Educational content, not medical advice. Never stop or change RLS medication without your doctor; dopamine agonist withdrawal must be medically supervised.

RLS Atlas

Full evidence protocol

The long-term restless legs syndrome management protocol

Every recommendation in the plain-language plan, in full, with its citation, PMID and evidence grade. This is the document a physician would want to see.

This is the full evidence-based protocol behind the plain-language plan: every recommendation carries its citation, PMID, and evidence grade (A–E), for a physician or a highly engaged reader who wants the primary sourcing.

← Back to the long-term plan overview

Long-term RLS management protocol

Synthesised from the three drafts in docs/synthesis/ and judged in docs/synthesis/JUDGING.md: built on the evidence-first draft, grafting the risk-first safety architecture and the patient-first practicality layer. Written against docs/00_MASTER_PLAN.md §2–3.

Preface — what this is, and what it is not

What it is. An evidence-based, long-term management path for adults with confirmed Restless Legs Syndrome already on daily medication — usually a dopamine agonist (pramipexole/Sifrol/Mirapexin, ropinirole, rotigotine) or an alpha-2-delta ligand (gabapentin, pregabalin, gabapentin enacarbil). It builds a non-drug foundation, removes what makes RLS worse, and aims to reduce or, where the evidence supports it, eliminate the need for daily dopaminergic medication — because those agonists carry an augmentation incidence of roughly 7–10% per year (2026-winkelman-rls-jama-clinical-review.md, PMID 41563785, grade D — the source uses that figure to argue against starting an agonist; extending it to de-prescribing established therapy is this document's own reasoning) and AASM 2025 now recommends against pramipexole, ropinirole and rotigotine as standard therapy for that reason (2025-winkelman-aasm-guideline-rls.md, PMID 39324694, grade D).

What it is not. Not a cure — RLS is chronic. Not individual medical advice. Above all, not a self-directed taper manual: every medication change, especially any dopamine-agonist reduction, must be made only under physician supervision and never abruptly, because dopamine agonist withdrawal syndrome (DAWS) can be severe, resists dopaminergic rescue, and has no proven treatment (2013-nirenberg-dopamine-agonist-withdrawal-review.md, PMID 23686524, grade D; 2026-lobinska-dopamine-agonist-withdrawal-syndrome-review.md, PMID 41870480, grade D). It does not claim everyone should stop medication — Phase 4 says when staying on a drug is right.

Evidence grades (as recorded on each knowledge-base note). A = meta-analysis or 2+ RCTs; B = single RCT or large cohort; C = small/observational; D = guideline consensus; E = mechanistic/animal.

The interpretive anchor for every number below. On the 0–40 IRLS the group-level minimal clinically significant change is about 3 points and the individual threshold about 6 (2013-allen-irls-minimal-clinically-significant-change.md, PMID 24051118, grade C), "much improved" about −9 to −11 (2016-ondo-irls-rls6-mcic-rotigotine.md, PMID 26446245, grade B), and the pooled placebo response across 85 RCTs is −6.58 points, nocebo rate 45.4% (2017-silva-placebo-nocebo-meta-analysis.md, PMID 28490647, grade A). Any individual improvement under about 6 points is inside the placebo band and is not proven benefit — including improvements produced by this protocol.


1. Who this is for, and red flags requiring specialist care first

For adults with IRLSSG-confirmed chronic-persistent RLS already taking a daily dopamine agonist or alpha-2-delta ligand, who want to lower medication burden, avoid or reverse augmentation, and build a durable non-drug foundation. Written for use with a physician, not instead of one.

Obtain specialist (sleep-medicine or neurology) assessment before any step below if any of the following apply — each is a situation where the generic path can cause harm.

  • Suspected augmentation with dose escalation — earlier daily onset (a 4-hour advance is the single best Max Planck criterion), spread to arms or trunk, shorter latency at rest, shorter drug-effect duration, or paradoxical worsening after a dose increase (2007-garcia-borreguero-max-planck-augmentation-criteria.md, PMID 17544323, grade D). Raising the dose to chase worsening is contraindicated by every guideline reviewed (2016-garcia-borreguero-augmentation-prevention-treatment-guidelines.md, PMID 27448465, grade D) and is this disease's most damaging error.
  • Impulse control disorder (gambling, shopping, hypersexuality, binge eating): 7.1% (10/140) by structured interview in a cohort treated with a dopamine agonist or levodopa — 8 of the 10 affected patients were on an agonist, so an agonist-specific rate is not separately reported (2011-voon-impulse-control-behaviours-cohort.md, PMID 21955669, grade C) — and 10% in an 8-year Mayo cohort where augmentation reached 42% and median dose rose 0.38 to 1.0 mg/day (2012-lipford-pramipexole-eight-year-cohort.md, PMID 23036265, grade C); it may persist after withdrawal and is a major DAWS risk factor (PMID 23686524, grade D).
  • Psychiatric comorbidity — RLS roughly doubles suicide/self-harm risk (adjusted HR 2.66, 95% CI 1.70–4.15, rising to 4.14 after excluding depression, insomnia and OSA; n=169,373) (2019-zhuang-suicide-self-harm-risk-rls.md, PMID 31441941, grade B), and a taper transiently worsens symptoms — exactly when this risk concentrates.
  • Pregnancy or planning pregnancy — pooled prevalence 21% in pregnancy and 4% at a cross-sectional postpartum time point — a population average, not a guarantee of resolution for any individual (2018-chen-pregnancy-prevalence-meta-analysis.md, PMID 29169861, grade A); iron-first, non-dopaminergic management under obstetric care (2014-picchietti-irlssg-pregnancy-lactation-guideline.md, PMID 25553600, grade D).
  • CKD or dialysis — RLS prevalence 51.7% on haemodialysis (2025-elangovan-ckd-dialysis-iron-prevalence.md, PMID 40677478, grade C), and gabapentin's half-life extends from 5–9 h to as much as 132 h, causing severe but reversible neurotoxicity if not aggressively renally dose-adjusted (2022-lehmann-gabapentin-overdose-dialysis-case-report.md, PMID 36518357, grade E).
  • Current opioid therapy, substance use disorder, or an older adult with falls, cognitive impairment, COPD/OSA or driving-dependent work — the gabapentinoid plus opioid combination raises dizziness (OR 3.26), cognitive dysfunction (OR 3.13), respiratory depression (OR 1.71, 95% CI 1.31–2.24) and, in non-randomised data, mortality (OR 2.76) (2022-hahn-gabapentinoid-opioid-combination-meta-analysis.md, PMID 36304170, grade A), plus falls in older adults (aHR 1.18, 95% CI 1.02–1.37 — Medicare beneficiaries treated for chronic pain and anxiety rather than RLS, median exposure 26 days, so an extrapolation) (2021-shah-raji-opioid-gabapentinoid-falls-medicare.md, PMID 35476819, grade B); the 2019 FDA respiratory warning remains active (2020-shrestha-palaian-fda-gabapentinoid-respiratory-warning.md, PMID 33728039, grade D). Ten per cent of long-term pramipexole users reported sleep attacks while driving (PMID 23036265, grade C), with no denominator-based crash study — unquantified, not absent.
  • Sudden severe worsening, new neurological signs, or suspected secondary cause — new anaemia, neuropathy or Parkinsonian features; RLS precedes Parkinson disease with OR 4.19 (95% CI 3.91–4.50) for any prior RLS diagnosis, and OR 3.73 (95% CI 3.39–4.09) when the RLS diagnosis fell 5–10 years before the PD diagnosis (2023-schrag-parkinson-prodromal-risk-factors.md, PMID 36342675, grade B).

