Low-Dose Naltrexone for the Treatment of Fibromyalgia: Investigation of Dose–Response Relationships

Abstract

Objective

This study explores dose–response relationships when treating fibromyalgia with low-dose naltrexone.

Design

A single-blinded clinical trial was carried out using the “up-and-down” method.

Subjects

Subjects included women with a diagnosis of fibromyalgia aged 18–60 years who had been referred to treatment at a public pain clinic at a Danish university hospital.

Methods

The test doses were in the range 0.75–6 mg, and the dosing interval was 0.75 mg. The method was sequential and allowed predicting the dose effective in 50% (ED50) and 95% (ED95) of the subjects when the dose had shifted direction 10 times, and six pairs of “up-and-down” data were available.

Results

A total of 27 subjects were included in the study; two subjects were withdrawn. After inclusion of 25 evaluable subjects, the dose estimates were calculated as 3.88 mg for ED50 and 5.40 mg for ED95. As a secondary outcome, the effects on 10 common fibromyalgia symptoms were evaluated. A high interindividual variation was observed both in the symptom presentation at baseline and in which symptoms were reduced by low-dose naltrexone.

Conclusions

This study is the first to explore dose–response relationships in the treatment of fibromyalgia with low-dose naltrexone. Future placebo-controlled randomized clinical trials are needed, and according to our findings, 4.5 mg, which has previously been used, seems to be a relevant test dose. We recommend that future studies include additional nonpain fibromyalgia symptoms as outcome measures.

Fibromyalgia, Low-Dose Naltrexone, LDN, Dose–Response

Introduction

Fibromyalgia (FM) is a chronic disorder characterized by generalized pain and tenderness, accompanied by a range of symptoms such as fatigue, insomnia, cognitive disturbances, increased sensitivity to other sensory inputs, anxiety, and depression [1]. FM is a common disorder affecting ∼2% of the population [2]. Accumulating evidence indicates that FM patients have changes in their pain regulatory system, with facilitated excitation of nociceptors and reduced central inhibition, leading to a widespread hyperalgesic state [3]. More recently, disturbances in the immune system with signs of an inflammatory state in the central nervous system (CNS) have been found [3–6]. FM patients have been shown to have increased levels of enkephalins in the cerebrospinal fluid (CSF) [7], and decreased mu receptor availability has been demonstrated [8], suggesting that these patients have a dysfunction of the endogenous analgesic mechanisms. No specific treatment of FM is available, and traditional pharmacological therapies aim at either reducing the release of facilitatory neurotransmitters (e.g., gabapentinoids) or blocking the reuptake of both serotonin and norepinephrine [1].

Low-dose naltrexone (LDN) is used widely as off-label treatment for various conditions [9,10], but the evidence for its use in FM is sparse [10]; to our knowledge, only three small clinical trials have been published [11–13].

Naltrexone is primarily known for its antagonistic effect on the opioid receptor [14], but it is also thought to blunt dopaminergic transmission in mesolimbic pathways [15], thereby attenuating craving and reinforcing effects of substances of abuse. Naltrexone is marketed as an additional therapy for supporting abstinence in patients with opioid or alcohol use disorders [16]. It has been known for decades that naltrexone can have a paradoxical analgesic effect when used in low doses [17]. The proposed mechanisms of action for LDN on FM are 1) improvement of opioid signaling and 2) an anti-inflammatory effect. Evidence, primarily from animal studies, has shown that LDN increases levels of both endorphin and met-enkephalin (opioid growth factor [OGF]) and increases expression of opioid receptors and OGF receptors [18,19]. Moreover, LDN binds to Toll-like receptor 4 (TLR4) on glial cells, where it exhibits antagonist properties, leading to reduction of pro-inflammatory cytokines such as interleukin-6 and tumor necrosis factor–alpha [13,20,21].

In the previous studies investigating the effect of LDN on FM, a dose of 4.5 mg was used. However, thus far, no dose–response studies have been carried out. The aim of this study was to explore dose–response relationships when treating FM patients with LDN, and thereby to estimate the dose effective in 50% of subjects (ED50) and the dose effective in 95% of subjects (ED95). A secondary aim was to investigate which FM symptoms were most commonly reduced by LDN.

