Registry Trace
Every letter on this site rests on published papers, and every paper is listed with its PubMed identifier so anyone can open it. This page asks a harder question about the same citations: how much of that evidence is independently traceable, without taking anyone's word for what the study was.
Two facts answer it, and neither of them is ours. The first is the study design that the National Library of Medicine's own indexers assigned to a paper, which is not what the authors called it and not what we filed it as. The second is whether the citation carries a public trial registration, meaning the study was declared before its results were known.
Both facts are published. Neither is usually shown. We are showing ours because the numbers are not flattering and a reader is entitled to them.
Last reviewed: August 2026
A study declared before it runs can be checked against what it later reports. A study first seen at publication cannot.
Since 2005, the major medical journals have required a trial to be publicly registered before enrolment as a condition of publication. Registration records what the trial intended to measure, in whom, and for how long. It is the only mechanism that makes two otherwise invisible things visible: a trial that quietly changed which outcome it called primary once the data arrived, and a trial that finished and was never published at all.
Neither of those is rare. The reason our homocysteine An amino acid in the blood that rises when vitamin B12, folate or B6 are low. entry says what it says is that somebody assembled nineteen unpublished datasets and found the answer differed from the entire published literature on the same genotype. That is not an anomaly in the evidence base. It is a property of it.
What sits under our letters.
Counted from our own database at build time, so these figures cannot drift from the citations they describe. The second row of the table below is the one worth reading slowly: an entity where nothing is indexed as a trial and nothing is registered is an entity whose evidence you have to take on the paper's own description of itself.
| Entity | Grade | Citations | Indexed as trial | Registered |
|---|---|---|---|---|
| Ashwagandha standardized root extract 300 mg twice daily | B | 3 | 1 | 0 |
| Hyaluronic Acid (Oral) | B | 9 | 5 | 1 |
| Mediterranean diet + extra-virgin olive oil | A | 5 | 3 | 1 |
| Stool-based colorectal screening (gFOBT and FIT) | A | 5 | 3 | 2 |
| Structured exercise program after treatment for non-metastatic cancer | B | 3 | 2 | 1 |
| REHIT / Sprint Interval Training | B | 7 | 5 | 1 |
| Fasting-mimicking diet (FMD, 5-day cycles) | B | 7 | 5 | 3 |
| Blood Flow Restriction Training | B | 8 | 6 | 0 |
| Psyllium husk (10 to 15 g/day) for LDL-C Low-density lipoprotein cholesterol, the cholesterol fraction most closely tied to plaque in the arteries. | B | 4 | 3 | 1 |
| Ashwagandha (standardized root extract) | B | 10 | 8 | 0 |
| Photobiomodulation (musculoskeletal) | B | 5 | 4 | 0 |
| Nattokinase | B | 5 | 4 | 1 |
| Concurrent training sequence: resistance before aerobic (same session) | B | 6 | 5 | 0 |
| Adequate dietary protein (1.2 to 1.6 g/kg/day) | B | 6 | 5 | 1 |
| Pneumatic Compression (sequential) | A | 7 | 6 | 0 |
| Transcranial Magnetic Stimulation (TMS) for Depression | A | 15 | 13 | 4 |
| EGCG (Epigallocatechin Gallate / Green Tea Extract) | B | 8 | 7 | 1 |
| Creatine Monohydrate | A | 13 | 12 | 3 |
| Photobiomodulation to Prevent Oral Mucositis | A | 14 | 13 | 2 |
| Vitamin D3 | B | 17 | 16 | 4 |
| Alcohol reduction | B | 7 | 7 | 0 |
| Amla (Indian Gooseberry, Emblica officinalis) | B | 2 | 2 | 0 |
| Beta-Alanine | B | 2 | 2 | 0 |
| Boswellia serrata | B | 2 | 2 | 0 |
| Cinnamon (Cassia / Ceylon) | B | 2 | 2 | 0 |
| Coenzyme Q10 (CoQ10) | B | 5 | 5 | 0 |
| Collagen Peptides | B | 8 | 8 | 0 |
| Extracorporeal Shockwave Therapy (ESWT) | B | 5 | 5 | 0 |
| Fenugreek (Trigonella foenum-graecum) | B | 4 | 4 | 0 |
| Ginkgo Biloba (EGb 761) | B | 7 | 7 | 0 |
| HRV Biofeedback | B | 4 | 4 | 0 |
| Home blood pressure monitoring | B | 2 | 2 | 0 |