2. Foundations

2.1 Confirm the diagnosis (grade D, universal guideline consensus)

All five IRLSSG 2014 essential criteria must be met: (1) urge to move the legs, usually with unpleasant sensations; (2) worsening at rest; (3) relieved by movement while it continues; (4) worse in the evening or night; (5) not solely accounted for by another condition — the criterion added in 2014 to exclude mimics (2014-allen-irlssg-updated-diagnostic-criteria.md, PMID 25023924, grade D). Exclude leg cramps, positional discomfort, painful peripheral neuropathy (RLS occurs in about 30% of polyneuropathy clinic patients; 2006-gemignani-polyneuropathy-rls.md, PMID 16685701, grade C), venous insufficiency, akathisia and foot tapping; where doubt remains, the Suggested Immobilization Test is the validated objective provocation (2002-michaud-suggested-immobilization-test.md, PMID 11835447, grade B).

Re-confirm at every worsening: worsening on a dopamine agonist is augmentation until proven otherwise, not disease progression. Success: five criteria documented, mimics excluded, plus a baseline IRLS and the usual symptom-onset clock time — the reference against which all future augmentation is judged. If it fails: refer; never taper an uncertain diagnosis.

2.2 Baseline laboratory panel

  • Ferritin and TSAT (morning draw, at least 24 h off iron) — a grade-D good-practice statement with A/B-graded downstream treatment. AASM 2025 directs testing in all clinically significant RLS (PMID 39324694), and the same 75 µg/L threshold appears in the IRLSSG iron guideline (2018-allen-irlssg-iron-treatment-guidelines.md, PMID 29425576, grade D) and the German DGN/DGSM guideline (2024-trenkwalder-german-dgn-dgsm-guideline.md, PMID 39501372, grade D). A "normal" ferritin does not exclude an iron contribution (2018-dauvilliers-hepcidin-severity-rls.md, PMID 29418021, grade C).
  • CBC, eGFR, B12, folate, TSH, fasting glucose/HbA1c — grade D, and honestly so: no RLS-specific trial supports any of these as treatment targets; they are comorbidity and mimic screening (type 2 diabetes carries 28.3% RLS prevalence; 2018-akin-type2-diabetes-rls-prevalence.md, PMID 30213520, grade C). Vitamin D: do not treat RLS with it — the dedicated RCT (50,000 IU/week, 12 weeks) raised serum levels but produced no RLS improvement (2018-wali-vitamin-d-replacement-rct.md, PMID 30430372, grade B; 2024-gonzalez-parejo-dietary-supplements-systematic-review.md, PMID 39064758, grade A).
  • Serum phosphate — mandatory before and after every ferric carboxymaltose course (Phase 1).

2.3 Medication review for aggravating drugs (grade A/B — highest yield, lowest risk)

In 340,099 monitored inpatients, over 80% of drug-induced RLS traced to two drugs: mirtazapine (39 cases) and quetiapine (16), usually within 1–2 days of a start or dose change (2026-fugger-secondary-rls-pharmacovigilance.md, PMID 42251748, grade B); prospectively, mirtazapine provoked or worsened RLS in 28% of starts versus 0% on reboxetine (2008-rottach-second-generation-antidepressants-rls.md, PMID 18468624, grade C — an uncontrolled, unblinded prospective office series whose total and per-drug denominators are not reported, so read both percentages as provisional). Metoclopramide is a third clear aggravator (RLS in 75% of systemic sclerosis patients taking it versus 26.3% not, p=0.02; 2013-ostojic-metoclopramide-systemic-sclerosis-rls.md, PMID 23456369, grade C), and acid-suppression drugs associate with RLS across two national blood-donor cohorts, independently of measured iron stores: PPIs in both (OR 1.43 US, 1.27 Denmark), H2 antagonists significantly only in the US cohort (OR 1.56; Denmark OR 1.18, 95% CI 0.92–1.53, p=0.2), with the combined PPI/H2 category significant in both (OR 1.41 and 1.29) (2021-earley-ppi-h2-antagonist-rls-denmark-us.md, PMID 33119070, grade B).

Replacements. Bupropion has the most consistent RLS-neutral-to-beneficial evidence (2017-kolla-antidepressants-rls-systematic-review.md, PMID 28822709, grade A). Do not blanket-ban SSRIs — in WHO VigiBase they were the only antidepressant class not significantly associated with RLS (2020-revet-antidepressants-movement-disorders-vigibase.md, PMID 32546134, grade C).

Ask explicitly about over-the-counter melatonin. 3 mg at about 19:00 significantly increased the SIT PLM index in a small, unblinded, non-randomised within-subject case series (n=8), while bright light modestly reduced sensory discomfort (2010-whittom-melatonin-bright-light-rls.md, PMID 20226733, grade C), with a corroborating ramelteon pharmacovigilance signal (2025-wu-ramelteon-faers-rls-signal.md, PMID 40624487, grade C). Patients rarely make the connection themselves.

