Naltrexone blocks opioid receptors. At about 50 mg a day it's a treatment for alcohol and opioid dependence. Low-dose naltrexone (LDN) is the same drug at about a tenth of that or less, most often 4.5 mg a day, prescribed off-label for fibromyalgia, ME/CFS, long COVID and some inflammatory conditions.
This page is about why anyone expects a small dose to do something a large one doesn't, and how the idea has held up in trials. It's general background. Whether it suits a particular person is a question for their prescriber.
Why a small dose might act differently
The body clears naltrexone quickly. Its half-life is about 4 hours, and about 12 for its main breakdown product. A full dose is still enough to keep opioid receptors blocked all day. The thinking behind LDN is that a small dose taken once a day blocks them only partly and for some hours, and that the useful effects come from what happens afterwards, or from other targets altogether.
Three explanations have been put forward. More than one could be true.
Idea 1: a rebound in the body's own opioids
This is the oldest. Block the receptors for a few hours and the body answers by making more of its own opioids, the endorphins and enkephalins, and more receptors for them. When the drug wears off, opioid signalling is stronger than before for the rest of the day. Endorphins dull pain, and immune cells carry opioid receptors too, so the effect could reach beyond pain.
What supports it. Mostly animal work. In one mouse experiment, a dose low enough to block the receptors for only 6 to 8 hours a day raised an enkephalin in the blood 2.5 times. In people the evidence is thin. In 40 people with multiple sclerosis who took LDN for six months, beta-endorphin rose, but the study had no placebo group to compare with.
Idea 2: calming the brain's immune cells
Microglia are the brain's resident macrophages (see The brain). When they're activated, the signals they release can cause pain, fatigue and the general feeling of being ill. One of the sensors that activates them is TLR4 (see How it spots trouble).
In 2008 a research group reported that naloxone, a close relative of naltrexone, and mirror-image versions of both drugs blocked TLR4 in cell lines. Given into the fluid around the spinal cord, they also reversed nerve-injury pain in rats. The mirror-image versions can't act on opioid receptors, so this wasn't an opioid effect. Ordinary naltrexone wasn't among the drugs given to the rats. From there came the proposal that LDN is an anti-inflammatory for the brain.
What supports it. Cells and rodents, and it's disputed. Three other laboratories tried to repeat the TLR4 block in cells and couldn't. The original group replied that the results in animals are what count. In human blood cells, naltrexone didn't change the response to a TLR4 trigger. It did damp the response through three other sensors, TLR7, 8 and 9, at 200 micromolar, a strength no tablet comes near (see idea 3). In people, eight women with fibromyalgia had lower blood levels of many cytokines, anti-inflammatory ones among them, after eight weeks of LDN, in a study with no separate placebo group.
Idea 3: a calcium channel in natural killer cells
This is the newest, and the one aimed at ME/CFS and long COVID. It comes from one research group, at Griffith University in Queensland, and it runs in four steps.
- To kill, a natural killer cell needs calcium to flow in.
- The group reports that one calcium channel, TRPM3, works poorly in NK cells from people with ME/CFS. It later reported the same in long COVID.
- In pain nerves, switching on an opioid receptor shuts TRPM3 down. Other researchers showed that in mice in 2017.
- So blocking the opioid receptor with naltrexone might let the channel work again.
What supports it. NK cells from 8 people with ME/CFS, kept in naltrexone for 24 hours, recovered their TRPM3 activity. In a later study of 10 patients, calcium flowed in normally again after the same treatment. In NK cells taken from 9 people with ME/CFS who were already on LDN, the channel behaved as it did in healthy volunteers, and the same was found in 9 people with long COVID on LDN.
What to hold against it.
- Every one of the naltrexone studies is small, and all are from the same group. Its largest study of the channel itself, 36 patients and 42 healthy people tested in two laboratories, found the same fault, but no outside group has repeated it.
- The naltrexone in the dish was at 200 micromolar, thousands of times stronger than the level a full 50 mg tablet reaches in blood. At that strength it acts on more than opioid receptors (it's the strength that damped TLR7, 8 and 9 under idea 2), so nothing shows the receptor is the route.
