Issues & changes, newest first.

If you get semaglutide from a compounding pharmacy rather than a branded pen, this one is worth two minutes of your time.

On 29 July, Apollo Care, LLC began a nationwide recall covering roughly 127,000 vials of compounded semaglutide, across eight different strengths of sterile multi-dose vial. The recall is still ongoing. The reason given is particulate matter in the vials — identified as nylon/polyamide and silk-like proteinaceous material. That is visible foreign material in a product people inject.

The FDA classes it Class II, meaning exposure may cause temporary or medically reversible harm. Not the most severe category, but not nothing either.

What this is not: it does not involve Ozempic or Wegovy, and it says nothing about semaglutide as a drug. Our semaglutide entry is unchanged — the evidence, the effect sizes and the safety profile of the molecule are exactly what they were yesterday. This is a manufacturing failure at one compounding pharmacy.

Why we’re recording it anyway: a large share of people taking semaglutide get it compounded, usually because the branded product is expensive. That route is legal and often sensible, but it carries a quality-assurance risk the branded pen does not, and this is what that risk looks like in practice. If you hold compounded vials, the lot numbers are in the FDA’s enforcement report.

The same week, a more serious one

Victory Medical Center Pharmacy recalled compounded injectable glutathione nationwide on 1 July after bacterial endotoxin came back out of specification — microbial contamination in a product that is supposed to be sterile.

That recall is Class I, the FDA’s most serious category, reserved for situations with a reasonable probability of serious harm or death. The quantity was small — 91 units — and again, the failure belongs to the compounder rather than to glutathione.

Our glutathione entry has always noted that IV glutathione supports a substantial clinic industry on limited controlled evidence. This is the sharper version of the point: the intravenous route runs through compounding pharmacies whose sterility assurance is not the same as a licensed manufacturer’s, and an endotoxin failure in something going into a vein is a different kind of problem from a capsule that absorbs poorly.

The pattern worth taking away

Both entries kept their ratings. Neither compound got worse this month. What changed is the evidence about the supply route — and for drugs people increasingly buy compounded to save money, the route is part of the risk. It is not an argument against compounding. It is an argument for knowing which pharmacy made what you are injecting, and for checking recall notices occasionally.

Also this week

The Lp(a)HORIZON trial has finished. The registry was updated on 28 August to mark it complete, with an actual completion date of 16 July. It is the first trial ever built to test whether lowering Lp(a) — an inherited cardiovascular risk factor carried by roughly one person in five, and one most people have never been tested for — actually prevents heart attacks and strokes. It enrolled 8,323 people.

No results have been released and no topline has been announced, so nobody outside the sponsors knows the answer yet. But it exists now, and it is being analysed rather than still being gathered. We have updated the Lp(a)-lowering therapies entry to reflect that, and the result will matter in either direction.

The database stands at 442 interventions.

Last week we told you the best human test of senolytics had narrowly failed. This week the idea gets a second attempt — through a different door than we said.

We have added RLS-1496 to the database at a D. It is a cream, and it is the first drug of its kind to be tested in people.

First, a correction. When we put RLS-1496 on our upcoming page last week, we described it as facing the same obstacle as foselutoclax — the senolytic that missed its primary endpoint and whose company then wound down. We said both had to be applied locally because the drug class damages platelets when given throughout the body.

That was wrong, and the reason matters. Foselutoclax blocks a protein called BCL-xL, which platelets need — hence the platelet problem, hence injecting it straight into the eye. RLS-1496 does something different: it targets GPX4, an enzyme that protects cells from a form of self-destruction called ferroptosis. Senescent cells lean on that protection far more heavily than healthy ones, so nudging GPX4 is meant to tip the old cells over while leaving the rest alone. Different pathway, no platelet issue. It is a cream because the conditions being treated are skin conditions, not because the drug is dangerous elsewhere.

We also missed that its first human results came out in March, five months ago. Our tracking was watching a later US trial and never checked whether an earlier European one had already reported.

What the trial actually found. A randomised, double-blind, vehicle-controlled Phase 1 in Europe treated three groups — plaque psoriasis, atopic dermatitis, and sun-aged skin — for one month. It met its safety goal, and there was a signal within four weeks: about 20% less epidermal thickening in psoriasis at the higher doses, and 25% of the eczema group reaching a meaningful drop in itch against none on the dummy cream. In sun-aged skin, collagen production rose and inflammatory markers fell, more so at higher doses.

Why it is still rated D. Every one of those numbers comes from a company press release, not a published paper that other scientists have checked. That is not a comment on the trial, which looks well designed — it is our standing rule. Sponsor announcements do not move an evidence grade, however good they sound. A larger US trial in precancerous skin lesions is due to finish this year.

And the honest ceiling: this is evidence about skin. It says nothing about lifespan, healthspan, or clearing senescent cells anywhere else in the body. If it works, it means senescent skin cells can be cleared with a cream — a real result, and a much smaller claim than the word “senolytic” tends to carry.

Also updated this week: creatine gained a 12-week randomised trial in 64 adults aged 45–65 — more lean mass and strength with or without exercise, though the memory and cognition findings in it are exploratory rather than proven. N-acetylcysteine got a clean null: 2,400 mg a day for 12 weeks did nothing for oxidative stress, insulin resistance, or liver fat in fatty liver disease. And NAD+ picked up a 90-day trial of a reduced form of NMN in healthy middle-aged adults, which raised blood NAD+ substantially — and, once again, produced only exploratory hints of anything a person would actually notice.

The database stands at 442 interventions.

We were missing the most informative human test the senolytic idea has had. That entry now exists: foselutoclax (UBX1325), graded C-.

Senolytics — drugs that kill off senescent “zombie” cells — are one of the most talked-about ideas in longevity. What people actually buy for it, fisetin and quercetin, rests on thin human evidence. So the obvious question is whether the idea has ever been properly tested in people. It has, once, seriously. It is worth knowing how that went.

What the trial found. ASPIRE was a Phase 2b trial in diabetic macular edema: 52 people, randomised, double-masked, with an active comparator — foselutoclax against aflibercept, the existing standard of care. Vision improved by 5.2 letters at 24 weeks and 5.5 at 36.

What most coverage says, and why it’s misleading. The result is usually reported as “encouraging,” with the detail that the drug was non-inferior to aflibercept at 9 of 10 timepoints. That sounds like a near-total success. But trials are judged on the endpoint they registered in advance, precisely so that results can’t be picked after the fact — and ASPIRE’s registered primary endpoint was the average of weeks 20 and 24. That is the one timepoint out of ten where non-inferiority was not established. The trial missed its primary endpoint. Counting the other nine is exactly the move pre-registration exists to prevent.

Then the company wound down. In May 2025 Unity Biotechnology cut its entire workforce, moved its executives into consulting roles, and began exploring a sale, licensing deal or dissolution. The programme stalled for financial reasons, not because of a safety problem.

Two things this says about senolytics generally. First, the drug had to be injected directly into the eye. That isn’t a quirk of the trial design — blocking Bcl-xL throughout the body drops your platelet count, so local delivery is a constraint the mechanism imposes. Second, even with that advantage — highest possible concentration exactly where it was needed — it did not beat an existing drug.

None of that makes senolytics a dead end. It does mean the honest summary of the class’s best human evidence is “narrowly missed, in one tissue, then defunded” rather than “shown to work.” Nothing here supports the idea that swallowing a senolytic supplement does something comparable.

Also added to Upcoming: Rubedo’s RLS-1496, a senolytic cream now in a small trial for sun-damaged skin, with results due about now. It is the second senolytic to reach a registered efficacy trial — and, like the first, it works by being applied straight to the tissue rather than taken systemically.

The database stands at 441 interventions.

A drug class that existed only as a theory in humans now has an approval.

On August 5 the FDA approved oveporexton, sold as Orzeyful, for narcolepsy type 1 in adults. It is the first orexin receptor agonist ever approved.

Narcolepsy type 1 is caused by the loss of orexin, a signalling molecule that keeps you awake. Every existing treatment works around that hole — amphetamines and modafinil stimulate the brain broadly, which raises alertness without touching the actual defect. Oveporexton replaces the missing signal directly. Our entry for it has carried that distinction for a while; what changed is that it is no longer a promise.