Honest gap: sedating first-generation antihistamines are widely assumed aggravating and are mechanistically plausible, but no dedicated controlled or cohort study was found — extrapolation, not evidence. Success: every aggravating agent stopped, substituted, or explicitly justified. If it fails: proceed, documenting the residual aggravator as a confounder of every later outcome judgement.


3. The phase plan

Phase 0 — Prepare (weeks −4 to 0). Grade D.

Complete section 2. Start a symptom diary recording onset clock time, body distribution, severity 0–10, medication and time, sleep, alcohol and any new medicine — diary methodology is validated to detect real triggering relationships (OR 6.94 for a same-night trigger) (2015-chen-migraine-rls-diary-study.md, PMID 26195587, grade C). Screen for ICD, mood and suicide risk, falls and driving; record baseline IRLS; agree a written plan with the prescriber. Structure matters: in routine practice 85% of augmented patients stay on a dopamine agonist and only 20% of those are dose-reduced (2025-martin-garcia-spanish-augmentation-cohort.md, PMID 41130145, grade C), whereas a guideline-based cross-titration cohort achieved 78% responders (2022-yeung-laiwah-augmentation-guideline-effectiveness.md, PMID 35532181, grade C).

What to ask the prescriber. "Can we check ferritin and transferrin saturation?"; "Do I meet the Max Planck augmentation criteria?"; "Which of my other medicines could be making this worse?" RLS e.V. (restless-legs.org) and Association France Ekbom (france-ekbom.fr) publish guideline summaries patients can bring. Success: complete diary, full lab panel, written plan. If it fails: fix adherence first — the diary is the only augmentation early-warning system available, and an unstructured taper is the failure mode this protocol exists to prevent.

Phase 1 — Iron repletion (weeks 0–16; judge at 12 weeks, not 4). Grade A efficacy, D thresholds.

Thresholds D. Ferritin at or below 75 ng/mL or TSAT under 20%: oral or IV iron; ferritin 75–100 ng/mL: IV only (PMID 39324694). IRLSSG positions FCM 1000 mg as possible first-line at ferritin below 300 µg/L (PMID 29425576).

Oral iron (grade A/B). About 65 mg elemental iron as a single morning dose on alternate days, with vitamin C, away from milk/tea/coffee — a dose above 60–100 mg raises hepcidin for roughly 24 h and blocks the next day's absorption (2017-stoffel-alternate-day-iron-absorption.md, PMID 29032957, grade B, non-RLS mechanism). First positive oral-iron RCT: IRLS −10.3 versus −1.14, p=0.01 (2009-wang-oral-ferrous-sulfate-rct.md, PMID 19230757, grade B); pooled across the 7 trials and 345 participants contributing that outcome (inside a Cochrane review of 10 trials / 428 participants overall), IRLS MD −3.78 (95% CI −6.25 to −1.31) (2019-trotti-cochrane-iron-review.md, PMID 30609006, grade A) — below the −6.58-point pooled placebo response (2017-silva-placebo-nocebo-meta-analysis.md, PMID 28490647, grade A), so oral iron is modest, not transformative.

IV ferric carboxymaltose, 1000 mg total (grade A/B). IRLS −11.9 versus −7.88, p=0.03, with 37.5% medication-free at 30 weeks (2016-cho-ferric-carboxymaltose-rct-remitters.md, PMID 27823710, grade B); −8.0 versus −4.8, p=0.0036, n=209 (2024-earley-iv-ferric-carboxymaltose-rct.md, PMID 38625730, grade B); meta-analytic WMD −6.03 (95% CI −10.11 to −1.96, p=0.004), 7 RCTs, 537 patients (2024-qadri-iv-fcm-meta-analysis.md, PMID 39326219, grade A). Dose may matter: 500 mg failed to beat placebo (−8.3 versus −4.8, p=0.100; 2018-cho-fcm-500mg-dose-rct.md, PMID 29458749, grade B), suggesting — by cross-trial comparison, not a head-to-head dose-ranging trial — that 1000 mg may be the minimum effective total dose, and the largest trial missed its co-primary CGI endpoint (35.5% versus 28.7%, p=0.30), so iron is not a guaranteed substitute for medication. Why iron precedes any taper: in that same protocolised-taper RCT iron recipients needed rescue medication far less often — 32.7% versus 59.4%, p=0.0002, ARR 26.7%, NNT about 4 (PMID 38625730, grade B).

Timeline. Do not judge IV iron at 4 weeks: a 1000 mg dose missed its week-4 endpoint (p=0.163) and reached significance at week 12 (p=0.021) (2017-trenkwalder-fcm-nonanemic-rct.md, PMID 28643901, grade B). Success at 12 weeks: ferritin above 100 ng/mL, TSAT above 20%, IRLS improvement at least 6 points. If it fails: re-check adherence and absorption (PPI/H2 antagonists, PMID 33119070), escalate oral to IV, and consider a coeliac/H. pylori/SIBO workup — but do not repeat infusions indefinitely.

Safety and stop-rules B. FCM causes hypophosphataemia below 2.0 mg/dL in 50.8% of recipients versus 0.9% with ferumoxytol, persisting at 5 weeks in 29.1% (2018-wolf-ferric-carboxymaltose-hypophosphatemia-rct.md, PMID 30518682, grade B — regraded from A: a single RCT, which this project's rubric grades B; and its n=1997 population was iron-deficiency anaemia, not RLS, so the absolute rates are extrapolated), and cumulative dosing has produced osteomalacia and fractures (2025-magagnoli-fcm-hypophosphatemia-systematic-review.md, PMID 39935027, grade C). Stop-rule: phosphate below 2.0 mg/dL, or bone pain or proximal weakness — stop FCM, correct phosphate, switch formulation; where repeated courses over years are likely, prefer a non-FCM formulation such as ferric derisomaltose. Long-term repeated-dose iron safety in RLS is not established.

Phase 2 — Supervised transition off the dopamine agonist. Grade C/D throughout.

Only under physician supervision. Never abrupt.

When indicated. Confirmed augmentation; impulse control disorder; dose at or above the licensed maximum; or informed patient preference. Not indicated as a reflex in a stable, low-dose, non-augmented patient — the German 2024 guideline still lists non-ergot agonists as co-first-line (PMID 39501372, grade D), and the dopamine-agonist items were the only AASM 2025 recommendations with substantial international dissent (2025-ferri-world-sleep-society-endorsement.md, PMID 40834711, grade C).