Methods

Procedure

A single-center study was conducted at a public university hospital pain center in Southern Denmark. Approval was obtained from the Danish Data Protection Agency, the Ethical Committee of Southern Denmark (S-20160121), and the Danish Health and Medicines Authority. The study was registered with the European Union Drug Regulation Authorities Clinical Trials Database (EudraCT-nr: 2016–002081-31) and was monitored by the Good Clinical Practice (GCP) unit at Odense University Hospital. Informed consent was obtained from all subjects entering the study. Enrollment began in June 2017, and the study was completed in September 2018. The study medication was produced at Glostrup Pharmacy, and the tablets were blinded in similar white gelatin capsules and labeled with blinding codes. The medicine was administered orally once daily in the evening.

Entry Criteria

Subjects eligible for the study were Caucasian women aged 18–60 years who had a diagnosis of FM. The subjects were required to understand and write Danish, and all fertile women were requested to use secure anticonception or to be sexually abstinent for the three weeks before entering and the one week after concluding the trial. Exclusion criteria included known allergy toward naltrexone hydrochloride, any known inflammatory rheumatic disease, any known neurological disease, other significant localized pain conditions, psychiatric disease, suicidal ideation, suicidal attempt during the past five years, pregnancy or breastfeeding, use of opioids within the eight weeks before entering the trial, and abuse of alcohol or other substances.

All subjects included in the study were referred to treatment at the pain center and were identified at the end of the treatment course as patients who had not benefitted sufficiently from traditional pharmacological treatment. Because male patients with a diagnosis of FM are rarely referred to the pain center, it was decided to include only female subjects. Subjects eligible were examined by a pain specialist (KBP) at a screening interview, and all included subjects fulfilled both the classification criteria approved by the American College of Rheumatology (ACR) in 1990 [22] and the diagnostic criteria approved by the ACR in 2011 [23]. A pregnancy test was performed in all fertile subjects at inclusion.

Subjects were allowed to continue their usual medication during the trial, but the treatment had to be stable during the trial. Any change in usual medication or initiation of new medicine that could influence pain, sleep, or fatigue during the trial would lead to exclusion from the trial.

The subjects did not receive any compensation for participation in the trial.

Study Design

The study was designed as a prospective dose–response study using the “up-and-down” method [24]. The method is sequential and allows predicting the ED50 and ED95, when the dose has changed direction 10 times and six pairs of “up-and-down” data are available. These ED50 and ED95 serve as a range for a reasonable dose. The method usually requires 20–30 patients [24]. In this study, the dosing interval was 0.75 mg, and the chosen test doses were 0.75 mg, 1.5 mg, 2.25 mg, 3 mg, 3.75 mg, 4.5 mg, 5.25 mg, and 6 mg. All subjects received active treatment but were blinded to which dose they received. The primary investigator (KBP) was not blinded to which dose the subjects received.

The subjects had three visits during the trial, and all visits were performed by the primary investigator. At the first visit (day 1) baseline data were collected and medicine was dispensed in separate containers for the first two weeks of treatment (14 capsules) and for the third week of treatment (seven capsules). Visit 2 was a telephone visit (day 15), and visit 3 (day 29) was a follow-up meeting one week after the end of treatment. At visits 2 and 3, the subjects were interviewed about adverse events and compliance with the treatment. At visit 3, the empty medicine containers were returned, and any noningested capsules were counted. A subject was considered compliant if no more than two capsules were returned after the first two weeks of treatment and if no more than one capsule was returned after the third week of treatment. At all visits, the shared electronic medical record was updated to ensure that concomitant medication was stable. Electronic questionnaires were sent out by e-mail after two and three weeks of treatment. Because of the sequential method, subjects who were withdrawn or dropped out during the trial were not evaluated but were replaced.

The first subject was randomized to treatment with 2.25 mg, 3 mg, 3.75 mg, or 4.5 mg using the envelope method. Evaluation of response was made after two weeks of treatment based on the electronic questionnaires. The evaluation was made before visit 2 to reduce bias. The primary end point was subjective evaluation of improvement of overall FM symptoms using a seven-point transition scale—the Patient Global Impression of Improvement Scale (PGI-I). A subject was considered a responder if she scored 1–3 on the PGI-I scale, which ranges from 1 (very much improved) to 7 (very much worse), with 4 being no change. An additional primary end point was chosen to be a >30% reduction in pain from baseline, measured as average pain during the last three days on a 0–100 visual analog scale (VAS).