The approval rests on two 12-week randomised, placebo-controlled trials in 273 adults, which improved daytime sleepiness, cataplexy and quality of life. We have moved the entry from investigational to prescription and taken it off our upcoming readouts list, since the readout it was waiting for has arrived.

Two things worth being straight about.

First, the side effects are common. On the label, insomnia affected 55–60% of people versus 1% on placebo, and urinary frequency 53–58% versus 5%. None of it is dangerous — the liver toxicity that killed earlier drugs in this class did not appear here — but “frequent and annoying” is the honest description, and it is not what the headlines convey. Drug Enforcement Administration scheduling is still pending, expected within 90 days of approval.

Second, we had the dosing wrong. Our entry said once daily, carried over from an early description of the trial programme. The approved label is twice daily, and the pivotal trials tested 1 mg and 2 mg twice daily. That is now fixed.

What this does not mean. The approval is for narcolepsy type 1 and nothing else. There is no evidence at all for oveporexton as a wakefulness or cognitive aid in people who produce orexin normally — that use has never been tested, in anyone. The reason this entry matters beyond sleep medicine is narrower and more interesting than a new stimulant: it is the first demonstration that targeting this system in humans works. What that opens up is a research question, not a prescription.

Also updated this week: retatrutide — two more Phase 3 trials succeeded and the FDA filing is now guided to early 2027, while the newly formalised compassionate-use route turns out to have criteria narrow enough that almost nobody asking will qualify. Vitamin B6 picked up its first randomised signal for depression — real, but confined to a subgroup and not a reason to start taking high doses indefinitely, given the nerve-damage ceiling that entry already describes. Resistance training gained a rare piece of randomised evidence that a lifestyle intervention can slow an epigenetic ageing clock.

The database stands at 440 interventions.

Photobiomodulation’s cognition claim just got its first real clinical backing — in a narrow population.

Our red light therapy entry has always drawn a hard line: strong evidence for skin, hair, and wound healing, but only preclinical (mouse and cell) data behind the cognition and mitochondrial claims that show up in longevity circles.

That line just moved, for one specific group. A 2026 network meta-analysis pooled 35 randomized controlled trials, 1,559 people, comparing three non-invasive brain-stimulation methods — photobiomodulation, transcranial direct current stimulation, and transcranial magnetic stimulation — head-to-head in Alzheimer’s disease. Photobiomodulation came out ahead of the other two on both standard cognitive tests used in dementia research, and beat one of them by a statistically significant margin.

What this does and doesn’t change: the entry’s headline rating is untouched — that’s still built on the skin/hair/wound-healing evidence, which this study has nothing to do with. What changed is the cognition clause, which now reads “Moderate” for Alzheimer’s disease specifically, backed by real RCT data, instead of “Preclinical” across the board. For everyone else — healthy adults using red light for a cognitive edge, not a diagnosed dementia — the evidence is exactly where it was: mouse studies and mechanism, no human trials.

The distinction matters. This is a controlled clinical protocol in a diagnosed patient population, not a case for a consumer red-light panel doing the same thing at home.

The database stands at 440 interventions.

Omega-3’s atrial-fibrillation signal is real, but narrower than we described it.

Our omega-3 entry has carried a dose-dependent atrial-fibrillation warning for a while, based on a 34-trial meta-analysis: roughly a 12% increase in risk at about 1 g/day, rising to around 50% at 1.8–4 g/day. That framing implied meaningful risk at the doses most people actually take.

The updated analysis — 35 trials, 114,592 people — cuts the data by dose and by baseline cardiovascular risk, and the picture changes. A statistically significant increase in new-onset atrial fibrillation shows up in only one of the four groups:

• High dose (>1500 mg/day EPA+DHA) in people at high cardiovascular risk: odds ratio 1.43 (95% CI 1.14–1.79). In absolute terms that is 0.8 extra cases per 100 people.
• Low dose in high-risk people: not significant (OR 1.07).
• High dose in low-risk people: not significant (OR 1.03).
• Low dose in low-risk people: not significant (OR 1.06).

So the risk concentrates in people combining a high dose with existing cardiovascular disease — much closer to the prescription-level indications than to a 1 g/day capsule.

We did not move the rating. Safety stays at Low-Med. The signal is better resolved, not smaller in the population where it appears, and the entry still covers prescription-level dosing up to 4 g/day where the concern is real. What changed is the description, which was previously more alarming than the evidence supports at ordinary doses.

Worth noting who ran it: the lead author is at the Fatty Acid Research Institute, an omega-3-focused research organisation. A finding of harm from a group aligned with the compound is reporting against its own interest, which counts in its favour rather than against it.

Two smaller updates the same day. The cold plunge entry gained a 22-trial meta-analysis with a useful time course: cold water immersion doesn’t change maximal isometric strength, briefly reduces explosive power right after exercise (recovering within 24–48 hours), and modestly reduces soreness — with its apparent effect on muscle-damage markers shrinking once publication bias is accounted for. And the NAD+ entry gained a short trial of an oral NAD+ formulation that raised intracellular NAD in blood cells by 53% while leaving plasma NAD unchanged — another data point that raising NAD+ is the easy part.

The database stands at 440 interventions.

What happened. On July 23–24 the FDA’s Pharmacy Compounding Advisory Committee voted on seven peptides for the 503A bulk drug substances list — the list of ingredients compounding pharmacies are permitted to make. Six were recommended, one was rejected:

Peptide	Vote	Outcome
BPC-157	8–6	Recommended
KPV	8–6	Recommended
TB-500	8–6 (1 abstention)	Recommended
MOTS-c	7–5 (2 abstentions)	Recommended
Epitalon	7–5 (1 abstention)	Recommended, for insomnia
Semax	8–5	Recommended, for cerebral ischaemia, migraine, trigeminal neuralgia
DSIP (emideltide)	6–7 (1 abstention)	Rejected

What it does and does not mean. The committee advises; it does not decide. The FDA is not bound by the vote, and actually adding a substance to the list would run through notice-and-comment rulemaking. Even then, being on the 503A list means a compounding pharmacy may prepare the ingredient — it is not FDA approval, not a finding that the peptide works, and not a determination that it is safe. The FDA’s own review staff had recommended against the four peptides voted on the first day, citing poor characterisation, weak human efficacy and safety evidence, and immunogenicity risk that has not been assessed. Press coverage of the meeting also reported financial ties between committee members and the peptide industry.

What we changed. We updated the legal-status field and the notes on all seven entries to record the vote and that framing. We did not move a single score. Not one of these votes is new evidence: no trial reported, no safety dataset appeared, no efficacy question got answered. BPC-157 still has no published human randomised trial. MOTS-c’s first registered human trial is still recruiting with no results. What changed is who may legally prepare these compounds — and that is a fact about supply, not about whether they work.

That distinction is the whole point of scoring evidence and legal status in separate columns. If easier access moved our grades, the grades would be measuring the wrong thing.

A couple of other updates this week, both negative results worth knowing:

• Low-dose aspirin picked up a healthspan analysis from ASPREE — 19,114 older adults, followed a median of 4.7 years. Rate of “healthspan loss” (first heart disease, diabetes, dementia, cancer, or death) was 82.1 per 1,000 person-years on aspirin versus 82.7 on placebo. No meaningful difference. This is about primary prevention in initially healthy older adults; it says nothing about aspirin after a cardiac event, where the case remains strong.
• NAD+ precursors picked up a Phase II trial of nicotinamide riboside in older adults with mild cognitive impairment. NR did exactly what it is supposed to do biochemically — blood NAD doubled — and produced no improvement in cognition, total cerebral blood flow, or blood pressure over 12 weeks. It is a tidy illustration of the split this entry has always carried: raising NAD+ is reliable, turning that into a measurable outcome is not.

The database stands at 440 interventions.