Sequencing — add before you subtract (grade D, safety-driven). Establish the replacement at an effective dose before reducing the agonist; even small reductions can trigger severe rebound (2022-winkelman-treating-refractory-augmented-rls.md, PMID 35609673, grade D). Add-then-taper versus taper-then-add has never been tested head-to-head.

Bridge and destination options.

  • Alpha-2-delta ligand — the destination (grade A/B). Pivotal head-to-head (n=719): augmentation 2.1% on pregabalin 300 mg versus 7.7% on pramipexole 0.5 mg over 40–52 weeks, p=0.001 (2014-allen-nejm-pregabalin-pramipexole-augmentation.md, PMID 24521108, grade B) — ARR 5.6%, NNT about 18/year. Report the harm beside the benefit: that trial recorded 6 suicidal-ideation cases on pregabalin versus 2–3 on pramipexole. Gabapentin enacarbil 1200 mg: CGI-I responders 77.5% versus 44.8%, p<0.0001, NNT about 3 (2011-lee-gabapentin-enacarbil-pivot-rls-ii.md, PMID 21677899, grade B), efficacy unchanged by prior agonist exposure (2015-ondo-gabapentin-enacarbil-prior-agonist-exposure.md, PMID 26788345, grade B), benefit holding at 9 months in randomised withdrawal (relapse 9% versus 23%, OR 0.353, p=0.02) (2010-bogan-gabapentin-enacarbil-long-term-maintenance.md, PMID 20511481, grade B). EU caveat: gabapentin enacarbil has no EMA authorisation (2019-wanner-nondopaminergic-vs-dopaminergic-europe-access.md, PMID 31229171, grade D), so EU plans default to off-label gabapentin or pregabalin; not every gabapentinoid trial is positive (2025-sunwoo-pregabalin-korean-negative-rct.md, PMID 40635537, grade B).
  • Rotigotine patch — intermediate bridge only C. Cross-titrate −1 mg ropinirole or −0.25 mg pramipexole against +1 mg rotigotine every two days: 85% switch success at 5 weeks, but only 50% retention at 12 months, and rotigotine augments too (13% at 5 years) (2016-winkelman-rotigotine-cross-titration-method.md, PMID 27810181, grade C; 2014-bogan-rotigotine-bench-to-bedside-review.md, PMID 24636821, grade D).
  • Low-dose opioid — reserved bridge B. Prolonged-release oxycodone-naloxone: IRLS −16.5 versus −9.4, difference 8.15 (95% CI 5.46–10.85), p<0.0001, n=306 (2013-trenkwalder-oxycodone-naloxone-rct.md, PMID 24140442, grade B); registry doses broadly stable (median 30.0 to 37.5 MME over 5 years, IRLS 13.3 to 13.1) (2026-winkelman-opioid-registry-5year.md, PMID 41543123, grade B). Use only with section 1's combination-safety monitoring and per appropriate-use consensus (2018-silber-opioid-appropriate-use-consensus.md, PMID 29304922, grade D). Anticipate a second flare if the opioid bridge is later tapered — 50.8% of opioid-detoxifying patients meet RLS criteria (2016-mackie-detox-alcohol-opioid-rls-prevalence.md, PMID 28017182, grade B).
  • Re-time before you re-dose (grade D, untested). Match each drug's time-to-peak to the recorded symptom-onset hour, not a fixed pre-bedtime rule (2024-garcia-malo-circadian-dosing-posology.md, PMID 39698169, grade D).

Taper rate (grade D — the weakest link here). No RCT has compared taper rates, and no numeric taper schedule is given by any guideline in this knowledge base. The IRLSSG/EURLSSG augmentation guideline and its narrative successor describe qualitative moves only — divide the dose, advance its timing, or switch drug class, once iron is optimised and aggravators removed (2016-garcia-borreguero-augmentation-prevention-treatment-guidelines.md, PMID 27448465, grade D; 2018-garcia-borreguero-treatment-failure-management.md, PMID 29602660, grade D) — and the Mayo algorithms state no mg/day rate or duration at all (2021-silber-mayo-clinic-rls-algorithm.md, PMID 34218864, grade D; 2026-silber-updated-rls-management-algorithm.md, PMID 42203073, grade D). A "10–25% every 2–4 weeks" rule of thumb circulates in clinical practice, but it could not be traced to any source in this knowledge base and is therefore not asserted here: the decrement size and the interval between steps are set by the prescriber, each step small enough to be tolerated and held until symptoms resettle before the next. A structured one-month protocol (rapid taper plus iron plus alpha-2-delta up-titration) achieved complete discontinuation in 77% (24/31), IRLS 27.1 to 15.1 (p<0.001), with failure predicted by longer prior augmentation (11.4 versus 3.5 years, p=0.003) (2026-martin-garcia-structured-dopamine-agonist-withdrawal.md, PMID 42673880, grade C) — one centre, uncontrolled, one month.

Withdrawal timeline (grade C/D). Rebound peaks within days to about two weeks of each reduction and subsides over 2–8 weeks in most people. DAWS — anxiety, panic, dysphoria, suicidal ideation, orthostatic hypotension, craving — is variable, can last months, and has no proven treatment (PMID 23686524, grade D). RLS-specific DAWS incidence is not established; the quoted "up to 24%" comes from Parkinson disease cohorts, where dopamine agonists are generally used at higher doses than in RLS — a comparison offered as editorial context in the cited review, not a separately verified dose figure (PMID 41870480, grade D).

What NOT to do (these are also the commonest real-world errors). Never stop abruptly; never increase the agonist to relieve worsening; never reduce it before the replacement works; never add a second dopaminergic agent; never co-prescribe a gabapentinoid with an opioid without documenting respiratory risk and counselling; never start a taper during acute psychiatric crisis or an unaddressed ICD; never add melatonin for taper insomnia; never judge iron at 4 weeks.

Stop / escalate rules — contact the specialist immediately for: new or worsening suicidal ideation; panic or dysphoria disproportionate to leg symptoms (DAWS); orthostatic syncope; new ICD; a fall; daytime sleep attacks or a driving incident; IRLS worsening more than 10 points sustained beyond 2 weeks; or excess sedation or respiratory depression on a gabapentinoid-opioid combination.

