The Drink You Know vs. The Pill You Don't

By fast2future · July 31, 2026 · Build-in-public series

The short version

Two things you swallow. Two stories that look similar and are actually opposites.

Alcohol is the fundamental where the evidence genuinely moved over the last decade — the old “a glass of red is good for your heart” finding has been substantially undermined — and where the official guidance is now openly, unresolvedly contested. The science moved; the labels are still arguing about it.

NMN (a NAD⁺ precursor sold as an anti-aging supplement) is the frontier product where the legal status flipped in the last ten months — from “FDA says this isn’t a supplement” to “FDA says it is” — while the human efficacy evidence didn’t move at all. The label moved; the science stayed exactly where it was.

That difference is the whole point of this piece. A change in what a regulator calls something is not a change in whether it works. And a genuine shift in evidence is not the same as a settled verdict.

🌿 STREAM ONE — Alcohol: the fundamental where the evidence really did move

What the old story was

For roughly thirty years, the dominant finding was the “J-curve”: people who drank a little appeared to live longer than both heavy drinkers and non-drinkers. It came out of large observational cohorts and got popularized as the “French Paradox.” The proposed mechanism was plausible — light drinking raises HDL cholesterol and has some antithrombotic effect.

(We’re deliberately not quoting a specific relative-risk figure from the flagship 2006 meta-analysis. Two different numbers circulate in secondary sources for it, and we could not confirm which is correct against the original paper. The shape of the finding is well established; a specific decimal we can’t verify is not something we’ll put in your hands.)

What undermined it

The comparison group was contaminated. The “zero drinks” reference group in many old cohorts included people who had quit drinking because they got sick, plus people who never drank because they were already unwell. That makes abstainers look unhealthy for reasons that have nothing to do with abstaining — and makes light drinkers look protected by comparison. This is the “sick-quitter” or abstainer-bias critique.

Zhao J, et al., JAMA Network Open, 2023. 107 cohort studies, 4,838,825 participants, 425,564 deaths. When the authors adjusted for study quality — critically, whether the study used a clean lifetime-abstainer reference group (only 21 of 107 did) — the apparent protection largely evaporated:

• Low-volume drinking (1.3–24.0 g/day), unadjusted: RR 0.85 (95% CI 0.81–0.88)
• Low-volume drinking, fully adjusted: RR 0.93 (95% CI 0.85–1.01), p = .08 — not statistically significant

“In this updated systematic review and meta-analysis, daily low or moderate alcohol intake was not significantly associated with all-cause mortality risk, while increased risk was evident at higher consumption levels, starting at lower levels for women than men.” — Zhao et al., JAMA Netw Open 2023 (abstract conclusion, verbatim)

Read that carefully, because it’s easy to get wrong in both directions. A confidence interval crossing 1.0 means no significant association found — not “drinking a little is protective” (the old claim) and not “drinking a little is proven harmful” (the overcorrection). It means the effect the old studies saw does not survive better methodology.

Harm at higher intakes is clearer, and starts lower for women: men at 45–64 g/day RR 1.15 (1.03–1.28); women at 25–44 g/day RR 1.21 (1.08–1.36), rising to RR 1.61 (1.44–1.80) at ≥65 g/day.

Biddinger KJ, et al., JAMA Network Open, 2022 — the causal check. Mendelian randomization uses genetic variants that influence alcohol consumption as a natural experiment, which sidesteps the reverse-causation and lifestyle-confounding problems that plague observational cohorts. In 371,463 UK Biobank participants, per 1-SD increase in genetically predicted alcohol consumption: hypertension OR 1.28 (95% CI 1.18–1.39), coronary artery disease OR 1.38 (95% CI 1.10–1.74). No protective inflection point at any intake level.

“In this cohort study, coincident, favorable lifestyle factors attenuated the observational benefits of modest alcohol intake. Genetic epidemiology suggested that alcohol consumption of all amounts was associated with increased cardiovascular risk, but marked risk differences exist across levels of intake, including those accepted by current national guidelines.” — Biddinger et al., JAMA Netw Open 2022 (conclusion, verbatim)

That last clause matters: marked risk differences exist across levels of intake. Light and heavy are not the same thing, even in the study that found no safe floor.

The part that is genuinely settled: cancer

This is the least disputed piece of the whole picture. The International Agency for Research on Cancer classified alcoholic beverages as a Group 1 carcinogen (carcinogenic to humans) in 1988, and IARC Monograph Volume 96 (2010) reaffirmed the original sites — oral cavity, pharynx, larynx, esophagus, liver — and added colorectal cancer and female breast cancer as causally related.