PCSK9 inhibitors can now come in pill form. On July 16, the FDA approved Lipfendra (enlicitide), the first oral, once-daily PCSK9 inhibitor. Until now, this drug class — some of the most effective LDL-lowering medicines available, on top of a statin — has meant an injection every two weeks or every six months. Lipfendra is a once-daily tablet with efficacy in the same range: Phase 3 trials showed 55–60% LDL-C reductions, comparable to the injectable antibodies (evolocumab, alirocumab) already in our database. A head-to-head trial also found it beat two other oral cholesterol add-ons, bempedoic acid and ezetimibe, by a wide margin. We’ve added Lipfendra to the PCSK9 inhibitors entry; cardiovascular-outcomes data (does it actually prevent heart attacks, not just lower a number) are still pending, so the injectable options with a longer track record remain the guideline-preferred first choice for now.

A few smaller updates this week: two GLP-1-class drugs already in our database — cagrilintide and orforglipron — picked up new head-to-head trial data (cagrilintide’s combination with semaglutide beat semaglutide alone on blood sugar control; orforglipron beat a standard oral diabetes drug). And three peptides drawing an FDA compounding-pharmacy hearing this week — BPC-157, TB-500, and MOTS-c — each picked up their first-ever registered human clinical trial; none have reported results yet.

The database stands at 440 interventions.

This pass changed the priority from adding more interventions to validating the evidence already attached to them.

An identifier audit found that some PubMed links resolved to real papers but were unrelated to the intervention row where they appeared. Four bad associations were replaced immediately; 49 more confirmed mismatches are suppressed from public citation lists while their rows are reviewed claim by claim. Affected pages now say that a citation correction is in progress. Automated title matching also produced a larger review queue, but those flags will not be deleted automatically: papers can be relevant even when their titles do not repeat a compound’s name.

Two additional PubMed identifiers attached to FMT and Betaine HCl returned no document summary from NCBI and were removed. Both entries retain other live sources; the invalid IDs were not treated as evidence and no replacement claim was inferred.

Four high-use rows changed now:

• Vitamin D no longer presents 2,000–5,000 IU/day or a 40–60 ng/mL blood target as a general default. The row now distinguishes deficiency treatment from routine supplementation in generally sufficient adults, uses the adult 4,000 IU/day upper limit as context, and moves from A+ to B.
• Hydration / electrolytes no longer tells all active people to consume 3–5 g sodium/day. It separates ordinary hydration from individualized replacement during substantial sweat loss and notes that forcing extra water is not a proven longevity intervention.
• Psyllium had two unrelated citations replaced with a claim-matched LDL meta-analysis, received clearer swallowing/medicine cautions, and moved from A+ to A.
• Rifaximin had an erroneous hepatic-encephalopathy dose corrected from 1,100 mg twice daily to the US-labeled 550 mg twice daily, with the current official label added as a source.

The public archive will track the replacement work and any resulting score changes. Until that audit is complete, the project will not treat database size as a proxy for evidence quality.

Read Off Label presents evidence; it does not select an intervention for an individual.

This update removes the former grade-replacement override. When a non-toxin entry has a Safety score of 3.5 or lower, the page adds a high-risk evidence profile label, but it still shows the ordinary letter grade, Evidence, Benefit, and Safety axes, reported effects, legal status, and studied, labeled, or reported dose context. The label describes the score rather than replacing it.

Comparison pages now report which entry has the higher arithmetic composite and where the axes differ without naming a winner. Toxin pages use the descriptive term exposure concern in place of the previous directive priority terminology. The quiz uses the same underlying score axes plus reader-selected view filters, and can surface high-risk topics when they match those settings; those cards show their Safety score and link to the full evidence context. Pregnancy, lactation, preconception, symptom, sex, and tested-athlete answers add visible context without automatically hiding matching entries.

The governing rule is now explicit on the About and Methodology pages: benefits, harms, uncertainty, use context, and legality remain visible together, while the reader retains the decision.

A quiet week for new compounds, but one real evidence-quality correction worth flagging.

LDN’s best fibromyalgia data just turned negative. Low-dose naltrexone (LDN) has circulated in biohacker and chronic-pain circles for over a decade on the strength of small, short crossover trials. This week, the INNOVA study — a 12-month, randomized, double-blind, placebo-controlled trial in 98 women with fibromyalgia, and the longest, best-controlled test of LDN for this use to date — found no clinically meaningful benefit over placebo on pain intensity or any secondary outcome (function, mood, cognition, disability). The drug was well tolerated, with no safety signal, but the efficacy case for fibromyalgia specifically just weakened. We’ve revised that clause from “Moderate” to “Weak-Moderate” in the database and added the trial to the row’s notes. LDN’s other off-label uses (autoimmune symptoms, and its already-thin weight-loss framing as part of Contrave) are unaffected.

Smaller citation updates this week: a second systematic review confirmed clomiphene citrate and testosterone gel raise testosterone similarly in hypogonadal men, though TRT still wins on libido; older-patient subgroup data reinforced that ezetimibe and bempedoic acid both remain well-tolerated, effective add-ons in patients 70+ and 75+; and a large Spanish cohort (PREDIMED-Plus, n=2,664) linked higher dietary spermidine intake to better liver and metabolic markers in older adults with metabolic syndrome.

The database stands at 440 interventions.

The headline this week is a fatty-liver readout, plus a regulatory signal worth watching if you’re in the biohacker-peptide world.

Survodutide’s fatty-liver data is in. Boehringer Ingelheim and Zealand Pharma’s dual GLP-1/glucagon agonist already had strong topline obesity numbers (−16.6% body weight at 76 weeks). This week, the full SYNCHRONIZE-1 results were published in the NEJM, and — more notably — SYNCHRONIZE-MASLD, the class’s first dedicated Phase 3 trial in people with obesity and metabolic dysfunction-associated steatotic liver disease (MASLD, the modern name for fatty liver), landed in Nature Medicine. Six in ten participants reached liver-fat normalization at 48 weeks. Survodutide’s mechanistic pitch has always been that the glucagon arm burns hepatic fat directly, not just via weight loss — this is the first real trial built to test that specifically, rather than just being consistent with it. We’ve updated the effectiveness rating from “projected from Phase 2” to “confirmed in published Phase 3.” Two Phase 3 arms (SYNCHRONIZE-2 in people with type 2 diabetes, and the cardiovascular-outcomes trial) are still pending, so it stays on the imminent-readouts watchlist.

A biohacker-peptide regulatory signal. Ahead of the FDA’s July 23–24 Pharmacy Compounding Advisory Committee hearing, the agency posted its briefing documents — and its stated position on BPC-157, TB-500, MOTS-c, and KPV (all in our database) is to recommend against adding them to the legal compounding list, citing weak human evidence and unassessed immunogenicity risk. The committee’s vote is advisory, not final. The affected entries record this as evolving regulatory context rather than a final legal determination.

Smaller citation updates this week: an independent 262-trial network meta-analysis in the BMJ found tirzepatide loses the most fat mass of the approved anti-obesity drugs but also the most lean mass, while semaglutide was the only one in the comparison to show a reduced all-cause mortality signal outside its original SELECT trial.

The database stands at 440 interventions.

Three updates from this week’s evidence sweep — one frontier milestone, one approval we needed to catch up on, and one fresh Phase 3.

ER-100 — the first human dose of cellular reprogramming. On June 9, 2026, Life Biosciences (co-founded by Harvard geneticist David Sinclair) announced it had dosed the first patient in the Phase 1 trial of ER-100, its lead “epigenetic restoration” candidate. This is believed to be the world’s first clinical test of partial epigenetic reprogramming in a human — the approach that uses a controlled burst of three Yamanaka factors (OCT4, SOX2, KLF4) to reset a cell’s epigenetic “age” without erasing its identity. The trial targets optic neuropathies (open-angle glaucoma and NAION) and measures safety first, with visual-function endpoints. To be clear about where this sits: it is a safety study with no efficacy data, in a tiny number of patients, and reprogramming carries real theoretical risks (uncontrolled dedifferentiation, cancer). Our grade stays at the bottom of the scale — this is the most ambitious idea in the field, not a proven one. But “first human ever dosed” is a genuine milestone, so we’ve updated the entry.

Setmelanotide — a new approval we’d missed. On March 19, 2026, the FDA approved setmelanotide (Imcivree) for acquired hypothalamic obesity — severe obesity caused by damage to the hypothalamus from a tumor or its treatment — in adults and children aged 4 and up. It’s the first and only approved therapy for that condition, cleared on the Phase 3 TRANSCEND trial (an 18.4-percentage-point placebo-adjusted drop in BMI). Until now our entry listed only the rare genetic obesity syndromes setmelanotide was originally approved for; we’ve added the acquired-disease indication. Its FDA approval is restricted to named forms of hypothalamic or genetic obesity and does not cover general obesity treatment.

