Peptide Evidence Map: methods, field definitions and change log =============================================================== Version 2, November 20, 2026 https://mypromise.com/research/peptide-evidence-map Files ----- peptide-evidence-map.csv — one row per compound entry (38 rows, 36 compounds) https://mypromise.com/data/peptide-evidence-map.csv peptide-evidence-map-studies.csv — one row per compound and record pair https://mypromise.com/data/peptide-evidence-map-studies.csv peptide-evidence-map.json — the same dataset as JSON, including PubMed's search warnings https://mypromise.com/data/peptide-evidence-map.json peptide-evidence-map-methods.txt — this file Promise chose 36 compounds marketed or widely discussed by U.S. telehealth and wellness sellers, including compounds prescribed through Promise. Semaglutide, tirzepatide and oxytocin serve as reference points with approved medicines. This is not a complete list of peptides. 5-Amino-1MQ, MK-677, SLU-PP-332, NAD+, NR and NMN are not peptides. Dihexa is a modified peptide and glutathione is a tripeptide. Of the 28 compounds on this list with no matching FDA product, 14 had no completed interventional study in the registry search. The table has 38 entries because it shows full-length thymosin beta-4, the TB-500 fragment name and their combined count. Compound totals use the combined entry once. NR and NMN count separately. Trials can appear under several compounds or in both databases. What we searched ---------------- On November 20, 2026 UTC, we searched ClinicalTrials.gov intervention names and PubMed titles and abstracts using compound names and aliases. Separate PubMed filters selected human randomized-trial candidates and animal-only records. The broad name-search count, earliest publication year and animal-only figures are automated. Two AI screening passes independently applied written rules and checked decisions against the registry or PubMed record. For low-count entries, the second pass checked proposed changes and a sample. For selected mid-count entries, it checked every candidate. A third pass settled disagreements against primary records. Follow-ups checked all six afamelanotide papers and all three candidates for Melanotan II and Selank. One Selank paper documented random assignment. The older Selank paper and the Melanotan II crossover confirmed exposure but left chance assignment unresolved. On November 20 we also reran nine exact saved randomized-trial queries. All nine returned zero candidates again. The main CSV and the JSON give the frozen and repeat counts side by side, with retrieval times. The JSON includes PubMed’s search warnings. An empty result has no papers to screen. What counts ----------- ClinicalTrials.gov calls a study interventional when researchers assign people a treatment or test. We include the compound given alone or in a studied combination. Registration includes plans at any stage. Completed means the registry says Completed. Posted results are results on the registry. The earliest start year can be a planned date. We excluded diagnostic probes, reference arms used solely to study another treatment, measurements of the body’s own peptide and different molecules. We held out unclear identities or roles. The two TESICO registrations describe one aviptadil comparison, so we counted the substudy once. We kept eight registrations from one sponsor that list the same single site. ClinicalTrials.gov does not verify recruitment for any sponsor. Two descriptions use the phrases “This example trial” (https://clinicaltrials.gov/study/NCT07437560) and “this public example” (https://clinicaltrials.gov/study/NCT07487363). The CSV also gives counts with these entries set aside. Reported enrollment adds actual enrollment across all study groups in completed records, including controls. A person can appear in more than one study. We report how many completed records lack an actual count. “Not reported” means none supplied one. A qualifying PubMed paper reports a randomized comparison in which people received the compound. Treatment, dose, route and physiology comparisons can qualify, as can secondary, interim and pooled reports of verified exposure. Several papers can represent one trial and vice versa. We grouped papers using study identifiers and published methods. Unresolved allocation and relevant publication notices kept some papers outside the count. Positive automated totals still count candidates. The supplementary review followed selected leads in other registries and publications. Each addition required documented human administration and a distinct cohort. The 101 distinct studies count the shared Selank/Semax study and overlapping thymosin entries once. All three thymosin additions involved the full-length form. “Other” includes unclear designs. Zero means the review found no eligible addition. “Not reviewed” means it did not cover that entry. How to reproduce a cell ----------------------- Open the compound’s search in the main CSV’s sources field. In the study CSV, filter by slug and count membership. Count unique PubMed trial IDs for a trial total. Sum participants on records marked “completed participants” for enrollment. The 5,056 study rows identify contributors and give reasons for removals. The build archive keeps the original broad and animal-only matches. The field definitions below describe each column. The change log at the end dates corrections. PubMed changes over time. Where the original search saved totals alone, same-day replays reconstructed candidate lists. Record histories identified later additions to set aside. Matching counts cannot establish identical membership. Older or non-U.S. studies may be missing. Russian-language publications can be hard to find under English names. Some records lack an abstract or accessible methods. Blank fields indicate missing information or review coverage. In the FDA column, “None” means no matching product was found. Marketing status comes from FDA records. WADA entries cover explicit mentions for the listed year or monitoring program. Field definitions ----------------- peptide-evidence-map.csv is a UTF-8 comma-separated file with one header and 38 data rows. It