Success. Agonist discontinued, or dose reduced by at least 50%, with IRLS no worse than baseline at 3 months (real-world benchmark: 78% responders at mean 28 months, more likely to have fully discontinued, 59% versus 40%; PMID 35532181, grade C). If it fails: hold the dose, re-check iron and aggravators, lengthen the interval between steps, consider a rotigotine bridge or physician-initiated low-dose opioid. Expect a blunted response to whatever comes next — prior long-term agonist exposure reduces response to gabapentin enacarbil even pre-augmentation (2018-garcia-borreguero-reduced-response-dopaminergic-pretreatment.md, PMID 30772697, grade B) and to dipyridamole and suvorexant (2025-garcia-borreguero-subclinical-augmentation-posthoc.md, PMID 40467956, grade B). The 2026 algorithm broadens fallbacks rather than returning to a dopamine agonist (2026-silber-updated-rls-management-algorithm.md, PMID 42203073, grade D).

Phase 3 — Evening non-drug toolkit. Grade A–C.

Run continuously from Phase 0, not after the taper, concentrated in the vulnerability window: symptoms and PLMs peak around midnight to 01:00 and trough around 09:00–11:00 (2004-michaud-circadian-rhythm-melatonin.md, PMID 14991815, grade C). Content and grading in section 4. Success: one modality used consistently for 4+ weeks with a diary-documented fall in evening severity. If it fails: substitute another modality rather than escalating drug dose. These are adjuncts, not substitutes.

Phase 4 — Maintenance, relapse plan, and when medication remains right. Grade C/D.

Quarterly review for one year, then 6-monthly: IRLS, diary onset time, ferritin/TSAT, renal function, mood, ICD, falls and driving. Repeat iron repletion as needed — brain iron handling appears to be a persistent trait, not a one-time correction (2011-catoire-meis1-iron-homeostasis.md, PMID 21710629, grade E). Relapse plan: (1) re-check ferritin and TSAT; (2) look for a new aggravating drug or OTC melatonin; (3) screen for augmentation against the recorded baseline onset time; (4) re-time the dose to the actual symptom-onset hour (PMID 39698169, grade D); (5) only then discuss a dose or agent change — never back to a dopamine agonist as the first move (PMID 42203073, grade D).

Medication remains the right choice when symptoms stay moderate-to-severe despite corrected iron, removed aggravators and consistent non-drug therapy, or when sleep and quality of life are materially impaired. An alpha-2-delta ligand that works does not need tapering to prevent augmentation — no reviewed trial links the class to augmentation — but monitor dose creep, renal accumulation and misuse risk in substance-use history (2017-evoy-gabapentinoid-abuse-misuse-review.md, PMID 28144823, grade C; 2021-tambon-gabapentinoid-addictovigilance-france.md, PMID 33613345, grade C). "Off all drugs" is not a success criterion; reducing medication is the goal only where it is safe.

Adherence and access. Tie iron to an existing morning habit; plan for constipation; treat the evening routine as a fixed appointment; involve a partner, since family members carry measurable burden and benefit from inclusion (2026-odzakovic-family-members-everyday-experiences.md, PMID 41969743, grade C). German statutory insurers generally cover strength training, heat/cold and electrotherapy; other modalities are self-pay at about €50 to €3,660 per 6 months (2023-iqwig-nichtmedikamentoese-verfahren-rls.md, HTA report, no PMID, grade C). IV iron in iron-replete RLS is off-label in Germany, gabapentin enacarbil is not EMA-authorised (PMID 31229171, grade D), and EU reimbursement for TOMAC and any Luxembourg RLS pathway were searched and not found.


4. Non-drug toolkit, ranked by evidence

  1. Tonic motor activation (TOMAC; bilateral high-frequency peroneal nerve stimulation) — grade A. The only non-pharmacological therapy recommended by AASM 2025. Pivotal sham RCT (n=133): CGI-I responder 45% versus 16%, difference 28% (95% CI 14–43), p=0.00011 — NNT about 4 (2023-bogan-tomac-restful-rct.md, PMID 37458698, grade B; manufacturer-sponsored); replicated in naive and refractory patients (2024-singh-tomac-naive-refractory-rct.md, PMID 39173210, grade B) and in two meta-analyses (IPD, 3 RCTs, n=252, IRLS mean difference 3.39 adjunctive / 3.80 monotherapy, p at most 0.005: 2026-karroum-tomac-individual-participant-meta-analysis.md, PMID 41581285, grade A; 2025-mohamed-tomac-meta-analysis.md, PMID 40381601, grade A). It enabled at least 20% opioid dose reduction in 70% of refractory patients (2023-buchfuhrer-tomac-opioid-dose-reduction.md, PMID 37990163, grade C). Honest caveat: that pooled effect sits below the 6-point individual MCID, and the source studies are largely manufacturer-linked.
  2. Structured exercise (aerobic plus resistance, 3x/week, at least 12 weeks) — grade A/B. Founding general-population RCT p=0.001 (2006-aukerman-exercise-rct.md, PMID 16951298, grade B); in haemodialysis RLS progressive cycling cut severity 58% versus 17% with sham cycling, p=0.003 (2013-giannaki-aerobic-exercise-uremic-rct.md, PMID 23929523, grade B), and exercise roughly matched ropinirole 0.25 mg (46% versus 54%) while adding lean mass (2013-giannaki-exercise-dopamine-agonist-comparison.md, PMID 24024727, grade B). Caveat: the largest network meta-analysis found intradialytic aerobic exercise alone non-significant (2022-chen-component-network-meta-analysis-eskd.md, PMID 34612498, grade A).

