most current guidelines."²⁵

The case that it's an artefact

The comparison group. People who don't drink include people who stopped, some because they were ill. That's the sick quitter problem, also called abstainer bias. Stockwell's pooled cohorts show it: former drinkers had a relative risk of 1.22 (1.14 to 1.31). "However, low-volume drinkers may appear healthy only because the 'abstainers' with whom they are compared are biased toward ill health." After adjusting for this and for study quality, the light drinkers' 0.86 became 0.97 (0.88 to 1.07), an interval that includes 1.²⁰ Jürgen Rehm, one of ICCPUD's authors, put the general point in a WHO Europe news item: "Potential protective effects of alcohol consumption, suggested by some studies, are tightly connected with the comparison groups chosen and the statistical methods used, and may not consider other relevant factors".²⁸

Zhao and colleagues updated this in 2023 with 107 cohorts.²¹ Against lifetime non-drinkers, low-volume drinkers (1.3 to 24 grams a day, up to about 1.7 US drinks) had a relative risk of 0.93, P = .07: just over the usual .05 bar (lesson 3 glossed P values). Their conclusion: "daily low or moderate alcohol intake was not significantly associated with all-cause mortality risk".²¹ It's "not significantly associated", not "no benefit": the point estimate is still below 1.

Healthy drinkers. Light drinkers differ in other ways. Biddinger and colleagues, in the UK Biobank: "Light to moderate alcohol consumption was associated with healthier lifestyle factors, adjustment for which attenuated the cardioprotective epidemiologic associations with modest intake."²⁴ Read at abstract level.

The genes. This is lesson 1's Mendelian randomisation. In East Asia two common gene variants "alter alcohol metabolism" and strongly affect how much people drink.²² If light drinking really protects, men whose genes make them drink a little should do better than men whose genes make them drink nothing.

Holmes and colleagues, pooling 261,991 people of European descent, found carriers of a variant linked to drinking less had lower odds of coronary heart disease, 0.90 (0.84 to 0.96): "This suggests that reduction of alcohol consumption, even for light to moderate drinkers, is beneficial for cardiovascular health."²³ The method has limits its own users state. Biddinger's team notes that "traditional approaches to genetic epidemiology assume a linear association and thus have not fully evaluated dose-response estimates of risk across different levels of alcohol intake", which matters when the question is a dip at low doses. Their non-linear analysis found "light alcohol intake was associated with minimal increases in cardiovascular risk, whereas heavier consumption was associated with exponential increases".²⁴ Both read at abstract level.

ICCPUD's draft report, published in January 2025 by K. Shield and five colleagues, is the fullest statement of this side's conclusion.¹⁹ This course read its Summary and its introduction on protective effects in full, and searched the rest. "Among the U.S. population, the risk of dying from alcohol use begins at low levels of average use." And: "In the United States, males and females have a 1 in 1000 risk of dying from alcohol use if they consume more than 7 drinks per week." Its Summary also reports that at one drink a day people had "an increased risk of liver cirrhosis, esophageal cancer, oral cancer, and injuries, but a lower risk for ischemic stroke", and women "a lower risk for diabetes mellitus".¹⁹ So its modelled evidence found protective associations too; its all-cause conclusion went the other way. Its own statement of method: "Most of the evidence evaluated in this report is from systematic reviews and meta-analyses of “observational” cohort studies. It did not include data from quasi-experimental (e.g., Mendelian randomization) studies or randomized controlled trials." Its introduction does cite them: "Mendelian randomization studies, including those using advanced methodologies, find no protection at low levels of consumption for ischemic heart disease, ischemic stroke, or diabetes mellitus".¹⁹

Why two federal reviews disagreed

Side by side, on this course's reading, most of the disagreement comes from the two reports' entry rules.

Each answers the other's strongest objection only partly. NASEM's never-drinker rule answers the sick quitter, but separating former drinkers doesn't remove the healthy-drinker difference Biddinger describes; NASEM itself rated the five studies in one of its analyses as having “some concerns” about bias, "primarily due to confounding".¹⁸ The genetic studies answer confounding, but the methods most used assume a straight line when the question is a dip, as Biddinger's team says, and, on this course's reasoning from lesson 1, variants that change how the body processes alcohol might reach the disease by a route other than the amount drunk.²²²⁴

No randomised trial has settled it. The Moderate Alcohol and Cardiovascular Health trial, MACH, was to randomise people to about one drink a day or none and count heart disease and diabetes. In June 2018 the NIH announced it would end its funding: "The decision is based on concerns about the study design that cast doubt on its ultimate credibility. This includes whether the study would effectively address other significant consequences of moderate alcohol intake, such as cancer."³¹ This course read the release through an Internet Archive copy.