Retatrutide — first peer-reviewed Phase 3. Lilly’s triple agonist published TRANSCEND-T2D-1 in the Lancet (537 adults with type 2 diabetes): as a standalone treatment it cut HbA1c by 1.94 points and body weight by 15.3% at the top dose over 40 weeks. The grade was already Strong, so this doesn’t move it — but it’s the program’s first peer-reviewed Phase 3 paper, after a year of topline-only press releases. Citation added.

The database stands at 440 interventions.

A reader prompt — “find the next Adderall, or the compound that stops aging” — turned into three database additions. All three are genuine gaps, not hype.

SGLT2 inhibitors (canagliflozin / empagliflozin / dapagliflozin) — the “stops aging” entry. These diabetes/heart-failure drugs have quietly become one of the most serious geroscience bets after rapamycin and metformin. In the NIH-funded Interventions Testing Program, canagliflozin extended median lifespan in male mice by about 14% (Miller, 2020). A 2024 Nature Aging paper showed why it might: the drug acts as an indirect “senolytic,” helping the immune system clear worn-out senescent cells, and it extended lifespan in fast-aging mice even when started in middle age. And unlike most longevity candidates, the human data is already strong — SGLT2 inhibitors cut all-cause death and heart-failure and kidney events across multiple large trials, including in patients with an average age of 90. We’ve graded the longevity use Moderate (the human trials are in sick patients, and the lifespan data is in mice), while noting the cardio-renal evidence is Strong. It’s cross-listed under both Longevity and Metabolic Health.

Centanafadine — a non-stimulant “next Adderall.” Otsuka’s candidate raises dopamine and norepinephrine like a stimulant but isn’t a scheduled drug — potentially ADHD relief without the abuse-potential baggage. It’s backed by four positive Phase 3 trials across adults and children, with an FDA decision due July 24, 2026. Graded Strong.

Oveporexton (TAK-861) — the wakefulness frontier. Takeda’s first-in-class orexin-2 agonist restores the exact brain signal narcolepsy patients are missing, rather than stimulating broadly like amphetamine or modafinil. Its two pivotal Phase 3 trials hit every endpoint, with roughly 80% reductions in cataplexy. Approval (FDA decision expected in the third quarter of 2026) would be for narcolepsy only — but it’s the first proof that orexin agonism works in people, which is what makes it interesting well beyond that one disease. Graded Strong.

Both centanafadine and oveporexton now appear on the Watchlist under imminent readouts. The database is now 431 interventions.

A round of filling in compounds we’d flagged but not yet added.

Two more obesity drugs — the field keeps getting more crowded.

• Enicepatide (CT388) — Roche/Genentech’s injectable dual GLP-1/GIP agonist (same class as tirzepatide), from their Carmot acquisition. Its Phase 2 trial showed clinically meaningful weight loss and it’s moving to Phase 3. Graded Moderate (Phase 2).
• Elecoglipron (AZD5004/ECC5004) — an oral GLP-1 pill from AstraZeneca (licensed from Eccogene), competing with Lilly’s orforglipron. Both of its Phase 2 trials — one in obesity, one in type 2 diabetes — hit their endpoints. Graded Moderate (Phase 2). (We’d earlier mislabeled this as a Lilly drug on our internal watchlist — corrected here.)

Two cognitive additions.

• Mazindol ER (Quilience) — NLS Pharmaceutics is reviving mazindol, a 1970s appetite-suppressant stimulant, as an extended-release drug for narcolepsy and ADHD. It blocks dopamine/norepinephrine reuptake and partially activates orexin receptors — an unusual non-amphetamine profile. Its Phase 2 ADHD trial met all endpoints and the FDA cleared a Phase 3 narcolepsy program. Another “stimulant alternative,” graded Moderate.
• Sarcosine (N-methylglycine) — an NMDA-receptor co-agonist sold OTC as a nootropic. Small trials show it helps the negative symptoms of schizophrenia and depression as an add-on, but there’s no controlled trial of cognitive enhancement in healthy people. Graded Weak-Moderate — the evidence is psychiatric, not nootropic, and we say so.

One we left out: troriluzole. This glutamate modulator was a candidate, but the FDA just issued a Complete Response Letter rejecting it for spinocerebellar ataxia, and it had already failed its Alzheimer’s and OCD trials. An ataxia drug with negative cognitive data isn’t a fit here.

The database is now 440 interventions.

The American Diabetes Association’s 2026 Scientific Sessions (June 5–8, New Orleans) brought a wave of obesity-drug data. Most of the headline compounds are already in our rankings — retatrutide, CagriSema, survodutide, amycretin, eloralintide, mazdutide. One was missing.

Petrelintide (Roche/Zealand) is a long-acting amylin analog — it works through the satiety hormone amylin rather than the GLP-1 (incretin) pathway most weight-loss drugs use. In the Phase 2 ZUPREME-1 trial (485 adults, reported June 5), once-weekly petrelintide produced up to 10.7% body-weight loss versus 1.7% on placebo at 42 weeks, with notably gentle tolerability: nausea in about 20% (mostly mild), and only 1.5% of participants stopped because of gastrointestinal side effects. Phase 3 is planned for the second half of 2026.

That tolerability is the point. Amylin monotherapy is emerging as a potentially better-tolerated alternative — or a combination partner — to the GLP-1 drugs, and petrelintide now sits alongside cagrilintide as the database’s amylin entries. We’ve graded it Moderate evidence: solid Phase 2 data, but this is conference data, industry-sponsored, with the peer-reviewed publication still to come.

Also updated: bimagrumab. The activin-receptor antibody’s BELIEVE Phase 2b trial was formally published in Nature Medicine (2026). The combination of bimagrumab plus semaglutide drove 17.8 kg of weight loss versus 3.3 kg on placebo at 48 weeks — and, true to bimagrumab’s purpose, the weight lost was fat while lean mass was preserved. Its grade is unchanged; we’ve added the published figures and citation.

The database is now 428 interventions.

Continuing the geroprotector coverage, this update adds one repurposed-drug evidence profile and one preclinical lifespan signal with human toxicity evidence.

ACE inhibitors / ARBs (telmisartan, losartan, enalapril) — the other repurposed cardiometabolic geroprotector. Right after adding SGLT2 inhibitors, we realized the entire renin-angiotensin blocker class — some of the most prescribed, cheapest generic drugs on earth — was missing, and it has a real aging story. Angiotensin II is a direct driver of vascular aging, chronic inflammation, and cellular senescence; blocking it extends healthspan in animal models, and high-dose ARBs can actually reverse age-related kidney and blood-vessel scarring independent of their blood-pressure effect. Telmisartan doubles as a PPAR-γ activator (a metabolic bonus); losartan lowers TGF-β and improved aged-muscle repair in mice. And unlike most longevity candidates, the human hard-outcome data is overwhelming: these drugs cut mortality, stroke, heart failure, and kidney decline across decades of trials. We graded the longevity use Moderate (the lifespan data is animal; human trials are in patients with disease) and the cardiovascular use Strong. Cross-listed under Longevity and Metabolic Health.

NDGA (nordihydroguaiaretic acid) — a lifespan signal alongside liver toxicity. NDGA is one of the compounds that extended lifespan in the NIA Interventions Testing Program, but only in male mice, with no human longevity data. Its natural source, the chaparral (creosote) bush, has also caused documented cases of acute liver failure in people, and the FDA has issued warnings about chaparral supplements. The Weak grade and low Safety score display the mouse-lifespan signal and human toxicity evidence as separate parts of the profile.

We also noted mitochondrial transplantation on our internal watchlist — a fascinating frontier idea, but its only human use so far is in pediatric cardiac surgery, with no healthy-aging data yet.

The database is now 436 interventions.

Continuing the discovery hunt, three more rows — one straightforward, two cautionary.