| Inositol (Myo-inositol) | B | 2 | 2 | 0 |
| Melatonin | B | 13 | 13 | 0 |
| Mobility and fall-prevention protocols | B | 2 | 2 | 0 |
| Oat beta-glucan (3 g/day or more) for LDL-C | B | 5 | 5 | 0 |
| Plant sterols or stanols 1.6 to 2 g/day | B | 1 | 1 | 0 |
| Pneumatic Lymphatic Compression | B | 1 | 1 | 0 |
| RIMSO-50 (Dimethyl Sulfoxide Bladder Instillation) | B | 5 | 5 | 0 |
| Resistance training 2–3×/week | A | 8 | 8 | 0 |
| Saffron (Crocus sativus) | B | 2 | 2 | 0 |
| Sleep restriction treatment (CBT-I) | B | 2 | 2 | 0 |
| Sodium reduction to under 2,300 mg/day | B | 4 | 4 | 0 |
| Transcutaneous Vagus Nerve Stimulation | A | 9 | 9 | 0 |
| Bright Light Therapy for Depression | B | 14 | 14 | 1 |
| VO₂max The most oxygen the body can use during all-out exercise, a standard measure of aerobic fitness. training (Zone 2 + intervals) | A | 7 | 7 | 1 |
| DASH dietary pattern | B | 6 | 6 | 1 |
| Whole-food plant-based diet | B | 10 | 10 | 2 |
| Berberine HCl | B | 9 | 9 | 2 |
| Aerobic exercise (Zone 2) 150 min/week | A | 4 | 4 | 1 |
| Custom Foot Orthotics | B | 4 | 4 | 1 |
| Ketogenic diet (very-low-carbohydrate) | B | 8 | 8 | 2 |
| MBSR (Mindfulness-Based Stress Reduction) 8-week program | B | 4 | 4 | 1 |
| Zinc | B | 4 | 4 | 1 |
| Neuromuscular Electrical Stimulation (NMES) after Knee Replacement | A | 11 | 11 | 3 |
| Berberine 1.0 to 1.2 g/day (with meals) | B | 7 | 7 | 2 |
| Inspiratory Muscle Training | B | 10 | 10 | 3 |
| Aged garlic extract (480 mg/day for blood pressure; 2.4 g/day in plaque trials) | B | 3 | 3 | 1 |
| Omega-3. High-dose EPA (icosapent ethyl, 4 g) | B | 3 | 3 | 1 |
| Zone 2 aerobic training | A | 3 | 3 | 1 |
| Mindfulness meditation for sleep (weekly group course with daily home practice) | B | 8 | 8 | 3 |
| Avoiding ultra-processed food | B | 7 | 7 | 4 |
| SGLT2 inhibitors (empagliflozin, dapagliflozin) | A | 3 | 3 | 3 |
| Semaglutide | A | 3 | 3 | 3 |
An unindexed trial is still a trial.
This page could be misread badly, so here is the limit of it stated first. A missing design type means the National Library of Medicine's indexers did not assign one. It does not mean the study was not a trial. Indexing is done by people, at scale, sometimes years after publication, and it is patchy.
One of our own citations is titled Pterostilbene on metabolic parameters: a randomized, double-blind Neither the participants nor the people assessing them know who received the real treatment until the trial ends., and placebo-controlled trial. PubMed assigns it no study design whatsoever. It also carries a public registration, which anyone can look up. So the paper says randomized, the registry backs it, and the index is silent. Three sources, one of them missing, and the missing one is not the study.
For that reason nothing on this page is treated as evidence against a study, and no letter grade on this site has ever been set or changed by it. The build check that produces these figures will only speak about an A or a B when every single citation on the record lacks both an indexed design and a registration, which is a much narrower claim than any one number here.
The same caution runs the other way. A registration proves a study was declared. It does not prove the study was good, that it was large, that it measured something that matters, or that what was published matches what was declared. Checking that last one requires reading the registration against the paper, which is a separate piece of work and is not what this page does.
Generated, not asserted.
The design types and registration accessions are harvested from PubMed by a script, committed as data, and rendered from that data at build time. The page is reproducible from the repository alone. A separate build check refuses the deploy if a graded citation is missing from the trace, if the trace carries an identifier no grade cites any more, or if an A or B grade has no independently recognisable trial anywhere beneath it.
Where that last check has found something, the record is named in the build log and left for editorial review rather than quietly excused. That register only shrinks.