"No safe level": where the words come from

You may have heard that the Lancet said there's no safe level of alcohol. The abstract of the 2018 Global Burden of Disease (GBD) paper in the Lancet, which this course read, doesn't use those words. It says "the level of consumption that minimises health loss is zero", with an uncertainty interval (GBD's version of a confidence interval) of 0.0 to 0.8 of its 10-gram standard drinks a week.²⁶ That's the lowest point of a risk curve combining 23 health outcomes.

The phrase itself, in official language, comes from the WHO's European office. Its news item of 4 January 2023 was headed "No level of alcohol consumption is safe for our health" and said that "when it comes to alcohol consumption, there is no safe amount that does not affect health."²⁸ This course read the news item, not the Lancet Public Health statement it summarises. The UK's Chief Medical Officers said something narrower in 2016: "there is no level of regular drinking that can be considered as completely safe in relation to some cancers."²⁹

Then the same research group changed its answer. The 2022 GBD paper weighted each disease by how common it is in each place and age group, reasoning that "the overall risk depends, in part, on background disease rates, which vary by region, age, sex, and year."²⁷ For people aged 15 to 39 the lowest-risk level ran from 0 to about 0.6 standard drinks a day, depending on the region; for those 40 and over the curve was J-shaped everywhere. The conclusion: "There is strong evidence to support recommendations on alcohol consumption varying by age and location."²⁷ This course's reading is that the change came from the weighting, a new method rather than new biology, and that the older-age dip rests on the heart attack association, the outcome where the genetic evidence is least settled.

What the guides say

• UK, 2016. "To keep health risks from alcohol to a low level it is safest not to drink more than 14 units a week on a regular basis", spread over three or more days if you do.²⁹
• Canada, 2023. A continuum of risk by drinks a week, and "No matter where you are on the continuum, for your health, less alcohol is better."³⁰ This course read the landing page only.
• US, as of September 2026. The 2025-2030 Dietary Guidelines, published in January 2026, dropped the old numeric limit for "Consume less alcohol for better overall health". Lesson 9 has the document.³²

On who shouldn't drink at all, they agree. The US text: "People who should completely avoid alcohol include pregnant women, people who are recovering from alcohol use disorder or are unable to control the amount they drink, and people taking medications or with medical conditions that can interact with alcohol."³² Canada: "When pregnant or trying to get pregnant, there is no known safe amount of alcohol use."³⁰ The UK: "the safest approach is not to drink alcohol at all".²⁹

And one safety point matters for anyone who drinks heavily and reads "less is better" as "stop today". The UK guidance: "Very heavy drinkers, who are likely to experience alcohol withdrawal, are advised to seek medical advice before they stop drinking."²⁹

Who paid, and who wrote

Who paid and who wrote is listed here so you can weigh it, not so you can dismiss a finding with it. NASEM's review was sponsored by the US Department of Agriculture; ICCPUD's was issued through the Substance Abuse and Mental Health Services Administration (SAMHSA), part of Health and Human Services; both GBD papers were funded by the Gates Foundation.^18192627 T. S. Naimi wrote for both ICCPUD and Zhao's meta-analysis, and Jürgen Rehm, an ICCPUD author, is the researcher WHO Europe quoted, so those sources aren't independent.¹⁹²¹²⁸ A House Oversight Committee majority report of January 2026 called the ICCPUD study biased; this course saw it only as a search result and can't say what it argues. As of September 2026 only the January 2025 draft of ICCPUD's report has been published. MACH was paid for in part by "private donations of $67.7 million raised to date by the FNIH", the Foundation for the NIH; the part of the release this course read doesn't name the donors.³¹ A 2020 analysis of the trial's correspondence by Mitchell, Lesch and McCambridge, researchers who study the alcohol industry, read at abstract level, does: "Alcohol companies agreed to fund the MACH trial to advance their commercial interests rather than to help answer a major scientific question."³³

What people get wrong

"Supplements can't hurt." Beta-carotene raised lung cancer in smokers, vitamin D3 with calcium raised kidney stones, and supplements are a recognised share of liver injury.¹⁴¹⁰