tabulates a dated snapshot of human study registrations, PubMed publications and additional verified human studies, alongside FDA product history and named sport-list entries. The 38 entries cover 36 compounds. Count the combined thymosin beta-4 entry once. NR and NMN are separate precursors. The counts describe research volume. You may quote and reuse these figures with a link to the source page Suggested citation Promise. Peptide Evidence Map. Version 2, November 20, 2026. https://mypromise.com/research/peptide-evidence-map When quoting a number, give the compound, the column and the snapshot date. How to read blanks and basis values hand_verified — the figure reflects a review against source records. Registry figures can apply checked corrections to an automated starting set. automated_not_hand_checked — the figure comes from an automated search or screen. A positive automated PubMed randomized-trial figure counts candidate papers. A zero for that search means the frozen filter returned no candidates. A blank value can signal unavailable information or a row outside review scope. Use notes to tell which applies. Zero is an observed count in the stated search. peptide-evidence-map.csv columns slug — stable row ID used to match the study download. Compound — display name for the molecule, group or related substance. Row kind — compound for a single entry, combined_union for the deduplicated thymosin beta-4 and TB-500 union, or separate_precursor for NR and NMN. Registered human interventional studies — ClinicalTrials.gov records at any status for studies assigning a named intervention to people. Registered studies excluding flagged sponsor records — the same count with the flagged single-sponsor entry set aside. Nine rows mark eight registrations because the TB-500 fragment and combined rows share a record. ClinicalTrials.gov does not verify recruitment for any sponsor. Completed registered studies — included records with registry status Completed. Reported enrollment — the sum of reported actual enrollment across all study groups, including comparison groups. People can appear in more than one study. Completed studies without actual enrollment — completed records missing a numeric enrollment labeled actual. Registered studies with posted results — included records showing results on ClinicalTrials.gov. Earliest registered start year — earliest start date among included records. This can be a planned date. PubMed randomized human trial records — indexed papers reporting eligible randomized human research, or candidate papers where the basis is automated. Distinct randomized human trials represented on PubMed — identifiable trials behind those papers. One paper can cover multiple trials and vice versa. Frozen PubMed candidates and Replay PubMed candidates — paired automated counts for the nine exact saved randomized-trial queries replayed on November 20, 2026. Both are zero in every pair. Replay UTC gives the retrieval time. Blank fields mean no final replay was requested. The frozen snapshot is the basis for map counts. PubMed animal-only query matches — automated hits from the animal-only filter. PubMed all-records query matches — automated name-search hits of all designs. First PubMed query publication year — year of the oldest indexed name-search hit. Additional randomized human studies — extra studies assigning groups by chance. Additional non-randomized human studies — extra trials using other assignment. Additional human case series — extra reports describing a group of cases. Additional human studies of other or unclear design — extra human studies whose design differs from those groups or could not be established. Total additional verified human studies — sum of the four design columns for the row. The review checked selected leads outside the main counts. Across entries, 105 study mentions represent 101 distinct studies. The full and combined thymosin rows overlap, and one Selank/Semax study appears twice. Each count or year column's basis — how the preceding value was obtained. Matching FDA product in its history — whether the FDA ever approved a product containing that exact molecule for a specific use. FDA product names — names recorded in the product review. First FDA product approval year — earliest confirmed year for a matching product. Approved product uses (label excerpts) — FDA label wording for matching products. FDA marketing status — status of the matching product or presentation. WADA 2026 Prohibited List and WADA 2027 Prohibited List — confirmed explicit entries on each World Anti-Doping Agency list. WADA 2026 Monitoring Program and WADA 2027 Monitoring Program — confirmed entries in each edition of the monitoring program. sources — primary records and searches supporting the figures. Multiple links are separated by a vertical bar. notes — limits and exceptions for the row, including unresolved record IDs. peptide-evidence-map-studies.csv columns The study download has 5,056 data rows. A row is a compound and record pair. It includes count contributors and candidates removed from the original counts in screening. The build archive retains the broad and animal-only query matches. slug — row ID in the main CSV. id — registry, publication or protocol ID for the record. registry or journal — where the record was filed or published. year — registry start year or publication year where known. design — the study design or the candidate's publication type. participants — number reported for the population described in the next field. participants basis — which group or enrollment measure the number describes. status — registry status, publication screening status or study status. counts — the v2 study figures the record contributes to. v1 counts — corresponding count membership in the original search or screen. PubMed trial IDs — distinct trials represented by the paper where traceable. Removal reason — why the review set aside a candidate that contributed to v1. Flagged registration lead sponsor — the registry name for the eight flagged IDs. URL — link to the source record. Change log ---------- November 20, 2026 — Version 2 published. Confirmed corrections will be listed here with their dates.