The primary end point was evaluated after two weeks, but the treatment was given for a total of three weeks to evaluate if significant changes in effect occurred between two and three weeks of treatment. If it turned out in the beginning of the trial that significant improvement occurred between two and three weeks of treatment, the protocol of the study allowed for changing the time for evaluation of the primary end point to three weeks instead of two weeks.

Secondary Outcomes

To evaluate the effect of LDN on other FM key symptoms, the 10 items from the symptom part of the Fibromyalgia Impact Questionnaire Revised (FIQR) [25] were chosen as secondary outcome measures. Each item evaluates the average severity of the symptom during the last seven days on a 0–10 numeric rating scale (NRS). The composite value of the 10 items from the symptom part of the FIQR (sFIQR) was calculated as a rough measure of symptom burden. The Insomnia Severity Index (ISI) questionnaire was used to evaluate changes in sleep quality [26].

Statistical Analyses

The primary analysis was made per protocol. Given that the “up-and-down” method [24] does not allow for imputation or related methods, subjects who were withdrawn or dropped out or who did not complete the treatment as intended were replaced. To account for the implicit assumption of increasing effect with increasing dose, the study applies isotonic regression, as suggested by Pace and Stylianou [24], which provides estimates of ED50 and ED95 with smaller bias than the traditional Dixon and Mood (1948) estimators [27]. The analysis was performed in STATA 15.0 using the IRAX module developed by van Putten and Royston [28]. As an extension of the isotonic regression, ED50 and ED95, together with their 95% confidence limits, were thus not calculated from data but predicted by the pooled-adjacent-violators algorithm (PAVA), as suggested by Pace and Stylianou [24]. Descriptive statistics (numbers, percentages, means, standard deviations, and interquartile ranges) were used to evaluate demographic data and secondary outcome measures.

Results

Patient Disposition

A total of 28 women were screened for the study, and 27 met the entry criteria. Two subjects were withdrawn from the trial, as they did not complete the treatment as intended. Both subjects were replaced. After inclusion of 25 evaluable subjects, the dose had changed direction 10 times, and six pairs of “up-and-down” data were obtained. Baseline data for all evaluable subjects and data on concomitant pain medication are presented in Table 1.

Table 1

Raw scores for demographics, clinical pain profile, level of the 10 items from the symptom part of the FIQR, symptom burden measured as composite value of the 10 items of the symptom part of the FIQR, degree of sleep disturbance measured by the ISI, and analgesic use at baseline

Demographic and Clinical Features at Baseline(N = 25)	Mean±SD (Range)	Interquartile Percentiles(25, 50, 75)
Age, y	47.0±9.4 (27–59)	(41, 49, 55)
BMI, kg/m2	29.4±6.4 (18–42)	(24, 28, 32)
Pain duration, y	13.6±11.1 (1–44)	(7, 10, 20)
Average pain intensity at baseline (NRS: 0–100)	67.2±17.0 (14–94)	(59, 66, 82)
FIQR items, average value during last 7 d (0–10 NRS)
Pain	7.0±1.4 (3–9)	(6–8)
Energy	7.4±1.7 (3–10)	(6, 8, 9)
Stiffness	6.7±1.9 (3–9)	(5, 7, 8)
Waking unrefreshed	7.9±2.1 (2–10)	(7, 8, 10)
Depressed	4.0±2.7 (0–9)	(2, 4, 6)
Concentration/memory	6.7±2.0 (1–9)	(6, 7, 8)
Anxiety	2.4±2.8 (0–10)	(0, 1, 5)
Tenderness to touch	7.6±2.2 (3–10)	(6, 8, 9)
Imbalance	3.9±2.8 (0–8)	(2, 4, 7)
Sensitivity to sensory inputs	6.1±2.8 (0–10)	(4, 7, 8)
Composite value for symptom part of FIQR at baseline (0–100)	59.6±12.0 (32–85)	(50, 60, 68)
ISI score at baseline (0–28)	17.6±5.0 (1–25)	(16, 18, 20)
Concomitant pain medication, No. using medication (%)
Paracetamol users	16 (64)
NSAID users	8 (32)
TCA users	1 (4)
Anticonvulsive users	2 (8)
Baclofen users	11 (44)
Tizanidine users	4 (16)
SNRI users	6 (24)
SSRI users	2 (8)

BMI = body mass index; FIQR = Fibromyalgia Impact Questionnaire Revised; ISI = Insomnia Severity Index; NRS = numeric rating scale; NSAID = nonsteroidal anti-inflammatory drugs; SNRI = serotonin noradrenaline reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor; TCA = tricyclic antidepressant.