"An I statement means supplements don't work." It means the evidence can't settle the balance of benefit and harm. The Task Force uses D when it concludes that.¹

"If a vitamin is low in your blood, a pill will prevent the disease it's linked to." For folic acid and neural tube defects, yes. For vitamin D and heart disease, trials found no benefit "even if you have low blood levels of the vitamin".⁶

"The Lancet said there's no safe level of alcohol." The 2018 paper said the level that "minimises health loss is zero"; the phrase is WHO's, from a 2023 statement in The Lancet Public Health that WHO Europe's news item summarised; and the same research group's 2022 paper found a J-shape for people 40 and over.²⁶²⁷²⁸

"The National Academies found moderate drinking is good for you." It found an association at moderate certainty, a higher breast cancer risk, and it "does not offer dietary recommendations or advice."¹⁸

"Moderate drinking is proven good for the heart." No trial has tested it. NASEM's finding of lower cardiovascular mortality is an association at moderate certainty, and the genetic studies find no protection against stroke.¹⁸²²

"Science has now shown that any drinking shortens your life." Zhao's result for light drinkers was "not significantly associated" with mortality, with a point estimate below 1, and NASEM found lower all-cause mortality in moderate drinkers against never-drinkers, at moderate certainty.¹⁸²¹ The part both federal reviews accept at low doses is the cancer risk.¹⁷¹⁸¹⁹

Practice

Connections

Back. Lesson 1's beta-carotene trials sit behind the Task Force's D grade, and its Mendelian randomisation, with the pleiotropy condition, did the heavy lifting on alcohol. Logic and Argument lesson 9's absolute risk is the tool for the Surgeon General's figures, and Mental Fitness lesson 2's question about the comparison group is the alcohol dispute in miniature: light drinkers compared with whom?

Forward. Lesson 11 joins this course's questions for reading a claim about your own food, and says when it stops being a self-help question.

Go deeper

• NASEM, Review of Evidence on Alcohol and Health, free. Read here: the Highlights and the mortality chapter's findings and conclusions. The case for an association, with its limits stated.
• ICCPUD draft report, free. Read here: the Summary and the introduction on protective effects. Read it beside NASEM's.
• USPSTF, vitamin supplementation, 2022, in JAMA, with a matching page on the Task Force's site. Read here: the abstract and that page. Short, and the clearest example of a D grade beside an I.
• FDA, Questions and Answers on Dietary Supplements, free. Read here: the page, current as of 27 August 2026. The regulator on its own limits.

