The EBL grade dataset
Every letter EBL publishes on its graded records, in one file with a version, for anyone who needs to cite a grade as it stood on a given day. Each row is one letter: the record, the outcome it was earned on, the letter and the PubMed IDs behind it.
The file is written from the EBL database on every build and checked against it before the site publishes, so it never disagrees with the record pages.
Download
Version 2026-10-ff52370836: 248 graded records and 390 letters for specific uses, 638 rows in all. The records’ headline letters: 27 A, 67 B, 112 C, 42 D.
- ebl-grades.csv, a spreadsheet, one row per letter
- ebl-grades.json, the same rows with the grade scale, the column definitions and this version’s details
Row fingerprint (SHA-256): ff5237083675f90baf21ad626dc7e0725b6ee5eb0c7e23c04389ce806edcc3af. The version changes whenever any letter, outcome or citation changes, so a version names exactly one set of letters.
How to cite it
Arnold Sandlow, D.C. Evidence Based Longevity grade dataset, version 2026-10-ff52370836. Evidence Based Longevity, Inc.; 2026. https://evidencebasedlongevity.co/dataset
A DOI has not yet been issued for this dataset. Until it is, cite the version and this page.
License
The dataset is published under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). The rest of this site stays under the Terms of Use.
What each column holds
record_key: Collection and record id, joined by a slash. Stable across versions.collection: supplements, protocols, device_modalities or longevity_rx.record_id: The record's id in the EBL database.record_name: The record's name as published.kind: Supplement, Protocol, Device or modality, Longevity Rx, or Not counted as a supplement.level: record: the record's headline letter. use: a letter for one specific use.scope: For a use row, the use the letter was earned on. For a record row, the record's headline scope as its page states it.grade: A, B, C or D, as defined in the grade scale in the JSON file.endpoint_type: hard, functional, clinician_rated, self_report or surrogate, as the record states it.evidence_type: The kind of evidence the record's headline letter rests on, as the record labels it. Record rows only.pmids: PubMed IDs behind this letter, separated by semicolons. Empty only for a D that rests on a documented search finding no trial; the search is on the record's page.last_reviewed: Month of the record's latest review (YYYY-MM).url: The record's page on evidencebasedlongevity.co.
What is not in it
The written evidence behind each letter (notes, as-studied details, grade readings) is published one record at a time on each record's page and is not in this dataset. Rows are ordered by collection and record name, never by letter. A letter answers one question about one outcome; two letters on different outcomes are not a ranking. Every change to a letter is published, with the trial that moved it, in the Evidence Log.