Viloxazine (Qelbree) — an approved, non-scheduled ADHD drug. Already flagged as a gap, now added. Viloxazine raises norepinephrine (and indirectly dopamine) and affects serotonin signaling. It is neither a stimulant nor a controlled substance and was FDA-approved for children in 2021 and adults in 2022. Trials report onset within one to two weeks, with smaller average effects than amphetamine or methylphenidate and a different adverse-effect profile. Graded Strong for its approved ADHD context.

Blarcamesine (ANAVEX 2-73) — mechanism, evidence, and regulatory context. This compound is popular in longevity circles because it activates the sigma-1 receptor and is proposed to affect autophagy rather than clearing amyloid like the antibody drugs. Its Phase 2b/3 Alzheimer’s trial showed modest, subgroup-dependent benefit. In December 2025 the European Medicines Agency recommended refusing approval, and Anavex withdrew its European application in March 2026; FDA discussions were still ongoing. The Weak grade places that mechanism alongside the clinical and regulatory evidence.

SLU-PP-332 — “exercise in a pill,” minus the humans. This ERR-agonist compound mimics the effects of endurance exercise in mice — more running capacity, less fat, better insulin sensitivity. It remains a preclinical research direction. As of 2026 it has never been tested in a single human, and improved versions are still being optimized for the clinic, while gray-market vendors already sell it to biohackers. We’ve graded it Preclinical to make the evidence boundary explicit: there is zero human safety or efficacy data, and the entry does not infer an individual use decision from that absence.

The database is now 434 interventions.

The Watchlist had grown into five sections and dozens of entries — more of a second database than the “what could move the consensus” list it’s meant to be. We rebuilt it around two questions readers actually ask.

1. What’s about to read out? Imminent readouts now means a specific trial or regulatory decision in roughly the next six months — not the old 12–18 month window. Six compounds qualify today: olezarsen (an apoC-III blocker whose severe-hypertriglyceridemia decision lands June 30), CagriSema (FDA decision ~October), survodutide (full SYNCHRONIZE-1 data at ADA this month), the Lp(a)-lowering class (pelacarsen’s first-ever cardiovascular outcomes trial reads out this year), obicetrapib (EU decision and its big outcomes trial both due in late 2026), and retatrutide (filing and more Phase 3 data through 2026).

2. What’s genuinely new? On the frontier collects the early, high-discovery-value stuff where the data is still thin — first-in-human trials and novel mechanisms. Today: the first in-body PCSK9 gene editor (VERVE-102, which cut LDL up to 62% from a single infusion in an early trial), the first partial-reprogramming therapy in humans (Life Biosciences’ ER-100), the first TRPML1 “autophagy switch” agonist (LW-1017), an HIV drug being repurposed to silence “jumping genes” for aging (TPN-101), plus two existing early bets — ACD-856 and therapeutic plasma exchange. These carry provisional grades on purpose: they’re ones to watch, not recommendations.

We dropped the Tier-shift watch, Active development, and Recently available sections. Compounds in active development with no near-term catalyst weren’t deleted — they simply live in the main rankings now, where you can find them by their investigational status.

13 new entries. Everything on the old watchlist that already had enough data to grade got promoted into the database: the next-wave obesity drugs amycretin (up to ~14.5% weight loss in Phase 2) and eloralintide (up to ~20%); the lipid programs olezarsen/plozasiran (apoC-III) and evinacumab/zodasiran (ANGPTL3); estetrol, a native estrogen heading for menopause; the topical anti-androgens pyrilutamide and clascoterone for hair loss; and two psychiatric drugs — lumateperone (newly approved as an add-on for depression) and olanzapine + samidorphan (olanzapine with the weight gain blunted). The four frontier programs above were added too, at honest preclinical/early grades.

A few “candidates” turned out to already be in the database under broader rows — the lecanemab + donanemab combination (our anti-amyloid antibody row), topical finasteride (our finasteride row), and mazdutide’s Western trials (our mazdutide row) — so those weren’t duplicated. And three diagnostics that had been parked on the watchlist (the Stelo glucose monitor, advanced lipid markers, and inflammatory aging clocks) were removed: useful tools, but the rankings grade interventions, not tests.

The database is now 427 interventions.

This run focused on the part of the database that goes stale fastest: the investigational drugs with near-term regulatory or trial events. Most rows checked out as accurate. One did not, and it was a meaningful miss.

Elamipretide (SS-31) is no longer investigational for every context. Our row still listed it only as an experimental Stealth BioTherapeutics compound. The FDA granted accelerated approval on September 19, 2025, as FORZINITY for Barth syndrome — the first approved mitochondria-targeted therapeutic. That approval is specific to an ultra-rare genetic disease and does not establish efficacy for mitochondrial-energy or longevity claims, which remain off-label and unproven. We corrected the legality and rewrote the note to display the approved indication and other proposed uses as separate evidence contexts. The headline grade still describes the longevity claim identified in the row.

Survodutide’s first Phase 3 read out. SYNCHRONIZE-1, the obesity trial of Boehringer Ingelheim’s GLP-1/glucagon dual agonist, posted topline results: −16.6% mean body weight at 76 weeks versus −3.2% on placebo, with full data due at the ADA 2026 Scientific Sessions. We added the readout to the note but left the evidence tier where it was — topline press numbers and a conference slot aren’t the same as a peer-reviewed efficacy paper, and we don’t move grades on either alone.

Three regulatory clocks updated. CagriSema (cagrilintide + semaglutide) had its FDA filing sharpened — Novo Nordisk submitted the NDA on December 18, 2025, with a decision expected around October 2026 — and we tagged the row so it now shows up on the watchlist’s imminent-readouts list. Obicetrapib, the CETP inhibitor, picked up its European filing: the EMA accepted marketing applications for it (with Menarini), with an EU decision expected in the second half of 2026, which would make it the first approved drug in its class. And the FGF21 MASH class got a note refresh — Novo Nordisk bought Akero (efruxifermin) for ~$5.2B last October, and the pegozafermin Phase 3 program now reads out in 2027–2028.

Everything else in the investigational batch — the Lp(a)-lowering class (pelacarsen’s outcomes trial still hasn’t read out), orforglipron, mazdutide, MariTide, bimagrumab, tesofensine — was verified accurate and left unchanged. The discovery scan for genuinely new compounds came up empty for the window, which is the normal result.

A second sweep today, run through a wider evidence net, caught an entry that had gone stale.

Plasma exchange / young plasma — upgraded. Our row was years behind: it cited only the 2020 Conboy dilution experiments and the 2019 FDA warning against Ambrosia-style young-plasma transfusions, and graded the whole idea “preclinical.” That undersold where the evidence now sits. A peer-reviewed, single-blinded, placebo-controlled randomized trial (Aging Cell, 2025, 42 participants) found that repeated therapeutic plasma exchange (TPE) measurably lowered biological-age clocks — roughly 1.3 years with TPE alone and 2.6 years when paired with IVIG — with an acceptable safety profile. A large May-2026 preprint then validated the IgG glycome (the “glycan clock”) as a modifiable aging biomarker across 20,405 people and ranked TPE the single largest per-month reducer, ahead of hormone therapy and caloric restriction. We’ve moved the evidence grade from Preclinical to Weak and flagged it as a tier-shift watch.

The honest caveats stay loud: this is one small trial on surrogate clocks, not hard outcomes (nobody has shown TPE makes you live longer or healthier — only that a biomarker moves), the glycan-clock validation is still a preprint, and TPE is invasive, expensive, and clinic-only. “Weak” is an upgrade from “preclinical,” not an endorsement.

A first-in-class autophagy drug entered the clinic. The discovery half of the sweep surfaced Lysoway Therapeutics’ LW-1017, the first TRPML1 agonist ever dosed in humans (Phase 1 began in Melbourne in May 2026). TRPML1 sits on the lysosome and controls the cell’s autophagy-driven waste-clearance system — the same proteostasis machinery that rapamycin, spermidine, and fasting all lean on. It’s being developed for Alzheimer’s and Parkinson’s, but a clean human readout would be early proof that pharmacologically boosting autophagy does something in people. It’s far too early for a ranking — it’s gone onto the watchlist alongside the partial-reprogramming and senolytic programs now in first-in-human trials.

