Restless legs syndrome: sources, guidelines, and methodology

A topic sweep produced 22 prose syntheses — one per subject area (iron therapy, dopamine-agonist augmentation, tapering and deprescribing, genetics and epidemiology, cross-guideline comparison, and 17 others) — each one citing the individual study notes behind it. A language sweep searched 25 non-English language groups (Arabic, Czech/Slovak, German, Greek, Spanish, Persian, Finnish, French, Hebrew, Hindi, Hungarian, Indonesian/Malay, Italian, Japanese, Korean, Dutch, Polish, Portuguese, Romanian, Russian, Scandinavian (Swedish/Norwegian/Danish), Turkish, Ukrainian, Vietnamese/Thai, and Chinese) for non-English findings, national guidelines, and patient organizations; verified entries currently exist in 11 of those languages — most sweeps that found nothing citable are recorded as such rather than silently dropped.

Every note — one per study, trial, or guideline — follows the same frontmatter-plus-sections template: a stable id, title, authors, year, journal, a PMID/DOI/URL, source language, study type, sample size, population, controlled topic tags, and an evidence grade, followed by a key finding, the intervention or exposure, outcome numbers, relevance to a non-drug long-term plan, and limitations (per the note template in docs/00_MASTER_PLAN.md §5).

Before anything is cited on this site, its PMID, DOI, or URL is independently checked against the source record. As of the last verification pass:

The single unverifiable note is excluded from this catalog and from the clinical protocol until it can be independently confirmed.

Browse the full, searchable catalog on the Knowledge base page →

The long-term protocol was produced in three independent drafts — risk-first, evidence-first, and patient-first — each written separately from the same knowledge base, then reviewed by an independent judging pass and synthesized into a single document (docs/00_MASTER_PLAN.md §6, phase R2-B). Every one of its 40 load-bearing claims was then re-checked, one by one, against the cited knowledge-base note, the PubMed record for its PMID, and any contradicting note elsewhere in the knowledge base — an adversarial verification pass, followed by a separate line-by-line honesty audit against the project’s own non-negotiable rules, including a full citation-existence check and a 15-note random numeric spot-check.

Verdict: 32 of the 40 claims survived unchanged. Seven were rewritten because they overstated their source — a percentage generalized past the population it was measured in, a small case series graded as if it were a large trial, a pooled effect size presented without noting it was driven by a different patient population. One numeric figure — a taper rate repeated widely elsewhere — was deleted outright because no source in this knowledge base contains it.

A separate mechanistic simulation models iron depletion, dopamine signalling, and dopamine-agonist withdrawal over time. It is explicitly a mechanistic model, not a predictor: by its own specification it cannot forecast what will happen to any particular person, carries no patient-specific calibration data, and its 0–10 symptom-index output is only loosely mapped to a real IRLS score — a difference of less than about one point is not meaningful. Every figure and data export it produces carries the same label: “Mechanistic model output — not clinical prediction.”

See the simulation →

Medical disclaimer. RLS is a chronic neurological condition. No cure is claimed here, and nothing on this site is individual medical advice. Every factual statement is tied to a published study in this knowledge base, with an evidence grade attached, so you can see how strong it is. Dopamine-agonist withdrawal and augmentation are dangerous when mishandled: any change to medication must happen only under physician supervision, and a dopamine agonist must never be stopped abruptly. Where content separates established evidence from emerging evidence from this project’s own simulation’s hypotheses, take that separation seriously — simulation output is never evidence.

This is the long form. The condensed version travels with you on every page, in the banner above the header — it can be collapsed to one line, but it is never removed.

• This project could not confirm the specific numeric thresholds in the current Spanish (SEN/SES) or full Japanese national guidelines — the Spanish document was an unreadable image scan, and the Japanese guideline was reachable only via a secondary narrative review — and found no dedicated, indexed national RLS guideline for Canada, Australia, China, or India. Those gaps are recorded, not filled in with a guess.
• One knowledge-base note remains unverifiable — its citation could not be confirmed against any indexed bibliographic database — and is excluded from the public catalog and the protocol until that changes.
• The last full citation-verification pass ran on 2026-09-12; this catalog was last regenerated 2026-09-15. Research published after those dates is not yet reflected here.
• Three knowledge-base notes carry a frontmatter evidence grade that the verification pass judged too high; the protocol already uses the corrected grade, but the notes themselves and the generated catalog have not yet been regraded in place (see the verification log above, “Follow-ups outside this document”).
• The brain-and-dopamine simulation is mechanistic, not predictive — see “About the simulation” above.