Is the RLS Atlas simulation a tool that predicts my personal outcome?

No. It is an explicitly mechanistic teaching model — a hypothesis built from published evidence, not a new fact about any individual. It has no patient-specific calibration data, cannot forecast what will happen to you, and must never be used to start, stop, or change a dose. Every number it produces is labelled "not clinical prediction."

Can restless legs syndrome improve or go away without treatment?

Sometimes. In a ten-year Finnish cohort, roughly half of the people who had frequent RLS at the start reported rare or no symptoms a decade later — with no treatment at all. Because the average placebo response in RLS trials is itself substantial (6.58 points on a 40-point scale), any uncontrolled improvement should be read cautiously.

What is augmentation in restless legs syndrome?

Augmentation is a drug-driven worsening seen with dopamine agonists: symptoms start earlier in the day (a shift of about four hours is the clearest sign), spread to the arms or trunk, appear sooner after sitting or lying down, and the drug wears off faster. Worsening after a dose increase is itself considered diagnostic.

Why has my restless legs medication (pramipexole/Sifrol) stopped working?

The most likely explanation is augmentation — a complication where the drug itself gradually drives symptoms harder over time, not the underlying condition progressing. It is not a sign you need more of the drug. Roughly 7–10% of people on dopamine agonists develop it each year, compounding to about 42% by eight years in one specialist cohort.

Interactive model

The restless legs syndrome simulation: what it predicts, and what it doesn't

Ten scripted scenarios, and a live model you can drive yourself, built from the same mechanistic simulation that produced every number and curve in this project's evidence-graded docs.

This interactive model simulates the mechanism believed to underlie RLS and dopamine-agonist augmentation over time. It is a teaching tool built from published evidence, not a predictor of any individual's outcome.

Every biological claim below carries a PubMed ID (PMID), graded A (strong, repeated trials) to E (mechanism/animal only). Every curve or number attributed to "the model" is our simulation's output — a hypothesis built from that evidence, not a new fact about you. Section 6 says why that distinction matters.

Model scope and limitations

This page describes a mechanistic computer model built for this project (simulation/, spec'd in docs/SIMULATION_SPEC.md). It is not a medical device, not a diagnostic tool, and not individual treatment advice. It cannot forecast what will happen to you. Never start, stop, or change a dose based on it. Dopamine-agonist tapering and augmentation management must happen only under physician supervision, and a dopamine agonist must never be stopped abruptly.

Known calibration limitation (0.25 mg vs. higher doses). The cohort's dose gradient is steeper than the 52-week RCT gradient it is meant to bracket: 0.5 mg gives 20.0% 12-month augmentation in this model, against a real 52-week trial figure of 7.7% at 0.5 mg — roughly a 2.6x over-statement. Only the 0.25 mg figure should be read as calibrated: “the model must not be read as a quantitative predictor of augmentation risk at any dose other than 0.25 mg; even at 0.25 mg, read it as order-of-magnitude only” (docs/SIMULATION_SPEC.md §6.3; also simulation/README.md’s “Known calibration limitation”).

Every number and curve below belongs to this project's honesty-rule category (iii): a hypothesis generated by our simulation, never a substitute for the evidence base, and never grounds to choose, change, or stop a medication on your own.

Ten scripted scenarios

Walk through S1-S10

Each scenario below is a precomputed run of the model (docs/SIMULATION_SPEC.md section 5), illustrating one specific, audited claim — from the disease's nighttime pattern to why abrupt withdrawal is dangerous.

Why does a dopamine agonist help fast and hurt slowly? (the core story)

Loading S3.json…

Build your own

Run the model yourself

Pick a baseline, a dose, and an exit strategy. This runs the real model, live, in a Web Worker in your browser — nothing is sent to a server. Uses a typical (population-median) patient, not the high-susceptibility demo patient behind the S3/S4/S6/S7/S8 scenarios above.

Baseline ferritin / brain iron45 µg/L — below the AASM 2024 IV-iron threshold (75 µg/L)

Pramipexole dose

Months on the drug12 months

Intervention (starts after the months above)

IV iron (1000 mg, two infusions 5 days apart)

Gabapentin bridge (300 -> 600 -> 900 mg titration)Evening device / exercise input (17:00-03:00)Taper length12 weeks

A 0-week taper (abrupt stop) is shown only to illustrate why abrupt cessation is dangerous (see the withdrawal-load spike this produces). In real life, a dopamine agonist must never be stopped abruptly and any taper must happen only under physician supervision.

What the model says

The way out

From simulation/EXPLAINER.md section 5 — what the model's best-outcome scenario (S6/S8) actually shows, and why each piece alone is not enough.

Three things happen together in the model's best-outcome scenario, and none of them work alone. First, iron repletion — IV iron raises ferritin for months, and brain iron follows with a longer lag still: in the model, brain iron doesn't plateau until roughly 15 weeks, and the maximum symptom benefit isn't reached until around 22 weeks. That's slower than it might sound, but durable: IV ferric carboxymaltose has produced real, measurable symptom improvement by 6 weeks and 60% response rates, with some patients medication-free 30 weeks later (Earley et al. 2024, PMID 38625730, grade B; Cho et al. 2016, PMID 27823710, grade B; Kim et al. 2023, PMID 37257418).

Second, a slow, physician-supervised taper with a gabapentin bridge already running — the model's simulated exit path reaches roughly a 5-point improvement on its 0–10 symptom scale by 6 months, compared with staying near the ceiling on the abrupt-stop path. Third, evening non-drug inputs stacked on top of that foundation add a further, real, if modest, benefit — a device therapy (TOMAC) reduced IRLS scores by 3.4–3.8 points against a sham control across three randomized trials (Karroum et al. 2026, PMID 41581285, grade A), and reduced sleep-time limb movements by nearly half (Rigot et al. 2026, PMID 42488517). In the model, adding that device on top of the iron-plus-taper path shaves off roughly another point of symptom severity in the same direction the real trials show.

Keep reading

The restless legs syndrome simulation: what it predicts, and what it doesn't

Model scope and limitations

Walk through S1-S10

Run the model yourself

The way out

Related pages and source documents

Related questions