Sources

1. US Preventive Services Task Force; C. M. Mangione and colleagues, JAMA 327(23), 2022, pp. 2326 to 2333, doi 10.1001/jama.2022.8970. Read: the abstract, and the Task Force's web page (released 21 June 2022), source of the scope and harm sentences.
2. J. E. Manson and colleagues (VITAL, vitamin D), New England Journal of Medicine 380(1), 2019, pp. 33 to 44, doi 10.1056/NEJMoa1809944. Read: the abstract.
3. J. E. Manson and colleagues (VITAL, omega-3), New England Journal of Medicine 380(1), 2019, pp. 23 to 32, doi 10.1056/NEJMoa1811403. Read: the abstract.
4. G. Bjelakovic and colleagues, Cochrane review of vitamin D and mortality, 2014, CD007470, doi 10.1002/14651858.CD007470.pub3. Read: the abstract. Searches to February 2012.
5. M. B. Demay and colleagues, Endocrine Society guideline, Journal of Clinical Endocrinology and Metabolism 109(8), 2024, doi 10.1210/clinem/dgae290. Read: the abstract.
6. NIH Office of Dietary Supplements, consumer fact sheets on vitamin D (updated 8 November 2022), vitamin B12 (15 December 2023) and iron (17 August 2023). Read: through Internet Archive copies of 18 to 20 September 2026.
7. US Food and Drug Administration, "Questions and Answers on Dietary Supplements", content current as of 27 August 2026. Read: the page.
8. NIH Office of Dietary Supplements, "Dietary Supplements: What You Need to Know" (updated 4 January 2023). Read: through an Internet Archive copy of 18 September 2026.
9. J. Tucker and colleagues, JAMA Network Open 1(6), 2018, e183337, doi 10.1001/jamanetworkopen.2018.3337. Read: the abstract.
10. V. J. Navarro and colleagues, Hepatology 65(1), 2017, doi 10.1002/hep.28813. Read: the abstract, which summarises a research symposium.
11. A. Likhitsup, V. L. Chen and R. J. Fontana, JAMA Network Open 7(8), 2024, e2425822, doi 10.1001/jamanetworkopen.2024.25822. Read: the abstract.
12. NHS, "Warfarin", last reviewed 24 February 2026. Read: the page.
13. MRC Vitamin Study Research Group, Lancet 338(8760), 1991, pp. 131 to 137, PMID 1677062. Read: the abstract.
14. US Preventive Services Task Force; M. J. Barry and colleagues, folic acid reaffirmation, JAMA 330(5), 2023, doi 10.1001/jama.2023.12876. Read: the abstract and the Task Force's web page.
15. NHS, "Pregnancy vitamins and supplements", last reviewed 3 June 2026. Read: the page, which gives the folic acid doses and who needs the higher one.
16. NHS, "Vitamin D". Read: the page, last reviewed 3 August 2020, review due 3 August 2023.
17. Office of the Surgeon General, Alcohol and Cancer Risk, January 2025. Read: the web page, through an Internet Archive copy of 14 September 2026, and the advisory's Summary, section 1, absolute-risk section and recommendations. The absolute risks come from the Australian 45 and Up cohort.
18. National Academies of Sciences, Engineering, and Medicine, Review of Evidence on Alcohol and Health, 2025 (released December 2024), doi 10.17226/28582. Read: the Highlights in full, and the findings, conclusions and summary of Chapter 3, "All-Cause Mortality"; not every table.
19. K. Shield, K. M. Keyes, P. Martinez, A. J. Milam, T. S. Naimi and J. Rehm, Draft Report: Scientific Findings of the Alcohol Intake & Health Study for Public Comment, ICCPUD, SAMHSA/HHS, January 2025. Read: the Summary and the introduction on protective effects in full; the rest searched.
20. T. Stockwell and colleagues, Journal of Studies on Alcohol and Drugs 77(2), 2016, doi 10.15288/jsad.2016.77.185. Read: the abstract.
21. J. Zhao, T. Stockwell, T. Naimi, S. Churchill, J. Clay and A. Sherk, JAMA Network Open 6(3), 2023, e236185, doi 10.1001/jamanetworkopen.2023.6185. Read: the abstract.
22. I. Y. Millwood and colleagues (China Kadoorie Biobank), Lancet 393(10183), 2019, doi 10.1016/S0140-6736(18)31772-0. Read: the abstract.
23. M. V. Holmes and colleagues, BMJ 349, 2014, g4164, doi 10.1136/bmj.g4164. Read: the abstract.
24. K. J. Biddinger and colleagues, JAMA Network Open 5(3), 2022, e223849, doi 10.1001/jamanetworkopen.2022.3849. Read: the abstract.
25. A. M. Wood and colleagues, Lancet 391(10129), 2018, doi 10.1016/S0140-6736(18)30134-X. Read: the abstract, source of the 100 grams to 12.5 units equivalence.
26. GBD 2016 Alcohol Collaborators, Lancet 392(10152), 2018, doi 10.1016/S0140-6736(18)31310-2. Read: the abstract.
27. GBD 2020 Alcohol Collaborators, Lancet 400(10347), 2022, doi 10.1016/S0140-6736(22)00847-9. Read: the abstract.
28. WHO Regional Office for Europe, "No level of alcohol consumption is safe for our health", news item, 4 January 2023. Read: the news item. The Lancet Public Health statement it summarises was not read.
29. Department of Health and Social Care, UK Chief Medical Officers' Low Risk Drinking Guidelines, August 2016. Read: the weekly, single-occasion and pregnancy sections, and the gov.uk page.
30. Canadian Centre on Substance Use and Addiction, "Canada's Guidance on Alcohol and Health", 2023. Read: the landing page only.
31. National Institutes of Health, news release on ending MACH funding, 15 June 2018. Read: through an Internet Archive copy.
32. HHS and USDA, Dietary Guidelines for Americans, 2025-2030, January 2026, PDF. Read: the alcohol section in full, for this lesson's one line; lesson 9 reads the rest.
33. G. Mitchell, M. Lesch and J. McCambridge, "Alcohol Industry Involvement in the Moderate Alcohol and Cardiovascular Health Trial", American Journal of Public Health, 2020, doi 10.2105/ajph.2019.305508. Read: the abstract, via Europe PMC, 24 September 2026. Supports: that alcohol companies funded MACH, in the authors' words.