The remainder, in the same rank order:

#ModalityBest evidenceGrade
3Temperature therapy (heat/cold) — cheapest globally accessible option6 trials, 177 patients, pooled SMD −1.520 (95% CI −2.122 to −0.918) (2023-mohammadi-temperature-therapy-meta-analysis.md, PMID 38296642); only whole-body cryotherapy at −60 °C reached significance in a dedicated RCT (2016-happe-cryotherapy-rct.md, PMID 27772790). Caveat: that pooled effect is driven substantially by haemodialysis-associated RLS (subgroup beta −2.006, p<0.05) and the source trials are predominantly secondary-RLS populations, so the magnitude of benefit in idiopathic RLS is much less certain — cheap and safe to try, but do not promise this effect sizeA (haemodialysis RLS) / C (idiopathic)
4Pneumatic compression — contestedSham RCT (n=35): severity 14.1 to 8.4, p=0.006, complete relief in a third of active users and none on sham (2008-lettieri-pneumatic-compression-sham-rct.md, PMID 19017878); "likely efficacious" per MDS 2018 (2018-winkelmann-mds-evidence-based-review.md, PMID 29756335) but not recommended by the German 2024 guideline (PMID 39501372)B / D
5CBT-I and progressive muscle relaxationFour CBT-I sessions beat one sleep-hygiene session on insomnia severity and sleep efficiency to 3 months, n=25 (2020-song-cbti-rls-comorbid-insomnia-rct.md, PMID 32861015); PMR replicated in haemodialysis, n=117, p=0.001 (2025-elfahl-progressive-muscle-relaxation-hemodialysis-rct.md, PMID 40954628) and pregnancy (2023-akbas-progressive-muscle-relaxation-pregnancy-rct.md, PMID 37146530). Safety caveat: CBT-I's sleep-restriction component has never been tested in RLS and experimental sleep loss worsens PLMs and sensory symptoms (1999-trenkwalder-circadian-sleep-deprivation-plm.md, PMID 9918351, grade C) — raise the time-in-bed floor, supervisedB
6Modifiable lifestyle and trigger avoidancen=55,540: adjusted OR 0.67 (95% CI 0.47–0.97) for 4 versus 0 healthy factors, p for trend <0.001 (2016-batool-anwar-lifestyle-factors-cohort.md, PMID 26446243). Caffeine restriction and generic "healthy eating" were not associated with RLS risk in that cohort. Avoid prolonged evening immobility: leg pO2 5.54 versus 7.19 kPa, p<0.01, correlating with severity rho 0.692 (2014-salminen-peripheral-hypoxia-rls.md, PMID 24789861)B
7Yoga; massage/reflexology; near-infrared light; spinal tsDCSOne or two small positive trials each, none replicated at scale (PMIDs 31957638, 35686176, 29776916, 41616577)B / C

Optional adjuncts: magnesium with B6, significant only at 2 months (2022-jadidi-magnesium-b6-rct.md, PMID 36587225, grade B) but contradicted by an inconclusive review (2019-marshall-magnesium-systematic-review.md, PMID 31678660, grade C); vitamin C/E in dialysis RLS only (2012-sagheb-vitamin-c-e-hemodialysis-rct.md, PMID 22317944, grade B); bright light timed away from the symptomatic window (PMID 20226733, grade C, n=8, unblinded).

Not supported: acupuncture (2008-cui-cochrane-acupuncture-review.md, PMID 18843716, grade A), cortical tDCS (2014-koo-tdcs-negative-rct.md, PMID 25576136, grade B), vitamin D (PMID 30430372, grade B), cannabidiol, benzodiazepines. No RLS-specific evidence found at all — distinct from "weak evidence" — for weighted blankets, mindfulness/ACT and evening-exercise timing.

On the horizon. Dipyridamole (adenosine augmentation) beat placebo in a crossover RCT, IRLS −13 versus −5.0, p<0.001 (2021-garcia-borreguero-dipyridamole-rct.md, PMID 34137476, grade B), with a conditional AASM 2025 recommendation; suvorexant likewise (2024-garcia-borreguero-suvorexant-rct.md, PMID 38246901, grade B). Single trials only.


5. Monitoring plan

WhatWhenThreshold / action
IRLS (0–40)Baseline, weeks 6, 12, 24, then 6-monthlyA change under about 6 points is inside the placebo band (PMID 24051118, C; PMID 28490647, A)
Diary with onset clock timeDaily in Phases 0–2, then one week per monthA 4-hour advance in onset time is the primary augmentation criterion (PMID 17544323, D)
Ferritin + TSATBaseline, 8–12 weeks after any iron course, then 6–12-monthlyRepeat repletion at 75 ng/mL or below, or TSAT under 20% (PMID 39324694, D)
Serum phosphateBefore, and about 2 and 5 weeks after, every FCM courseStop-rule below 2.0 mg/dL (PMID 30518682, B)
Augmentation signsEvery visitEarlier onset, spread to arms/trunk, shorter rest latency, shorter drug effect, worsening after a dose rise; quantify with the ASRS (augmented 7.4 versus 2.0, p<0.0001) (2007-garcia-borreguero-augmentation-severity-rating-scale.md, PMID 17543579, B)
ICD, mood/suicide risk, falls, drivingBaseline, every visit, every dose change, throughout any taperICD 7.1% (mixed agonist/levodopa cohort) and 10% (8-year agonist cohort), about 10% sleep attacks while driving (PMID 21955669 / 23036265, C); suicide/self-harm aHR 2.66 (PMID 31441941, B)
PLMS / actigraphy — optionalOnly if sleep fragmentation dominatesThe placebo effect is smallest for objective measures (2008-fulda-wetter-placebo-response-meta-analysis.md, PMID 17932100, A); use a validated ankle actigraph (SOMNOwatch r=0.98 versus PSG) over multiple nights (2024-spektor-somnowatch-actigraphy-plmi-validation.md, PMID 39449312, C). Consumer wrist wearables do not measure PLMS (2017-de-zambotti-fitbit-charge2-validation.md, PMID 29235907, C)

6. What this protocol does not claim; limitations; where the evidence is thin

No cure is claimed. The most claimed here is that a subset can reduce or eliminate medication need: 37.5% medication-free at 30 weeks in one FCM trial (2016-cho-ferric-carboxymaltose-rct-remitters.md, PMID 27823710, grade B) and 77–78% of augmented patients moved off a dopamine agonist in specialist hands, on one month to about two years of follow-up (PMID 42673880, grade C; PMID 35532181, grade C). Most patients are not full remitters.

No attribution is claimed. In a 10-year Finnish cohort, prevalence fell from 21% to 15% and roughly half of those with frequent RLS at baseline reported rare or no symptoms a decade later with no intervention (2011-juuti-ten-year-natural-course-remission.md, PMID 21725862, grade B). Read any uncontrolled improvement against that base rate plus the −6.58-point placebo response.

The four thinnest points.

