Verified against primary sources
Seven pages, 105 compounds, every number pulled from PubMed and ClinicalTrials.gov with the query declared.
What sets these pages apart from the rest of the site
Most of this site is an organized translation of a commercial secondary source. It is declared on every page, and it is the structural weakness of the material: if the source got it wrong, I repeat the error.
The pages listed below are different. Every figure in them was gathered by me directly on PubMed and on ClinicalTrials.gov, with the query used declared inside the page itself, so that anyone can repeat it and contradict me. They cover 179 distinct compounds, plus 106 substances and treatments surveyed in the round of September 12, 2026, across the nine pages opened that day.
This is not a featured section out of methodological vanity. It is because, of the two things a site like this can do — list a dose or say what the dose is worth —, only the second is hard to find in Portuguese.
The pages in this section
| Page | Compounds | The central finding |
|---|---|---|
| The GLP-1s against the package insert | 6 | The only page on this site checked against an official answer key. The titration ladders match the FDA, step by step. The black box warning was missing, the 7.2 mg dose that exists only in Brazil was missing. |
| Mazdutide | 1 | The compound with the most mature evidence on the whole site, and the opposite of the pattern here: 12 phase 3 trials, with GLORY-1 in NEJM, GLORY-2 in JAMA and the two DREAMS in Nature — and no FDA registration. All 12 have Chinese sponsors and populations, and the meta-analysis pooling 2,292 participants rates the certainty in obesity as very low in the same text in which it reports double-digit weight loss |
| Exosomes | 1 | The site's most extreme asymmetry: 45,674 articles on PubMed and 83 randomized trials, less than 0.2%. Of the 378 records on ClinicalTrials.gov, 142 are observational — they measure exosome as a marker, they don't treat anyone with it. I opened the nine phase 3 ones: one is a PET/CT study in prostate cancer that entered the search for citing “exosome analysis”, and seven of the other eight have 135 participants or fewer. The FDA has said since 2019 that no exosome product is approved, with a new warning letter in February 2026 |
| Monoclonal antibodies | 3 | The three that show up alongside the GLP-1s when the subject is lean mass, and the evidence scale does not keep up with the conversation: trevogrumab has 2 articles on PubMed, none of them a trial, and a single registered study. The only one with an FDA-approved label is garetosmab, and not for weight — it is for a rare disease, by 60-minute intravenous infusion every four weeks. The molecule weighs about 146 kDa against semaglutide's 4,113.58 g/mol, is produced in cell culture and not synthesized: there is no vial to reconstitute here |
| Next-generation incretins and amylins | 13 | The inversion between literature and development. HRS-9531 has 0 articles on PubMed and 9 registered phase 3 trials; liraglutide, which serves as the yardstick, has 5,909. Anyone who judges these molecules by article count concludes that half of them don't exist. 9 of the 11 with a published INN have no active registration |
| The GH axis outside the peptide vial | 7 | GHRP-2 and hexarelin have zero registered studies on ClinicalTrials.gov, with 256 and 313 articles on PubMed. The molecule has been studied; the treatment hasn't. On the other side, somatropin has 183 phase 3 trials and 9 active registrations |
| Geroprotectors | 17 | 6 of the 17 lines have no phase 3 trial registered, among them recombinant klotho and plasma exchange. Rapamycin has 259 — and none of the ones I opened is about aging. The 2 phase 3 trials for fisetin declare zero participants enrolled |
| Liver and lipids | 13 | The opposite of the site's usual pattern: literature spanning dozens of articles and outcome trials of up to 17,300 participants — the four largest add up to 46,047. And, of the thirteen, only inclisiran has active registration |
| Baldness — the old and what's coming | 18 | Clascoterone is registered in Brazil in the anti-acne class, not for hair. PP405, which already circulates as a research powder, has 1 article on PubMed and 0 phase 3 trials; VDPHL01 has the opposite — 0 articles and 3 phase 3 trials with more than a thousand people enrolled |
| Repair beyond BPC and TB-500 | 5 | Phase 3 of thymosin β-4 is eye drops for dry eye — 700 and 601 participants, not tendon repair. AHK-Cu has 1 article on PubMed. The only phase 3 record that the search for polydeoxyribonucleotide returns is defibrotide in hepatic veno-occlusive disease |
| Neuro and cognition — the Russian bloc and what has a label | 13 | Noopept, phenylpiracetam, bromantane, and cortexin together have 0 registered studies on ClinicalTrials.gov, despite totaling 435 articles on PubMed. In the same survey, lecanemab and donanemab already have active registration |
| Muscle, bone, HPG axis and immune system | 15 | The trial tested a rare disease; the vial promises recomposition. The 6 phase 3 trials of afamelanotide that I opened are for erythropoietic protoporphyria, and the ones for setmelanotide are for genetic-variant obesity. 8 of the 14 searches come up empty |
| Treatments that aren't a substance | 5 | The comparison that almost never comes up: bariatric surgery has 36,916 articles and 2,039 registered studies. Hyperbaric oxygen therapy crossed with aging, on the other hand, has 1 phase 3 trial, with 30 participants |