Primary Outcome

Of the 25 subjects analyzed, 11 were classified as responders, reporting either a minimum 30% decrease in pain from baseline or having a PGI-I score of 1–3 after two weeks of treatment. The up-and-down curve is shown in Figure 1, and the observed and fitted doses with 95% confidence intervals are shown in Table 2. In Figure 1, each circle is a patient, and the doses are shown in order from left to right. The first three patients did not respond, but the one given a dose of slightly over 5 did; hence the fifth patient was tried on a lower dose. That one failed, so the dose was increased to >5. Then patients 6, 7, 8, and 9 all succeeded, on increasingly lower doses, until number 10 failed, etc. The response rate (equal to the patient sequence number divided by the final sample size) is shown on the horizontal axis. The 95% confidence interval (CI) of Figure 1 and Table 2 is an estimated pointwise, partition-wise confidence interval for the fitted step function, which IRAX fits by making auxiliary regression on the partition groups; see van Putten and Royston for details [28]. The PAVA estimates of ${\mu }_{50}$ and ${\mu }_{95}$ with 95% confidence limits were obtained as ${\mu }_{50}=3.88\mathrm{ }(95\mathrm{\%}\mathrm{ }\mathrm{CI}=3.39\mathrm{ }\mathrm{to}\mathrm{ }4.35)$ and ${\mathrm{\mu }}_{95}=5.40(95\mathrm{\%}\mathrm{ }\mathrm{CI}=4.66\mathrm{ }\mathrm{to}\mathrm{ }6.13)$⁠.

Figure 1

The up-and-down curve, showing isotonic regression of dose (mg) vs response rate with 95% confidence limits.

Table 2

Observed and fitted doses (mg) with lower and upper 95% confidence limits

ID	Effect	Response Rate	Dose Observed	Dose Fitted	Lower 95% CI	Upper 95% CI
1	N	0.04	3.000	3.000	1.348	4.652
2	N	0.08	3.750	3.750	2.098	5.402
3	N	0.12	4.500	3.875	3.398	4.352
4	Y	0.16	5.250	3.875	3.398	4.352
5	N	0.20	4.500	3.875	3.398	4.352
6	Y	0.24	5.250	3.875	3.398	4.352
7	Y	0.28	4.500	3.875	3.398	4.352
8	Y	0.32	3.750	3.875	3.398	4.352
9	Y	0.36	3.000	3.875	3.398	4.352
10	N	0.40	2.250	3.875	3.398	4.352
11	N	0.44	3.000	3.875	3.398	4.352
12	Y	0.48	3.750	3.875	3.398	4.352
13	N	0.52	3.000	3.875	3.398	4.352
14	N	0.56	3.750	3.875	3.398	4.352
15	N	0.60	4.500	4.500	2.848	6.152
16	Y	0.64	5.250	4.875	3.707	6.043
17	N	0.68	4.500	4.875	3.707	6.043
18	N	0.72	5.250	5.250	3.598	6.902
19	N	0.76	6.000	5.400	4.661	6.139
20	Y	0.80	6.000	5.400	4.661	6.139
21	Y	0.84	5.250	5.400	4.661	6.139
22	N	0.88	4.500	5.400	4.661	6.139
23	N	0.92	5.250	5.400	4.661	6.139
24	Y	0.96	6.000	5.625	4.457	6.793
25		1.00	5.250	5.625	4.457	6.793

CI = confidence interval.

Of the 11 responders, one subject fulfilled only the pain reduction criteria, four subjects fulfilled both criteria, and six subjects fulfilled only the PGI-I criteria (Table 3). After three weeks, 10 of the responders still fulfilled the criteria; for the last responder, data were lost due to problems with the electronic questionnaire. Three of the subjects who were classified as nonresponders after two weeks fulfilled the criteria for being responders after three weeks. All three subjects changed their PGI-I score from 4 (after two weeks) to 3 (after three weeks), and one of these patients also had a >30% decrease in average pain after three weeks compared with baseline (data not shown).