In January 2023, WHO stated the position plainly:

“We cannot talk about a so-called safe level of alcohol use. It doesn’t matter how much you drink — the risk to the drinker’s health starts from the first drop of any alcoholic beverage.” — Dr Carina Ferreira-Borges, WHO/Europe, published on who.int, 4 January 2023

(Source-honesty note: we’re quoting WHO’s own published statement directly. The underlying journal piece — Anderson BO, et al., Lancet Public Health 2023;8:e6–e7 — was not accessible to us in full. So we quote WHO quoting itself, which is what we can actually stand behind, rather than a journal sentence we haven’t read.)

The part that is NOT settled — and almost nobody tells you this

Here is where most health coverage of alcohol, in both directions, is dishonest by omission.

Two US federal evidence reviews looked at alcohol and all-cause mortality — landing about a month apart, as parallel inputs to the same 2025–2030 Dietary Guidelines process — and reached opposite headline conclusions.

• The National Academies (NASEM) review, briefed to Congress in December 2024, concluded: “Based on data from the eight eligible studies from 2019 to 2023, the committee concludes that compared with never consuming alcohol, moderate alcohol consumption is associated with lower all-cause mortality (moderate certainty)” — RR 0.84 (95% CI 0.81–0.87). (We verified this directly against the primary source.)
• The ICCPUD (an HHS interagency committee) review, drafted January 2025, reportedly found no age group with a net mortality benefit, with risk rising progressively from low average intake.

Note the timing: these were not a debate where one side answered the other. They were two commissioned reviews running in parallel, published within weeks, feeding the same guidelines process, and disagreeing. They disagree largely because they included different studies. NASEM used a deliberately narrow set of eight it judged highest-quality; critics say that filter is what produced the protective result. NASEM’s own committee graded its finding “moderate certainty” and noted the underlying studies mostly captured the lower end of the moderate range.

We are not going to resolve that for you, because it is not resolved. Anyone telling you the science has definitively concluded that moderate drinking is beneficial is ignoring Zhao and the Mendelian randomization work. Anyone telling you the science has definitively concluded there is no benefit at any level is ignoring a serious congressionally-mandated federal review that found the opposite. On cancer, the evidence is clear and one-directional. On all-cause mortality at low intake, it is a live methodological dispute among competent people.

Where the guidance landed

Different countries with access to the same literature drew strikingly different lines, which is itself informative:

Canada (CCSA, 2023) — a continuum-of-risk framing, and by far the most conservative in the English-speaking world:

Plus: no more than 2 drinks on any single occasion. (We are not quoting a gram figure for a Canadian standard drink — the primary source didn’t state one, and “standard drink” is not internationally uniform. A US standard drink is 14 g of ethanol; do not apply that number to a non-US guideline.)

United States — the 2020–2025 guidelines defined moderate drinking as ≤2 drinks/day for men and ≤1 for women.

⚠️ A live item we could not fully verify. Multiple secondary sources report that the 2025–2030 Dietary Guidelines, published January 2026, dropped the specific numeric daily limits in favor of vaguer language, and reduced explicit cancer-risk wording relative to earlier drafts. We were unable to access the primary government guidance pages directly to confirm this ourselves. We’re surfacing this as reported and unconfirmed rather than either asserting it or hiding it — because if true, it’s a meaningful change to the guidance most Americans encounter, and because the two contradictory federal reviews above are the obvious context for why the language got vaguer. Check the current guidance yourself before relying on any number in this section.

A US Surgeon General advisory on alcohol and cancer was issued in January 2025, calling for updated cancer-warning labels. We are deliberately citing no figures from it — we could not access the primary document directly, and the numbers circulating in secondary coverage disagree with each other by a wide margin (three different case counts, two different death counts). We would rather give you no number than the wrong one.

🚩 Three framing traps in alcohol coverage — spot them and you can read any headline

1. Reporting a non-significant result as an effect. Zhao’s adjusted low-volume RR of 0.93 has a CI touching 1.01. Writing “7% lower mortality” from that is wrong. It’s a null.
2. Collapsing population fractions into personal risk. “X% of cancers are attributable to alcohol” is a population-attributable fraction. It tells you almost nothing about how much your risk changes from your intake.
3. Relative risk without the baseline. A large-sounding percentage increase on a small absolute risk is still a small absolute change. Any honest source gives you both.