- It took 24 hours. Added to the cells for only the minutes of a recording, naltrexone didn't open the channel, and it changed nothing in healthy people's cells.
- In the mouse nerves the brake came on only when an opioid drug was added. A blocker on its own changed nothing, so there was no standing brake to lift. Whether human NK cells carry opioid receptors at all is disputed.
- A low dose is thought to block the receptor only partly and for some hours a day, so on this idea the brake would be off part of the time at most.
- The studies of people on LDN compared different people at one moment. They didn't follow anyone from before treatment to after.
- They measured a channel, not whether the cells killed any better. When the group opened TRPM3 directly in cells from 15 patients, killing didn't change. The only report on symptoms is what nine patients recalled about how they were before LDN.
- The calcium that NK killing is known to depend on comes in through a different channel, ORAI1.
What trials in people have found
A mechanism in a dish only matters if patients do better than they would on a dummy pill. In the first small placebo-controlled trials, they did. The larger trials since then haven't repeated it.
| Trial | Who | Result |
|---|---|---|
| Crohn's disease, 2011 | 40 adults, 4.5 mg a day for 12 weeks | 88% improved on LDN and 40% on placebo |
| Fibromyalgia, 2013 | 31 women, each taking 4.5 mg a day and placebo in turn | Pain fell 29% on LDN and 18% on placebo |
| FINAL, 2024 | 99 women with fibromyalgia, 6 mg a day for 12 weeks | Pain fell 1.3 points out of 10 on LDN and 0.9 on placebo, a difference that could be chance |
| INNOVA, 2026 | 98 women with fibromyalgia, 4.5 mg a day, followed for a year | At 3 months pain had fallen 0.33 points out of 10 on LDN and 0.64 on placebo. No meaningful benefit later either |
| Crohn's disease, 2026 | 41 adults, 4.5 mg a day for 12 weeks, stopped early | No more healing of the gut on LDN, and fewer people in remission (22% against 70%). Fatigue scores were better on LDN |
| Post-COVID fatigue, 2026 (Canada, still a preprint) | 160 adults whose ME/CFS symptoms began after COVID, up to 4.5 mg a day for 16 weeks | No difference in fatigue, pain, symptom score, daily steps or quality of life |
A 2026 review of 105 LDN studies across all conditions, 15 of them randomised trials, reached the same conclusion: early positive findings were rarely repeated in later placebo-controlled trials, and the evidence doesn't support routine use.
Much of the encouraging evidence has no placebo group at all. A 2026 review of LDN in long COVID found four studies, of 155 people in all, that compared people with themselves before and after starting it. Fatigue improved moderately, and the review rated the certainty as low. For ME/CFS that didn't follow COVID, no placebo-controlled trial has been published. The evidence there is a review of 218 patients' records, most of whom reported some improvement, and case reports.
Before-and-after studies can mislead here for ordinary reasons. These illnesses go up and down, people tend to start something new when they're at their worst, and expecting to improve changes how symptoms get rated. In INNOVA, the hints of improvement were among the women who believed they were taking LDN.
Whichever of the three is right, none has been shown to make patients better, and the larger trials haven't confirmed what the small early ones reported. An average can hide a minority who respond, which is why the 105-study review suggests testing the drug one person at a time, with periods on and off it. Nobody can yet say who those people would be, and a lab reading of TRPM3 or NK function isn't an established way to tell. The same review leaves room for trying LDN when standard treatments have failed, as long as it's understood to be experimental.
It's generally well tolerated. In INNOVA, side effects were mostly mild and passing, and about as common on placebo (72%) as on LDN (69%). Vivid dreams are often reported. It blocks opioid painkillers, so doctors need to know before any surgery or if opioids are prescribed.
LDN is a small dose of an opioid blocker. The three ideas for how it could help are a rebound in the body's own opioids, calmer microglia, and a restored calcium channel in NK cells. Each rests on lab or animal work. Small early trials looked promising, and the larger placebo-controlled trials since haven't confirmed them.