Table 3

Responder classification

Responders (N = 11)	30% Change in Pain After 2 Weeks (0–100 NRS;Yes/No)	Patient Global Impression of ChangeAfter 2 Weeks (1–7)	30% Change in Pain After 3 Weeks (0–100 NRS;Yes/No)	Patient Global Impression of Change After 3 Weeks (1–7)
1ID 4	Yes	4	Missing value	Missing value
2ID 6	No	3	No	3
3ID 7	No	1	No	1
4ID 8	Yes	2	Yes	2
5ID 9	Yes	2	Yes	2
6ID 12	Yes	3	Yes	3
7ID 16	No	3	No	3
8ID 20	Yes	2	No	2
9ID 21	No	2	No	3
10ID 24	No	3	No	3
11ID 25	No	2	No	3

This table shows how the responders were classified. Primary outcomes were Patient Global Impression of Improvement Score, ranging from 1 (very much better) to 7 (very much worse) and average daily pain during the past three days on a 0–100 NRS. A responder had to have either a 30% decrease in pain on a 0–100 NRS or a Patient Global Impression of Change Score between 1 and 3. All the patients who were classified as responders after two weeks also fulfilled the criteria after three weeks.

NRS = numeric rating scale.

Secondary Outcomes

Data for the 11 responders are shown in Table 4. The two items from the FIQR with the highest mean change score after two weeks were self-perceived “tenderness,” with a mean change of –2.3 (0–10 NRS) and five of 11 reporting a >30% improvement, and “waking unrefreshed,” with a mean change of –2.3 and eight of 11 reporting a >30% improvement. For the item “pain,” the mean change was only –1.4 (0–10 NRS), and only four of 11 reported a >30% improvement after two weeks. After three weeks, the improvement in pain was even less prominent. Most of the other items remained stable or improved further after three weeks. The item self-perceived “tenderness,” especially, was further improved after three weeks, with a mean change of –3.5 (0–10 NRS) from baseline and eight of 11 reporting a >30% improvement.

Table 4

Raw scores for symptom burden at baseline for the 11 responders, measured by mean score of each of the 10 items from the symptom part of the FIQR, the composite value of the 10 items of the FIQR, and mean changes in these scores after two and three weeks of treatment with LDN

Responders (N = 11)	Baseline (N = 11), Mean±SD (Range)	After 2 Weeks (N = 11), Mean Change±SD	After 3 Weeks (N = 10), Mean Change±SD	>30% Improvement After 2 Weeks, No./Total	>30% Improvement After 3 Weeks, No./Total
FIQR items, average value during last 7 d (0–10 NRS)
Pain	6.7±1.5 (3–9)	–1.4±1.3	–0.4±2.0	4/11	2/10
Energy	7.4±2.1 (3–10)	–1.5±2.0	–1.7±2.3	4/11	4/10
Stiffness	6.3±2.1 (3–9)	–0.6±2.5	–1.2±2.4	2/11	4/10
Waking unrefreshed	7.8±2.0 (3–10)	–2.3±2.1	–2.2±1.8	8/11	5/10
Depressed	3.7±3.1 (0–9)	–1.5±2.0	–1.2±2.1	6/11	4/10
Concentration/memory	6.3±1.9 (3–9)	–1.2±1.9	–1.2±2.4	4/11	4/10
Anxiety	2.0±2.5 (0–7)	–0.5±1.1	–1.3±2.5	3/11	4/10
Tenderness	7.2±2.1 (3–10)	–2.3±1.9	–3.5±1.8	5/11	8/10
Imbalance	3.8±2.9 (0–8)	0, 5±2.5	–0.8±1.8	1/11	4/10
Sensitivity to sensory inputs	5.5±2.8 (0–9)	–0.3±2.1	–0.2±2.6	2/11	3/10
Composite value of the 10 items from the symptom part of FIQR (0–100)	56.6±13.6	–10.8±9.3	–13.6±12.1	2/11	5/10
Insomnia Severity Score (0–28)	16.9±2.6	–3.3±4.6	–5.4±4.6	4/11	4/10

FIQR = Fibromyalgia Impact Questionnaire Revised; LDN = low-dose naltrexone; NRS = numeric rating scale.

All responders had a minimum improvement of 30% on at least one of the 10 FM symptoms, but most of the patients showed a minimum improvement of 30% on several symptoms. This is illustrated by the mean change in the composite value of the 10 items from the sFIQR, which showed a mean value of 56.6 ± 13.6 (0–100 NRS) at baseline, with a mean change of –10.8 ± 9.3 after two weeks and a mean change of –13.6 ± 12.1 after three weeks. After two weeks of treatment, the average number of domains with >30% improvement was 3.5, and after three weeks, the average number of domains improved was 4.2. Nonresponders (N = 14) had a mean change of 3.3 ± 9.9 in sFIQR score after two weeks and a mean change of 1.1 ± 10.2 on the sFIQR after three weeks (data not shown).