The rest of today’s rotation — 18 long-stable supplements across detox, essentials, and hormones — turned up no new human trials or safety signals, which is the usual and correct result.

A reader-driven coverage audit caught 25 popular interventions that were either missing entirely or only present as part of a larger combination row. Each now has its own entry with mechanism, evidence tier, dosing, and the safety notes that actually matter at the supplement counter.

Essentials (10 new rows)

Filling in the rest of the vitamin and mineral aisle:

• Vitamin E (alpha-tocopherol / mixed tocopherols) — Weak for CV/cancer prevention after HOPE, SELECT, and ATBC; Moderate only for NASH. The Miller 2005 meta-analysis signal at >400 IU/day is the practical reason mixed tocopherols beat isolated alpha.
• Vitamin B1 (thiamine), B2 (riboflavin), B5 (pantothenic acid), B6 (pyridoxine / P5P), B7 (biotin), B12 (cobalamin) — each split out from the combined B-complex row. The two practically important entries are B2 (400 mg/day has real migraine prophylaxis data) and B6 (chronic

  200 mg/day causes peripheral sensory neuropathy — the most common reason to flag a high-dose B-complex). B7 also flags the FDA biotin/lab-assay interference warning that masks troponin and thyroid panels.

• Vitamin K1 (phylloquinone) — added as a distinct row from the existing K2 (MK-7) entry because the two have completely different evidence stories. K1 = blood clotting (Strong); K2 = arterial calcification, bone (Moderate).
• Calcium (citrate / carbonate) — separate from the existing Calcium-D-glucarate row. The Bolland 2010/2011 CV signal at high-dose isolated calcium is why the entry separates dietary calcium, co-nutrient context, and standalone tablet exposure.
• Manganese — the entry separates nutritional adequacy from the neurologic effects reported with chronic high exposure; many multivitamins already contain manganese.

Metabolic Health (3 new rows)

• Cinnamon (Cinnamomum cassia / verum) — Moderate evidence for ~0.1-0.3% HbA1c reduction in T2DM. The entry distinguishes Cassia and Ceylon because Cassia generally contains more coumarin, with liver toxicity reported at sustained high exposure.
• Allulose — small studies report attenuation of postprandial glucose when it is co-ingested with carbohydrate. Long-term outcomes remain less certain, and retail cost is often substantially higher than sucrose.
• Monk fruit (mogroside V) — a non-nutritive sweetener with limited direct comparative human-outcome evidence against aspartame, sucralose, or stevia; taste and commercial formulations vary.

Longevity (3 new rows)

• Luteolin — flavonoid often named in senolytic stacks alongside fisetin and quercetin, but with much thinner human data. The available human evidence does not establish standalone effects.
• Cocoa flavanols (Theobroma cacao) — Strong on endothelial function; COSMOS (n=21,442) missed its primary CV composite endpoint but hit ~27% CV death reduction as a secondary. Trials generally used standardized flavanol preparations; supermarket chocolate has variable flavanol content.
• Lutein + Zeaxanthin — Strong for AMD progression (AREDS2 2013). Notably replaced beta-carotene in AREDS2 because the carotenoid caused lung cancer in smokers in ATBC. Comparative studies report higher lutein bioavailability from an egg-yolk matrix than from spinach.

Cognitive (1 new row)

• Matcha (Camellia sinensis) — contains both L-theanine and caffeine. Contaminant and catechin concentrations vary by origin, grade, and product testing; “ceremonial” is not a clinical quality category. Rare liver-toxicity cases are reported mainly with concentrated catechin extracts, and exposure differs from matcha powder used as a beverage.

Mood, Anxiety & Stress (1 new row)

• GABA (gamma-aminobutyric acid) — distinct from phenibut and gabapentin, which were already in the database. The BBB-penetration question is the central debate; effect size is small either way. The row records a relatively favourable Safety score alongside small, uncertain effects and does not equate supplement evidence with evidence for clinical anxiolytics.

Sleep & Recovery (1 new row)

• Lysine (L-lysine) — old but solid HSV recurrence prevention data; competes with arginine for viral protein synthesis.

Immune & Inflammation (6 new rows)

• Lactobacillus acidophilus — single-strain version of the existing multi-strain probiotic row. Modern probiotic data is strain-specific — LGG (rhamnosus) and S. boulardii outperform acidophilus for AAD, but the vaginal/urogenital indication is the cleanest.
• Reishi (Ganoderma lucidum), Maitake (Grifola frondosa), Shiitake (Lentinula edodes) — the three classical immune-modulating mushrooms previously only mentioned inside the Beta-glucans row. Lentinan (from shiitake) is the standout: actual Phase III oncology data as an IV gastric-cancer adjuvant in Japan. Reishi’s hepatotoxicity case reports with concentrated extracts are the most important safety note.
• Ginger (Zingiber officinale) — one of the few herbs with consistent positive RCT data. Strong for nausea (pregnancy, postoperative, chemotherapy-induced) and in major obstetric guidelines.
• Lycopene — cooked tomato matrix beats supplement on bioavailability. The prostate-cancer story is genuinely mixed in the literature.

Also: Iodine row updated

The Iodine row now documents potassium iodide (KI) and potassium iodate (KIO3) 65-130 mg tablets as the pharmacologic thyroid-blockade form for radioactive iodine exposure (Chernobyl/Fukushima protocol) — a different use from daily nutritional iodine.

Items already in the database

22 items from the original 47-compound list were already covered by existing rows and were not edited this run: NR and NMN (inside the NAD+/NMN/NR row), ubiquinol (inside CoQ10), sodium hyaluronate (inside Hyaluronic acid), lithium orotate (inside Lithium), glucoraphanin (covered by the Sulforaphane row, which already explains the myrosinase requirement), Lion’s mane, Spermidine, Taurine, Glycine, Glucosamine, Glutathione, Astaxanthin, MCT oil, Curcumin, Fisetin, Vitamin D, Vitamin K2, Niacin (= B3), Zinc, Selenium, Boron, and Alpha-ketoglutarate (= CAAKG).

The database now tracks 414 interventions.

The discovery pass surfaced six useful interventions that were not yet present as database rows. They now have sourced entries, evidence/risk summaries, dose notes, and secondary category visibility where appropriate.

Added:

• CBT-I as a first-line behavioral insomnia protocol, visible under Sleep & Recovery, Mood, and Cognitive.
• Beta-alanine and sodium bicarbonate as evidence-backed ergogenic buffers for high-intensity exercise.
• MCT oil / ketogenic MCT drink for the mild-cognitive-impairment/ketone fuel use case.
• 2’-fucosyllactose (2’-FL) as a targeted HMO prebiotic with early older-adult microbiome/metabolic data.
• Dietary nucleotide supplementation as a promising but early aging/cognition/muscle nutrition entry.

The database now tracks 389 interventions.

A reader caught the pattern: Pinealon was listed only under Longevity even though its own evidence notes include cognitive/neuroprotective claims, and it is commonly searched as a pineal/circadian peptide.

The database already has a mechanism for this. Each intervention keeps one canonical CSV category, while CROSS_CATEGORIES decides where else it appears in the rankings filters. The issue was not duplicate rows or scoring. It was under-listing.

This pass reviewed all 383 rows and expanded secondary category membership where the row already had a credible use case, mechanism, or reader-search reason to appear elsewhere. Examples:

• Pinealon now appears under Longevity, Cognitive, and Sleep & Recovery. The sleep listing is deliberately evidence-weak: it reflects pineal/circadian positioning, not strong human sleep-trial evidence.
• Creatine now appears under Cognitive and Muscle & Strength, not only Essentials/Sleep & Recovery.
• Omega-3, NAC, taurine, GlyNAC, sulforaphane, curcumin, boswellia, SPMs, PEA, thymosin alpha-1, and KPV now show up in the inflammation, detox, mitochondrial, mood, or metabolic lists they logically touch.
• Progesterone, testosterone, HGH and GH secretagogues, resmetirom, and thyroid-support minerals now appear in the endocrine or metabolic lists readers would expect.
• Urolithin A, methylene blue, alpha-lipoic acid, CoQ10, MOTS-c, Humanin, and ALCAR now surface across the mitochondrial, cognitive, metabolic, muscle, or longevity filters where appropriate.