And we owe you that third one about our own numbers. Every figure in this section — Zhao’s RRs, Biddinger’s ORs, NASEM’s RR — is a relative risk or odds ratio measured across large populations. None of them tell you your personal absolute risk, and none of them were sourced with the individual-level baseline data that would let us convert them for you. Treat them as what the direction of the evidence is, not as what will happen to you. The person who can put a number on your situation is a clinician with your history in front of them.

🔬 STREAM TWO — NMN vs NR: the label moved, the science didn’t

NAD⁺ is a coenzyme central to cellular energy metabolism. Its levels decline with age. The longevity-supplement thesis: take a precursor, raise NAD⁺, slow aging. The two commercial precursors are NR (nicotinamide riboside) and NMN (nicotinamide mononucleotide).

What the human trials actually found

This is the honest core, and it is not what the marketing implies. Every flagship NR trial that tested a hard metabolic primary endpoint missed it.

“Insulin sensitivity, endogenous glucose production, and glucose disposal and oxidation were not improved by NR supplementation.” — Dollerup et al., AJCN 2018 (verbatim)

Remie’s authors put their own result just as plainly: skeletal muscle mitochondrial function was not elevated by NR supplementation.

The Elhassan result is the one worth sitting with. NR raised NAD⁺ in blood — the biomarker the marketing sells — while failing to raise it in the muscle tissue that was the actual point, and producing no change in mitochondrial function. A moving biomarker is not a moving outcome.

NMN’s trials are smaller and mixed. Yoshino et al., Science 2021 (n=25, postmenopausal prediabetic overweight/obese women, 250 mg/d) met its primary endpoint — muscle insulin sensitivity improved; muscle glucose disposal was 25±7% greater after 10 weeks of NMN than before it (p<0.01). Single-site, women-only, one narrow clinical population. (Primary-verified: trial registration, sample size and sponsor on ClinicalTrials.gov NCT03151239; the effect size read directly from the NIH public-access full text, PMC8550608.) Yi et al., GeroScience 2023 (n=80) showed dose-dependent NAD⁺ increases. Katayoshi et al., Sci Rep 2023 (n=36) found serum nicotinamide rose but arterial stiffness only “tended to” improve — i.e. did not clearly reach significance.

The meta-analyses are where it gets decisive. Both of these we verified directly against the primary text:

“This systematic review on 8 small-scale RCTs involving mainly relatively healthy adults did not find short-term NMN supplementation improved markers of glucose control and lipid profile… Our findings do not support the use of NMN supplementation among general population to improve glucose and lipid metabolism.” — NMN meta-analysis, 8 RCTs, n=342 (PMC11557618)

“Current evidence does not support NMN and NR supplementation for preserving muscle mass and function in adults with mean age of over 60 years.” — NMN + NR skeletal muscle meta-analysis, 10 RCTs (PMC12022230)

No significant effect on skeletal muscle index, grip strength, gait speed, or chair-stand performance.

🛑 The claim that does not exist

There is no human trial — none — showing that NMN or NR extends lifespan, extends healthspan, or reduces the incidence of any disease or death.

Every human trial is a surrogate endpoint (NAD⁺ blood levels, insulin sensitivity, arterial stiffness, muscle mass) over 3 weeks to 12 weeks, in samples of 12 to 80 people. The lifespan and healthspan data exist in mice only. When you see longevity marketing, this is the gap it is stepping across.

The regulatory flip — and why it proves nothing about efficacy

Here is the timely part, and the reason this piece pairs with alcohol.

In November 2022, FDA took the position that NMN was excluded from the legal definition of a dietary supplement under FD&C Act §201(ff)(3)(B)(ii) — the “drug preclusion” or “race to market” clause, which excludes an ingredient that was authorized for investigation as a new drug before it was marketed as a supplement. A pharmaceutical company had an active NMN investigational new drug application. Amazon delisted NMN supplements in March 2023.

Then, reportedly on September 29, 2025, FDA reversed itself in response to a trade-association citizen petition, concluding NMN had been marketed as a supplement before the drug authorization, and therefore is not excluded. Follow-up letters to ingredient suppliers in December 2025 formalized it. NMN is back on the shelves.

Three things to hold onto:

1. “Excluded from the supplement definition” was never a ban, and the reversal is not an approval. Both the 2022 exclusion and the 2025 reversal are rulings about a definitional/timing question — when was this thing first marketed versus first investigated as a drug. FDA said nothing about whether NMN works, in either direction, at any point. Any product copy reading “FDA-approved NMN” or “FDA confirms NMN” would be false.
2. NR was never subject to this at all. ChromaDex’s NR cleared the New Dietary Ingredient notification pathway and holds a GRAS determination. The NR/NMN legal asymmetry was always about drug-preclusion timing — never about one being better evidenced than the other. As the trials above show, neither has hard-outcome human evidence.
3. A regulatory event generates marketing. A legal status change is a news hook, and news hooks sell product. The evidence base for NMN in July 2026 is the same evidence base it was in 2022.