  1. Taper rate is expert opinion only D, and DAWS in RLS is unquantified. No RCT has compared taper schedules, or add-then-taper with taper-then-add; no guideline in this knowledge base supplies a numeric taper rate at all — the widely repeated "10–25% every 2–4 weeks" figure was untraceable to any source here and has been removed from this document — and the best structured-protocol result (77%) is a single-centre, uncontrolled, 31-patient, one-month cohort. No RLS-specific DAWS incidence figure exists, and no DAWS treatment is proven.
  2. Brain iron deficiency, the mechanistic foundation, is contested in vivo. The largest MRI study plus meta-analysis (72 RLS versus 72 controls) found no pooled evidence of lower brain iron, with funnel-plot asymmetry suggesting publication bias (2022-beliveau-mri-metaanalysis-revisited.md, PMID 35500370, grade A), and serum ferritin is an imperfect brain-iron proxy.
  3. Most non-drug effect sizes sit inside the placebo band, and most come from the wrong population. TOMAC's pooled IRLS effect (3.39–3.80) is below the individual MCID; the best exercise, vitamin C/E, massage and PMR trials were done in haemodialysis or pregnancy, and a Cochrane review of CKD-RLS interventions concluded uncertain effects (2016-gopaluni-cochrane-ckd-rls-interventions.md, PMID 27819409, grade A). No RCT has tested a bundled non-drug package with medication reduction as its primary endpoint, and no device trial has used tapering or augmentation prevention as a prespecified endpoint. Device evidence is also short and largely manufacturer-linked.
  4. Long-term safety of this protocol's own components is under-evidenced, and guidelines disagree. Repeated multi-year FCM dosing in RLS has no safety cohort; the gabapentinoid-opioid bridge carries the one clear regulatory respiratory warning; gabapentin enacarbil is unavailable in the EU; no prospective 5–10-year de-prescribing cohort exists. AASM 2025 recommends against dopamine agonists first-line while German 2024 and French guidance remain more permissive (2018-chenini-french-consensus-pharmacoresistant.md, PMID 30075957, grade D), and mortality data are contradictory and must not be used selectively to manufacture urgency (2017-katsanos-meta-analysis-cardiovascular-mortality.md, PMID 28948600, grade A).

Simulation outputs are hypotheses, never evidence. Anything from simulation/ or docs/SIMULATION_SPEC.md is category (iii) of the project honesty rules: a lumped ODE caricature with no patient-level calibration, licensed only for ordinal and timescale claims (docs/SIMULATION_SPEC.md §1.2–1.3). No clinical statement here rests on it.


7. References

Guidelines (knowledge-base/guidelines/): 2025-winkelman-aasm-guideline-rls.md (39324694) · 2024-trenkwalder-german-dgn-dgsm-guideline.md (39501372) · 2022-heidbreder-dgn-s2k-leitlinie-rls.md (AWMF 030/081) · 2018-allen-irlssg-iron-treatment-guidelines.md (29425576) · 2016-garcia-borreguero-augmentation-prevention-treatment-guidelines.md (27448465) · 2014-allen-irlssg-updated-diagnostic-criteria.md (25023924) · 2007-garcia-borreguero-max-planck-augmentation-criteria.md (17544323) · 2026-silber-updated-rls-management-algorithm.md (42203073) · 2021-silber-mayo-clinic-rls-algorithm.md (34218864) · 2014-picchietti-irlssg-pregnancy-lactation-guideline.md (25553600) · 2018-chenini-french-consensus-pharmacoresistant.md (30075957).

Iron: 2019-trotti-cochrane-iron-review.md (30609006) · 2024-qadri-iv-fcm-meta-analysis.md (39326219) · 2019-avni-iron-supplementation-meta-analysis.md (30798983) · 2024-earley-iv-ferric-carboxymaltose-rct.md (38625730) · 2016-cho-ferric-carboxymaltose-rct-remitters.md (27823710) · 2011-allen-fcm-multicenter-preliminary-rct.md (21978726) · 2018-cho-fcm-500mg-dose-rct.md (29458749) · 2017-trenkwalder-fcm-nonanemic-rct.md (28643901) · 2009-wang-oral-ferrous-sulfate-rct.md (19230757) · 2013-lee-oral-iron-vs-pramipexole-rct.md (24267148) · 2017-stoffel-alternate-day-iron-absorption.md (29032957) · 2018-wolf-ferric-carboxymaltose-hypophosphatemia-rct.md (30518682) · 2025-magagnoli-fcm-hypophosphatemia-systematic-review.md (39935027) · 2018-dauvilliers-hepcidin-severity-rls.md (29418021) · 2021-dowsett-hepcidin-danish-blood-donors-negative.md (34736065) · 2022-beliveau-mri-metaanalysis-revisited.md (35500370) · 2011-catoire-meis1-iron-homeostasis.md (21710629) · 2021-earley-ppi-h2-antagonist-rls-denmark-us.md (33119070).

Augmentation, tapering, DAWS: 2013-nirenberg-dopamine-agonist-withdrawal-review.md (23686524) · 2026-lobinska-dopamine-agonist-withdrawal-syndrome-review.md (41870480) · 2018-garcia-borreguero-treatment-failure-management.md (29602660) · 2026-martin-garcia-structured-dopamine-agonist-withdrawal.md (42673880) · 2025-martin-garcia-spanish-augmentation-cohort.md (41130145) · 2022-yeung-laiwah-augmentation-guideline-effectiveness.md (35532181) · 2022-winkelman-treating-refractory-augmented-rls.md (35609673) · 2016-winkelman-rotigotine-cross-titration-method.md (27810181) · 2014-bogan-rotigotine-bench-to-bedside-review.md (24636821) · 2012-lipford-pramipexole-eight-year-cohort.md (23036265) · 2011-voon-impulse-control-behaviours-cohort.md (21955669) · 2011-hogl-pramipexole-6-month-augmentation-rct.md (21354368) · 2018-garcia-borreguero-reduced-response-dopaminergic-pretreatment.md (30772697) · 2025-garcia-borreguero-subclinical-augmentation-posthoc.md (40467956) · 2026-winkelman-rls-jama-clinical-review.md (41563785) · 2025-ferri-world-sleep-society-endorsement.md (40834711).