Scores, evidence tiers, doses, notes, and primary CSV categories did not change. This is a visibility correction: the same intervention now appears in every category it has earned, without creating duplicate compound pages.

The retatrutide row was tagged “awaiting readout” because TRIUMPH-1, Eli Lilly’s first registrational Phase 3 obesity trial, was due to report this year. The topline announcement landed today, and it landed firmly: a 28.3% mean weight loss at 80 weeks on the 12 mg arm — about a third of body weight — and 30.3% (an average of 85 pounds) at 104 weeks for participants who continued. For comparison, tirzepatide’s SURMOUNT-1 reached ~22.5% at 72 weeks and semaglutide’s STEP-1 reached ~14.9% at 68 weeks. Before retatrutide, ~30% weight loss was the territory of bariatric surgery.

The trial randomized 2,339 adults with obesity (or overweight plus at least one weight-related comorbidity) and no diabetes to retatrutide 4 mg, 9 mg, or 12 mg weekly, or placebo. Almost half (45.3%) of the 12 mg arm reached ≥30% weight loss, vs 0.5% on placebo. Nearly two thirds (65.3%) reached a BMI below 30 — i.e., crossed out of the obesity category entirely.

Safety follows the GLP-1 class playbook. At 12 mg, nausea hit 42.4% (vs 14.8% placebo), vomiting 25.3% (vs 4.8%), and diarrhea 32.0% (vs 13.5%). Discontinuation for adverse events ran 11.3% vs 4.9% on placebo. The novel signal is dysesthesia at 12.5% vs 0.9% on placebo — a first-of-class read on this side effect that warrants monitoring as the program expands.

Important caveats. This is a sponsor topline announcement, not yet a peer-reviewed paper — we’ve labeled the row accordingly. It’s also one Phase 3 trial in one indication; TRIUMPH-2 (type 2 diabetes), TRIUMPH-3 (established cardiovascular disease), TRIUMPH-4 (knee osteoarthritis), the TRIUMPH-Outcomes CVOT, and the head-to-head against tirzepatide are all still pending through 2026. No FDA filing has been announced. We’re keeping the “awaiting-readout” tag on the row for those, and adding “tier-shift watch” to surface it in the public watchlist’s tier-shift section.

This run also touched two existing rows on the quieter side: omega-3 picked up two confirmatory citations on the EPA-vs-EPA+DHA and dietary-vs-supplemental atrial-fibrillation distinction the row already takes, and alpha-ketoglutarate gained a 2026 Aging Cell cohort study from the Kennedy lab pegging delayed-release Ca-AKG at ~1.8 years lower epigenetic Age Residual after adjustment — a more conservative real-world number than the 2022 single-arm ~8-year claim already on the row. The ABLE RCT remains the data point worth waiting for there.

We audited the site against the WADA 2026 prohibited-list CSV and added coverage for the missing performance-relevant gaps: long-tail anabolic steroids, EPO/HIF oxygen-delivery drugs, myostatin and activin blockers, GH analogs and secretagogues, exercise mimetics, stimulant families, diuretics and masking agents, opioids, THC/cannabinoids, systemic glucocorticoids, beta-blockers, high-volume IV infusions, and gene/cell doping concepts.

The goal was coverage without turning the database into a phone book of obscure near-duplicates. Common compounds still get their own pages; long-tail WADA examples now live inside class-level rows with aliases in the notes.

When an intervention has evidence that varies by use, its database row records that as a multi-part rating — for example Strong (acute poisoning); Very Weak (detox supplement). The score reads the first part. The convention is to lead with the use a reader of this site would actually adopt.

Thirteen rows broke that convention. They led with a narrow emergency-room or clinical indication — and so they scored as if that were the point of taking them.

The clearest case: activated charcoal sat at A+. It earned that on “Strong evidence, high effectiveness” — for treating acute poisoning in an ER. As the “detox” supplement it’s actually sold as, the evidence is Very Weak (it never leaves the gut). The row led with the ER use, so the score followed.

The whole Detox category ran on this pattern — chelators and adsorbents genuinely are strong emergency drugs and inert supplements at the same time.

Each row’s rating clauses were reordered so the reader-relevant use leads. What moved:

• Activated charcoal — A+ → C
• Milk thistle — A+ → C+
• Glutathione — A → B
• NAC (both entries) — A → B+
• Thymosin α-1 — A → C+
• Digestive enzymes — A− → C+
• CBD — B → C+
• HGH — B− → C−
• DMSA / DMPS chelators — B− → D
• IGF-1 LR3 — C → D
• Anti-amyloid antibodies — C− → D−
• EDTA chelation — C+ → F

A note on the last two. Anti-amyloid antibodies were scoring on amyloid clearance shown on PET scans — a surrogate marker; the rating now reflects clinical benefit, which a 2026 Cochrane review found trivial-to-small against a meaningful risk of brain swelling and bleeding. EDTA chelation was scoring on lead-poisoning treatment; marketed as a heart-disease or “detox” therapy it has no demonstrated benefit (the TACT2 trial was negative), and it lands at F.

Nothing about the scoring formula or the grade bands changed — this was a data correction. The authoring convention that prevents it recurring is now written into the data workflow.

A reader pointed out that creatine showed up twice in the rankings — once under Essentials, once under Sleep & Recovery. A full scan of the database turned up 14 compounds with the same problem: a compound that legitimately belongs to two categories had been entered as two separate rows, which meant two ranks, two scores, and two compound pages for one substance.

That is now fixed. Each of these 14 compounds is a single entry:

• Creatine, vitamin D, magnesium, zinc, omega-3, collagen peptides — kept in Essentials, also shown under Sleep & Recovery or Immune & Inflammation.
• Ashwagandha — kept in Mood, Anxiety & Stress, also shown under Essentials.
• Berberine, tesofensine, T3, CoQ10 — consolidated into their primary category, cross-listed under the second.
• Quercetin and lithium — single entries, cross-listed across the Longevity and Immune/Mood lists they each touch.
• GHK-Cu — had two near-identical rows inside Longevity itself; now one.

The surviving entry keeps the fuller of the two write-ups, and any citations that only existed on the other row were carried over, so nothing was lost. When a merge changed the picture — for example, T3’s off-label fat-loss use is now folded into the single thyroid-hormone entry — the notes were updated to say so.

A consolidated compound still appears under every category it earns a place in: filter the rankings by Essentials or by Sleep & Recovery and creatine shows up in both. It is just one row now, ranked once, instead of two competing copies.

One deliberate non-change: glycine is still two entries. The longevity dose (10–15 g/day, as a methionine-restriction mimetic) and the sleep dose (3 g before bed) are different enough in purpose to stand on their own — that is not a duplicate.

NAC (N-acetylcysteine) had two separate database rows — one filed under Essentials, one under Immune & Inflammation — a cross-category duplicate the earlier consolidation pass missed.

They are now a single row. The canonical entry lives in Essentials (where the other daily-supplement staples sit) and is cross-listed under Immune & Inflammation, so it still appears in both category rankings. The merged row keeps the fuller mechanism, risk, legality and notes detail from the two originals and the combined citation list.

The grade is unchanged — NAC was B+ in both rows after the evidence-context correction, and it remains B+.

S-tier had no members. That was a bug, not a verdict.

The composite score runs 0–10, and the letter scale mapped S to 9.4 and up. But the database’s interventions clump near the top — more than a dozen sit at exactly 8.4 — and the single best score anywhere was 8.8. The top grade was, in practice, impossible to earn. A reader’s first question about a tier list is “what’s at the top?”, and the honest answer was “nothing,” which isn’t an answer.

We fixed it in two parts.

The top of the scale was recalibrated. The four highest bands (A− through A+) were tightened to 0.3-point steps, and S now begins at 8.5 — the realistic ceiling of the database rather than a number nothing reaches. Bands B+ and below are untouched, so 245 of 350 interventions keep exactly the grade they had. The composite formula itself — how Evidence, Benefit, and Safety combine — was not changed.