⚠️ Source-honesty note, and it’s a significant one. We were unable to access the primary FDA and federal register documents directly. The regulatory timeline above is therefore assembled from legal and trade press that quotes those FDA documents (Venable LLP, Natural Products Association, NutraIngredients, Nutritional Outlook) — not from the FDA documents themselves. Multiple independent outlets converge on the same dates, which raises confidence, but this section is reported, not primary-verified, and we’re labeling it rather than letting it wear an authority it hasn’t earned.

Safety — short answer, honest answer

Across the trials above, NR and NMN were well tolerated with no serious adverse events at doses up to 2000 mg/d (NR) and 900 mg/d (NMN). That is genuinely reassuring as far as it goes.

How far it goes: no trial cited here ran longer than about 12 weeks. There is no long-term human safety dataset.

There is also an unresolved theoretical concern worth naming without inflating: NAD⁺ supports cell proliferation broadly, including in malignant cells, and some preclinical models raise the question of whether NAD⁺ precursors could support tumor growth. No human trial has been designed or powered to detect a cancer signal, so the human literature has neither confirmed nor ruled this out. It is a mechanistic open question, not a demonstrated harm. It is also exactly the kind of question that belongs to a physician who knows your history — particularly if you have an active cancer diagnosis or significant personal risk.

One more thing worth knowing before you buy

A 2021 analysis of 22 top-selling Amazon NMN brands reported that only about 14% met their label claim, roughly 64% contained under 1% of the claimed NMN, and 14% contained none at all.

That analysis was commissioned by ChromaDex — which sells NR and directly competes with NMN. We’re giving you the finding and the conflict of interest, because you need both to weigh it. Independent reporting has separately flagged label-claim failures across NMN products, which is consistent, but a competitor-funded study is not neutral third-party testing and shouldn’t be presented as such.

🪒 The which-is-which razor

The one-sentence version: with alcohol, less is better-supported than it used to be and the cancer link is not in serious dispute; with NAD⁺ precursors, a regulator changed a definition and nothing about your body changed at all.

🛑 We are not doctors, and this is not medical advice

We say this every time, and this piece is one where it carries real weight.

On alcohol — please read this part.

• If you drink heavily or daily, do not simply stop on your own. Abrupt alcohol withdrawal in a physically dependent person can cause seizures and delirium tremens, and it can be fatal. Medically supervised withdrawal exists precisely because this is dangerous. Talk to a doctor before changing a heavy or daily drinking pattern. This is the single most important sentence in this article.
• If you are pregnant or trying to become pregnant, this article is not your source — your clinician is.
• If you take medications (many interact with alcohol), have liver disease, a personal or family history of the cancers named above, a history of alcohol use disorder, or a mental health condition, your personal calculus is different from any population average and needs a professional who knows your history.
• If you’re worried about your own drinking, that concern deserves a real conversation with a real clinician, not an internet article. It is a completely ordinary thing to ask a doctor about, and asking early is easier than asking late.
• Nothing here is a reason to feel ashamed. People drink for reasons — social, cultural, celebratory, and sometimes for pain. The evidence changing does not make anyone a bad person, and this piece is not an accusation.

On NMN and NR.

• We are not telling you to take these, and we are not telling you to stop if you already do. We’re telling you what the trials found so you can decide with accurate information.
• If you have an active cancer diagnosis or significant cancer risk, the unresolved proliferation question above is a real conversation to have with your oncologist before starting a NAD⁺ precursor. Do not resolve that one from an article.
• If you take prescription medication or have a chronic condition, run any new supplement past your doctor or pharmacist. Supplements are not inert just because they’re sold without a prescription.
• If you have persistent fatigue, weakness, or symptoms you’re hoping a longevity supplement will fix, please get them diagnosed. Symptoms deserve a diagnosis, not a workaround.

Anything touching an actual medical decision belongs with a professional who knows your body, your history, and your medications. We can tell you what the studies say. We cannot tell you what you should do — and we won’t pretend otherwise.

🌱 The hopeful close

There’s something genuinely freeing in this pairing once you see it.

The frontier product costs money every month, has no human outcome data, and its biggest news in three years was a paperwork reversal. The fundamental costs nothing to act on and the honest guidance is mostly subtractive — a little less, and the decision is yours to make with your own doctor.