Alpha-2-delta ligands, opioids, combination safety: 2014-allen-nejm-pregabalin-pramipexole-augmentation.md (24521108) · 2011-lee-gabapentin-enacarbil-pivot-rls-ii.md (21677899) · 2015-ondo-gabapentin-enacarbil-prior-agonist-exposure.md (26788345) · 2010-bogan-gabapentin-enacarbil-long-term-maintenance.md (20511481) · 2025-sunwoo-pregabalin-korean-negative-rct.md (40635537) · 2019-wanner-nondopaminergic-vs-dopaminergic-europe-access.md (31229171) · 2013-trenkwalder-oxycodone-naloxone-rct.md (24140442) · 2026-winkelman-opioid-registry-5year.md (41543123) · 2023-winkelman-national-rls-opioid-registry.md (36697248) · 2018-silber-opioid-appropriate-use-consensus.md (29304922) · 2016-mackie-detox-alcohol-opioid-rls-prevalence.md (28017182) · 2022-hahn-gabapentinoid-opioid-combination-meta-analysis.md (36304170) · 2021-shah-raji-opioid-gabapentinoid-falls-medicare.md (35476819) · 2020-shrestha-palaian-fda-gabapentinoid-respiratory-warning.md (33728039) · 2017-evoy-gabapentinoid-abuse-misuse-review.md (28144823) · 2021-tambon-gabapentinoid-addictovigilance-france.md (33613345) · 2022-lehmann-gabapentin-overdose-dialysis-case-report.md (36518357).

Aggravating drugs and circadian: 2026-fugger-secondary-rls-pharmacovigilance.md (42251748) · 2008-rottach-second-generation-antidepressants-rls.md (18468624) · 2017-kolla-antidepressants-rls-systematic-review.md (28822709) · 2020-revet-antidepressants-movement-disorders-vigibase.md (32546134) · 2013-ostojic-metoclopramide-systemic-sclerosis-rls.md (23456369) · 2010-whittom-melatonin-bright-light-rls.md (20226733) · 2025-wu-ramelteon-faers-rls-signal.md (40624487) · 2024-buntak-melatonin-discontinuation-case.md (no PMID) · 2004-michaud-circadian-rhythm-melatonin.md (14991815) · 2024-garcia-malo-circadian-dosing-posology.md (39698169) · 1999-trenkwalder-circadian-sleep-deprivation-plm.md (9918351).

Non-drug toolkit: 2023-bogan-tomac-restful-rct.md (37458698) · 2026-karroum-tomac-individual-participant-meta-analysis.md (41581285) · 2025-mohamed-tomac-meta-analysis.md (40381601) · 2024-singh-tomac-naive-refractory-rct.md (39173210) · 2023-buchfuhrer-tomac-opioid-dose-reduction.md (37990163) · 2008-lettieri-pneumatic-compression-sham-rct.md (19017878) · 2023-mohammadi-temperature-therapy-meta-analysis.md (38296642) · 2016-happe-cryotherapy-rct.md (27772790) · 2006-aukerman-exercise-rct.md (16951298) · 2013-giannaki-aerobic-exercise-uremic-rct.md (23929523) · 2013-giannaki-exercise-dopamine-agonist-comparison.md (24024727) · 2022-chen-component-network-meta-analysis-eskd.md (34612498) · 2020-innes-yoga-educational-film-rct.md (31957638) · 2020-song-cbti-rls-comorbid-insomnia-rct.md (32861015) · 2025-elfahl-progressive-muscle-relaxation-hemodialysis-rct.md (40954628) · 2023-akbas-progressive-muscle-relaxation-pregnancy-rct.md (37146530) · 2016-batool-anwar-lifestyle-factors-cohort.md (26446243) · 2022-ghanbari-reflexology-swedish-massage-rct.md (35686176) · 2018-mohammadi-near-infrared-light-therapy-rct.md (29776916) · 2026-wang-tsdcs-spinal-cortical-rct.md (41616577) · 2022-jadidi-magnesium-b6-rct.md (36587225) · 2019-marshall-magnesium-systematic-review.md (31678660) · 2012-sagheb-vitamin-c-e-hemodialysis-rct.md (22317944) · 2018-wali-vitamin-d-replacement-rct.md (30430372) · 2024-gonzalez-parejo-dietary-supplements-systematic-review.md (39064758) · 2008-cui-cochrane-acupuncture-review.md (18843716) · 2014-koo-tdcs-negative-rct.md (25576136) · 2018-winkelmann-mds-evidence-based-review.md (29756335) · 2023-iqwig-nichtmedikamentoese-verfahren-rls.md (HTA report, no PMID) · 2021-garcia-borreguero-dipyridamole-rct.md (34137476) · 2024-garcia-borreguero-suvorexant-rct.md (38246901) · 2014-salminen-peripheral-hypoxia-rls.md (24789861).

Outcomes, comorbidity, natural history: 2013-allen-irls-minimal-clinically-significant-change.md (24051118) · 2016-ondo-irls-rls6-mcic-rotigotine.md (26446245) · 2017-silva-placebo-nocebo-meta-analysis.md (28490647) · 2008-fulda-wetter-placebo-response-meta-analysis.md (17932100) · 2007-garcia-borreguero-augmentation-severity-rating-scale.md (17543579) · 2002-michaud-suggested-immobilization-test.md (11835447) · 2024-spektor-somnowatch-actigraphy-plmi-validation.md (39449312) · 2017-de-zambotti-fitbit-charge2-validation.md (29235907) · 2011-juuti-ten-year-natural-course-remission.md (21725862) · 2017-katsanos-meta-analysis-cardiovascular-mortality.md (28948600) · 2019-zhuang-suicide-self-harm-risk-rls.md (31441941) · 2018-chen-pregnancy-prevalence-meta-analysis.md (29169861) · 2023-safarpour-ckd-rls-systematic-review.md (37008995) · 2025-elangovan-ckd-dialysis-iron-prevalence.md (40677478) · 2016-gopaluni-cochrane-ckd-rls-interventions.md (27819409) · 2023-schrag-parkinson-prodromal-risk-factors.md (36342675) · 2018-akin-type2-diabetes-rls-prevalence.md (30213520) · 2006-gemignani-polyneuropathy-rls.md (16685701) · 2015-chen-migraine-rls-diary-study.md (26195587) · 2026-odzakovic-family-members-everyday-experiences.md (41969743).

No note listed in knowledge-base/UNVERIFIED.md is cited anywhere in this document.