Five interventions were re-graded on the evidence. Sleep, resistance training, protein intake, creatine, and VO2 max / HIIT were carrying “Strong” evidence ratings when their evidence base is genuinely among the most robust of anything in the database — replicated across hundreds of trials and decades of epidemiology. They were moved to “Very Strong.” Sleep’s risk rating was also corrected: a note about the danger of sleep deprivation had been sitting in the risk field, where it doesn’t belong — the act of sleeping 7–9 hours is very low risk.

The new S-tier, eight interventions:

• Sleep — 9.7
• Creatine — 9.7
• Resistance training — 9.3
• Protein intake — 9.3
• Statins (rosuvastatin / atorvastatin) — 8.8
• Tadalafil — 8.8
• PCSK9 inhibitors — 8.7
• VO2 max / HIIT — 8.5

Statins, tadalafil, and PCSK9 inhibitors were not re-rated at all — they already carried top evidence and effectiveness scores and simply landed in S once the band was reachable.

This was a calibration of the letter scale and a correction of five under-rated rows — nothing about the underlying scoring math changed.

The full-table review produced one big access change: orforglipron is no longer a pipeline-only obesity drug. FDA approved it as Foundayo on April 1, 2026, so the row now uses the approved tablet titration schedule and Rx status.

Semaglutide also changed. Wegovy HD adds a 7.2 mg weekly option, with FDA specifically calling out altered skin sensation as more common at the higher dose.

Discovery added FGF21 analogs as a Metabolic Health row. Efruxifermin, pegozafermin, and efimosfermin are still investigational, but the class now has enough human Phase 2 data and Phase 3 MASH programs to belong in the database rather than the watchlist.

The toxins table gained four consumer-channel hazards: DMAA/DMHA stimulant adulterants, yellow oleander-adulterated weight-loss supplements, nitrite poppers, and recreational nitrous oxide inhalation.

This audit fixed a failure mode in the rankings: a large effect size had too much room to pull dangerous interventions upward. That made the score less faithful to an evidence-first, safety-visible product.

The formula still uses Evidence, Benefit, and Safety, but the weights now put more pressure on Evidence and Safety: 0.45·Ev + 0.15·Bn + 0.40·Sf. There are no hidden caps or vetoes. A compound can show high Benefit and terrible Safety at the same time, and the visible breakdown should make that tradeoff obvious.

The row corrections were deliberately focused:

• Caffeine now reflects ordinary dietary use as low risk, with a separate warning for pure or highly concentrated caffeine powders.
• Clomiphene and enclomiphene are now separate entries.
• DNP and several high-risk performance drugs gained clearer safety context in their notes and visible score breakdowns.
• BPC-157 and TB-500 now use parser-recognized Unknown-High risk language.
• Fadogia now correctly shows no human efficacy trials.
• PFAS now parses as high-prevalence toxin exposure.
• Duplicate Astaxanthin and Yohimbine rows were consolidated.

This is a correction, not a recommendation. The point is to make the ranking table less gameable by raw efficacy while keeping the component scores honest and visible.

“Protocols” was doing two jobs at once: a category (its own tab in the rankings) and a kind of thing (a class, like Peptide or Hormone). The double duty meant sauna sat in a generic “Protocols” bucket instead of next to the cardiovascular and longevity interventions it actually competes with.

It’s now just a class. Every protocol — sauna, cold plunge, Zone 2, resistance training, time-restricted eating, breathwork — keeps the Protocol class label but is ranked inside its real evidence categories. Resistance training shows up under Muscle Building, Longevity, and CV/Lipid; sauna under CV/Lipid, Longevity, and Recovery. The standalone Protocols tab is gone.

Nothing about the scores changed — only where these rows are filed.

Read Off Label ranks things you take or do — compounds, supplements, and protocols — scored on evidence, benefit, and safety. Diagnostic tests (bloodwork, DEXA, CGM, whole-body MRI, multi-cancer screening, epigenetic age tests) are a different kind of decision: their value depends on what you do with the result, not on a benefit-versus-risk profile. Scoring them on the same rubric was always a stretch, so the category is out.

What changed: 16 diagnostic entries removed. The rankings page drops its Diagnostics mode and is now a two-way Interventions / Toxins view. The database goes from 19 categories to 18, and from 384 to 368 ranked entries.

This narrows the site to what it does best. A dedicated treatment of testing — which tests are worth it, and how to act on the results — may return later as its own format, but not inside the intervention rankings.

First full evidence audit since launch. Worked category by category against PubMed, ClinicalTrials.gov, AdisInsight, and bioRxiv. Net result: the database grew from 357 to 378 interventions across 19 categories, and 31 malformed CSV rows (silent unescaped commas) were repaired.

Score-relevant rerates:

• Rapamycin — downgraded. The PEARL trial (April 2025) missed its primary endpoint of visceral adiposity reduction. Secondary signals on lean mass and well-being held up; the headline longevity claim did not.
• Multi-cancer early detection (Galleri) — downgraded from Moderate to Weak–Moderate. NHS-Galleri 2026 — the first large RCT of MCED — did not reduce late-stage cancer incidence.
• Cagrilintide (CagriSema) — re-framed. REDEFINE 1/2 (NEJM, June 2025) came in below Phase 2 expectations (-20.4% vs ~25% projected) and below tirzepatide 15 mg’s SURMOUNT-1 number. Removed the prior “rivals tirzepatide” framing.
• Tirzepatide — refined with SURMOUNT-5 head-to-head vs semaglutide (-20.2% vs -13.7%) and SURMOUNT-1 176-week extension data.
• Testosterone (TRT) — FDA Feb 2025 label change folded in. Boxed cardiovascular warning removed; class-wide blood pressure warning added; AF / AKI / PE signals from TRAVERSE acknowledged.
• Niacin — Lp(a) niche eclipsed by RNA-targeted agents (olpasiran, pelacarsen, lepodisiran et al). Notes refreshed.
• Retatrutide — TRIUMPH Phase 3 program detail (4 trials + TRANSCEND-CKD + TRIUMPH-Outcomes CVOT) updated.

New entries — Weight loss: orforglipron (Lilly oral small-molecule GLP-1), survodutide (BI/Zealand GLP-1/glucagon), maridebart cafraglutide / MariTide (Amgen monthly peptide-antibody), mazdutide (Innovent dual agonist, approved China 2025).

New entries — CV/Lipid: Lp(a)-lowering therapies (olpasiran / pelacarsen / lepodisiran / zerlasiran / muvalaplin) as a class entry; obicetrapib (next-gen CETP inhibitor).

New entries — Hormones: fezolinetant (Veozah, NK3 antagonist for VMS); elinzanetant (Lynkuet, dual NK1/NK3 antagonist); resmetirom (Rezdiffra, first FDA-approved MASH/NASH drug); romosozumab (Evenity, sclerostin antibody for osteoporosis).

New entries — Mental Health: dextromethorphan-bupropion (Auvelity, oral rapid-acting antidepressant); zuranolone (Zurzuvae, oral PPD); xanomeline-trospium (Cobenfy / KarXT, first new schizophrenia mechanism in 50+ years).

New entries — Cognitive: anti-amyloid mAbs (lecanemab / donanemab / aducanumab) as a class entry, with the 2026 Cochrane meta-analysis showing only trivial cognitive benefit; solriamfetol (Sunosi); pitolisant (Wakix); phenibut (with explicit dependence/withdrawal warning).

New entries — Recovery: suzetrigine (Journavx, first new acute pain mechanism in decades); glucosamine + chondroitin.

New entries — Longevity: pyrroloquinoline quinone (PQQ).

New entries — Immune: bovine lactoferrin.

Minor refreshes: bimagrumab, enclomiphene/clomiphene, estradiol HRT (NK3/NK1 cross-ref), TMS (SAINT clearance), microplastics (2026 Polimeni cardiovascular review).

Full per-row entries with DOIs and PMIDs live in the repo at src/data/audit_log.md. Eight cross-category duplicates (tesofensine, ashwagandha, collagen peptides, quercetin, berberine, astaxanthin, CoQ10, microdose lithium) are flagged for consolidation in a future pass.

First published ranking. 357 compounds and protocols across 19 categories, scored against the composite rubric. Methodology and disclaimer finalised. Weekly newsletter, Off Label Weekly, launching 28 April.

Initial scores are the author’s best synthesis as of this date. Expect movement as readers flag studies I missed. Every subsequent change will land here.