And underneath both of them, the things with the strongest evidence for a long, good life are still the same unglamorous ones: sleep, real food, movement and strength, sunlight, water, managed stress, and people who love you. None of them are for sale. None of them need a regulatory ruling. None of them need you to resolve an argument between two federal committees before you can start.

That’s not a consolation prize. That’s the actual finding — the best-evidenced things remain free, available today, and nobody’s advertising budget depends on you believing in them. Which is exactly why they’re so quiet.

Be well. Be honest with yourself. And talk to your doctor. 💪

Sources

Alcohol

• Zhao J, Stockwell T, Naimi T, et al. Association Between Daily Alcohol Intake and Risk of All-Cause Mortality: A Systematic Review and Meta-analyses. JAMA Netw Open. 2023;6(3):e236185. doi:10.1001/jamanetworkopen.2023.6185 (PMC10066463) — primary source verified
• Biddinger KJ, Emdin CA, Haas ME, et al. Association of Habitual Alcohol Intake With Risk of Cardiovascular Disease. JAMA Netw Open. 2022;5(3):e223849 (PMC8956974) — primary source verified
• IARC Monographs Volume 96 (2010), Alcohol Consumption and Ethyl Carbamate (NCBI Bookshelf NBK326557) — primary source verified; Volume 44 (1988) original Group 1 classification — secondary-sourced
• WHO/Europe, “No level of alcohol consumption is safe for our health,” 4 January 2023, who.int — primary statement verified. Underlying journal piece: Anderson BO, et al. Lancet Public Health. 2023;8:e6–e7 — not independently verified
• Canada’s Guidance on Alcohol and Health, CCSA, 2023, ccsa.ca — primary source verified
• National Academies of Sciences, Engineering, and Medicine, Review of Evidence on Alcohol and Health, Ch. 3 (NCBI Bookshelf NBK614690; congressionally briefed December 2024, volume catalogued 2025) — primary source verified
• ICCPUD Alcohol Intake and Health Study (draft, Jan 2025), SAMHSA — secondary-sourced
• 2025–2030 Dietary Guidelines for Americans — reported only; primary guidance pages not accessible to us at time of writing
• US Surgeon General Advisory on Alcohol and Cancer Risk (Jan 2025) — primary document not accessible to us; no figures cited from it here by deliberate choice
• Di Castelnuovo A, et al. Arch Intern Med. 2006;166(22):2437–2445 — not independently verified; no figures cited from it here by deliberate choice
• Rumgay H, et al. Lancet Oncol. 2021 — primary document not accessible to us; no figures cited from it here by deliberate choice

NAD⁺ precursors

• Dollerup OL, et al. A randomized placebo-controlled clinical trial of nicotinamide riboside in obese men. Am J Clin Nutr. 2018 — primary verified via PubMed
• Elhassan YS, et al. Nicotinamide Riboside Augments the Aged Human Skeletal Muscle NAD+ Metabolome. Cell Rep. 2019 — primary verified via PMC
• Remie CME, et al. Nicotinamide riboside supplementation alters body composition and skeletal muscle acetylcarnitine concentrations in healthy obese humans. Am J Clin Nutr. 2020 (PMID 32320006) — primary verified
• Martens CR, et al. Nat Commun. 2018 — not independently verified; described here only in general terms
• Yoshino M, et al. Science. 2021 — primary verified via NIH public-access full text (PMC8550608) and ClinicalTrials.gov NCT03151239
• Yi L, et al. GeroScience. 2023 — secondary-sourced
• Katayoshi T, et al. Sci Rep. 2023 — secondary-sourced
• NMN glucose/lipid meta-analysis, 8 RCTs, n=342 (PMC11557618) — primary source verified
• NMN/NR skeletal muscle meta-analysis, 10 RCTs (PMC12022230) — primary source verified
• FDA regulatory timeline (Nov 2022 exclusion → Sept 29, 2025 citizen-petition reversal → Dec 2, 2025 supplier letters) — primary FDA documents not accessible to us; assembled from Venable LLP, Natural Products Association, NutraIngredients, Nutritional Outlook. Reported, not primary-verified. Independent corroborating paper trail: NPA v. FDA, D.D.C. No. 1:24-cv-02479 (filed Aug 28, 2024; stayed Oct 24, 2024 pending the citizen-petition response; voluntarily dismissed after the Sept 2025 reversal) — the docket chronology matches every date above.
• 22-brand NMN label-claim analysis (2021) — ChromaDex-commissioned; conflict of interest disclosed in-text; secondary-sourced