| SARMs | 16 | The field's only two phase 3 trials were never published — 651 patients that exist only as a table on ClinicalTrials.gov, one of them posted seven and a half years late. And before the molecule comes the merchandise: only 52% of products contained any SARM. Andarine, YK-11 and S-23: zero trials |
| Semax — the evidence | 1 | The Russian trials used 6,000 to 18,000 mcg/day; the community uses 250 to 1,000. The range in practice has never been tested. Zero trial registrations, and one of the four studies is negative |
| KLOW — the evidence | 4 | The blend still has no trial, but BPC-157, TB-500 and GHK-Cu entered clinical phase in 2026 — three Hudson Biotech studies recruiting. It is no longer only preclinical |
| Nootropics | 44 | 32 have a trial, 12 have none. Modafinil leads with 425; noopept has 115 articles and zero trials. Four counts were inflated and were redone |
| Effect size | 19 | The other half of the page above: what the trials found, not how many exist. Where there is an effect, it sits between SMD 0.2 and 0.4. Three popular compounds have a large negative trial — oxiracetam (n=500), Cerebrolysin (n=1779) and piracetam |
| Khavinson bioregulators | 11 | Zero registered trials on ClinicalTrials.gov. Eight of the eleven have no clinical trial article at all. Seven do not even have an indexed chemical sequence. All of the family's clinical literature is concentrated in two compounds |
| Thymalin | 1 | 293 articles and 13 trials — the outlier of the family. But zero trial registrations, and the study behind the entire longevity claim is signed by the inventor himself |
| Meldonium | 1 | 357 articles, 35 trials, 7 trial registrations. The finding that decides its use is not about efficacy: the urinary detection window reaches 117 days after six days of use |
| Eastern European drugs | 17 | 29 registered trials, several of them phase 3 — the opposite of the bioregulators. And a forgotten signal: calcium hopantenate has 47 cases and 11 deaths from encephalopathy reported since 1986 — with 7 randomized trials, several in children |
| Over-the-counter supplements | 43 | Trial count item by item, with dose, for all 43. Resveratrol has 391 trials and disappointed; glycine, alpha-GPC, butyrate, fisetin and magnesium L-threonate have between 2 and 5 each |
| Herbal medicines and nootropics | 18 | Fadogia agrestis: zero. No trial, no human study — I checked the three articles PubMed tags as human and none of them is. Lion's Mane has two |
| Prescription-only items | 12 | Prescription drugs that appeared mixed in with vitamin D and creatine on a supplement list. The only page on the site without dosing, by decision |
The column doesn't add up to 179. Two pages are a second cut of the same compounds, not new compounds: the effect-size one reanalyzes 19 that already appear in the others, and the registration one covers the site's 44 protocol compounds, which aren't counted as primary-source material.
Three things this survey taught
1. A lot of literature is not the same as evidence
The eleven Khavinson bioregulators add up to nearly two hundred articles on PubMed and zero registered trials. Prior registration is what keeps a bad outcome from turning into a different outcome at publication — without it, there is no way to know how many studies stayed in the drawer. Article count measures academic activity, not reliability.
2. High volume is not a favorable result
Resveratrol has 391 randomized trials and meta-analyses, and the promise did not hold up. Ginkgo has 88 and is still not recommended for preventing dementia, because the large trials came back negative. Probiotics have 5,248 — the highest number on the site — and are the most misleading, because the effect is strain-specific and does not transfer to the bottle you bought.
3. Searching by brand name inflates the count
While surveying the bioregulators, a randomized trial turned up for Ovagen. I went to read it: it was a superovulation study in cows, because Ovagen is also a brand of veterinary FSH. In the Eastern Europe block, cocarboxylase and cytochrome C show up with hundreds of articles about the biochemistry of metabolism and apoptosis — not about the injectable. When the molecule has a biological function of its own, the count measures the biology, not the drug.
On the nootropics page the same problem showed up in another form: PubMed expands the search for armodafinil to the MeSH term modafinil, and returns 2,437 versus 2,411 — almost the same set. Restricted to title and abstract, armodafinil has 260. And four supplement counts were inflated for the same reason: L-tyrosine dropped from 199 to 14, uridine from 62 to 6, forskolin from 30 to 5, and agmatine from 9 to zero.
What is still missing
- I read abstracts and metadata, not full articles, except where the finding depended on it — Fadogia and calcium hopantenate, which I opened and checked one by one.
- Count is not quality. A high number can be a hundred small, poorly done trials. It is the central limitation of the method used here.
- There are no more items without a count. The 27 that were blank on the supplements, herbal medicines and Eastern Europe pages have been closed, and the two on the KLOW page as well. Closing them corrected a wrong number of mine — Cortexin had 26 articles declared and has 127 — and showed that cocarboxylase, cytochrome C and Ethoxidol have no randomized trial at all.
- Russian literature outside PubMed was not consulted. The Khavinson school publishes heavily in non-indexed journals — absence here does not prove absolute absence.
- The other 57 pages of the site still depend on the secondary source. This method has not yet been applied to them.