Refractory Fibromyalgia: Low-Dose Naltrexone or the Third Approved Agent First?
She has already moved through two of the three FDA-approved fibromyalgia drugs without lasting relief. The question is whether an unapproved, glial-modulating agent with a thin trial base deserves to go next, ahead of the one approved option still untried.
Priya K., a 51-year-old woman, has spent the last two summers teaching her granddaughter to knit, an afternoon ritual that's gotten harder to finish as her hands and forearms tire faster than they used to. She was diagnosed with fibromyalgia three years ago and has been under this clinic's care for the eighteen months since her symptoms failed to settle with exercise and NSAIDs alone. Pregabalin, tried first, was stopped after six weeks for dizziness severe enough that she fell once getting up from her desk. Duloxetine, tried next and continued for five months at the target dose, brought her FIQR down from 68 to 58 — a real but partial response she describes as "enough to notice, not enough to matter" — and she still leaves work early at least once a week when the pain becomes unmanageable. She has read about low-dose naltrexone online, mostly in fibromyalgia support-group forums, and brought it up herself at today's visit before anyone on the team did.
The pharmacology behind that request is real, even if the evidence supporting it is thinner than for anything already tried. At the doses used for fibromyalgia — a small fraction of naltrexone's usual opioid-blockade dose — the drug is thought to act primarily on glial cells rather than neurons, antagonizing toll-like receptor 4 on microglia and quieting the neuroinflammatory signaling increasingly implicated in central sensitization, with a secondary, paradoxical opioid-receptor upregulation thought to follow the brief nightly blockade. Younger and Mackey's original 2009 crossover pilot, and the larger randomized crossover that followed from the same group in 2013, both found meaningful pain reduction against placebo in fibromyalgia specifically — real, peer-reviewed findings, not just forum enthusiasm — though both were small, single-center crossover studies. The confirmatory work has since arrived, and it did not confirm: Due Bruun and colleagues' FINAL trial, published in Lancet Rheumatology in 2024 and the largest randomized test of the drug in this condition, compared naltrexone 6mg against placebo in women with fibromyalgia over twelve weeks and found no superiority on its primary pain endpoint. Pooled analyses of the small randomized trials still favor LDN overall, so the honest summary is a positive small-trial literature and a negative large one — not a drug that works, and not a drug that doesn't. The genuinely open question isn't whether that early evidence is real; it's whether a drug whose best-powered trial missed should go ahead of milnacipran, the one FDA-approved agent she hasn't tried, because she asked for it by name.
Follow-up visit, eighteen months in
I'd try milnacipran before LDN. She's tried two mechanisms, not three, and duloxetine at five months with a real drop in FIQR is a partial response, not a failure of the whole drug class — milnacipran's own norepinephrine-weighted profile is different enough from duloxetine's that a partial responder to one doesn't tell us much about the other. LDN's evidence, meanwhile, is two small single-center crossover trials — and since those, the FINAL trial, the biggest randomized test anyone has run in fibromyalgia, which found 6mg no better than placebo on pain at twelve weeks. That is the trial I'd want her to hear about before she takes anything.
I take the evidence-tier point — I'm not claiming LDN and milnacipran are on equal footing, and I wouldn't reach for LDN in a treatment-naive patient ahead of any approved agent.
But Younger and Mackey's 2009 pilot and their larger 2013 crossover weren't anecdote — both were randomized, placebo-controlled, and found real pain reduction specifically in fibromyalgia, at doses low enough that the mechanism is thought to run through microglial toll-like receptor 4 antagonism rather than classic opioid blockade. And I won't pretend FINAL didn't happen — but a single negative trial doesn't erase a literature, and the pooled analyses across all the randomized data still come out favoring the drug. What we have is a genuinely mixed record, which is a different thing from a refuted one. She's also not a naive patient asking on a whim; she's someone who's already tolerated one intolerable drug and one partially-working one, and is asking for this specific option by name, informed about what it is.
I don't think this needs to be milnacipran or LDN. Her duloxetine response, while incomplete, is real, and there's no pharmacologic reason LDN can't be added to it rather than substituted for it — the mechanisms don't overlap in a way that predicts an interaction, and naltrexone at these microgram-per-kilogram doses carries a side-effect profile in its own small trials that's been closer to placebo than to naltrexone's usual opioid-blockade dose.
That reframes the actual decision: not which drug deserves the next open slot, but whether it's reasonable to layer a thin-evidence, low-risk option onto an existing partial response while holding milnacipran in reserve if the combination doesn't move things further.
Low-dose naltrexone added to her existing duloxetine at 3mg nightly, titrated toward 4.5mg as tolerated, with FIQR reassessed at twelve weeks rather than the usual eight, matching the window FINAL used — so that if this doesn't work, it will have failed on the same timetable as the trial that failed it.
Not agreed, and left explicitly open rather than smoothed over: whether a true non-response to the combination at twelve weeks should mean stopping LDN and trying milnacipran next, or stopping duloxetine and trying milnacipran plus LDN together. The rheumatologist favors returning to a single, better-evidenced agent before adding complexity; the pain medicine specialist would rather not discard a drug that's shown even a partial effect without first testing whether it works better paired with something else. Neither position was resolved before today's visit ended, and the plan for that branch point was deliberately left for the twelve-week visit rather than decided in advance.