Adverse Events

No serious adverse events occurred during the trial, but side effects were common. Two subjects were withdrawn from the study because of noncompliance. Both the subjects had many side effects. One reported severe nausea, abdominal pain, and headache and scored 6 on the PGI-I scale after two weeks and did not want to continue the treatment for the last week; the other only ingested five capsules and reported fatigue, depression, headache, and abdominal pain to be the reason for withdrawal from the study. Both patients received doses on the low end of the dose range (3 mg and 3.75 mg). Adverse events were common but were generally graded mild and tolerable. Gastrointestinal symptoms were the most commonly reported side effects. The side effects reported after two weeks of treatment are listed in Table 5.

Table 5

Reported side effects after two weeks of treatment

Reported Side Effects (27 Subjects)	Dose Received, 2.25 mg (1 Subject)	Dose Received, 3 mg (5 Subjects)	Dose Received, 3.75 mg (5 Subjects)	Dose Received, 4.5 mg (6 Subjects)	Dose Received, 5.25 mg (7 Subjects)	Dose Received, 6 mg (3 Subjects)
Abdominal ache	1	1	1	2	2
Diarrhea		1		1	1
Constipation					2
Nausea		1	2
Headache		2	1	1
Vivid dreams			1
Mood disturbance			1
Increased pain					1
Increased fatigue			1
Dizziness	1
Palpitations		1
Increased appetite						1
Sleeping difficulty				1

The subjects received different doses; the number of subjects reporting a given side effect is listed for each dose. Some subjects reported more than one side effect. Side effects were common, and only six of 27 subjects did not report any side effects.

Discussion

In this study, we explored dose–response relationships when treating FM with LDN using the “up-and-down” method. We estimated the ED50 to be 3.88 mg and the ED95 to be 5.40 mg. Since its introduction in the 1980s [29], LDN has been used widely as off-label treatment for fibromyalgia and other chronic pain conditions [10]. Several case reports have been published reporting a pain-relieving effect of LDN [30–32]. The evidence is sparse, however, with few clinical trials. The doses used in the published case reports varied from 1 mg to 5 mg [10], but in all previous clinical trials testing the effect of LDN on FM, a dose of 4.5 mg has been used [12]. To our knowledge, no dose–response studies have ever been published. For future studies of the effect of LDN on FM, it is important to estimate a dose that is sufficiently effective for many patients and still not too high to give rise to intolerable side effects, which would lead to high dropout rates and poor quality of the studies. Based on our findings in this study, a dose of 4.5 mg seems to be a reasonable test dose in FM patients, as it lies in the range between our estimates of ED50 and ED95.

Given the sequential method, a relatively short treatment period was needed. Time-effect curves from previous studies [11,12] show that FM symptoms seem to improve during the first two weeks of treatment, with some further improvement during the following six to 10 weeks. Previous studies have shown that 57–60% of patients get >30% relief of pain when treated with LDN for eight to 12 weeks [11,12]. We found that only a smaller percentage of the responders in this study reported a reduction >30% in pain after two weeks, and the average reduction of pain after three weeks was minimal. Instead we found that the FIQR item self-perceived “tenderness” was the symptom most reduced on average after both two and three weeks of treatment. The design of this study does not allow us to make any conclusions about secondary outcome measures, but it does give rise to some hypothetical considerations. The decrease in self-perceived tenderness could reflect an improvement of the hyperalgesic state seen in FM patients, and we hypothesize that this, together with improvement of energy and sleep, might lead to increased activity, which could blunt the expected improvement of pain. Previous studies exploring the effect of LDN on FM have been longer, and patients might find a new balance in daily activities over a longer period of time, which might explain that studies with longer duration find pain to be significantly reduced by LDN. The dose of LDN may also play a role, as previous studies have used 4.5 mg, and some of the responders in this study received doses below that.

It is well known that there is high interindividual variability in which symptoms FM patients report to be most severe [33]. From clinical experience, there is also interindividual variability in which symptoms are relieved by LDN. Many patients report improvement of sleep and/or fatigue instead of pain relief as the major benefit from the treatment. In this study, we also found a great variability in both symptoms at baseline and, as an expected consequence of this, interindividual variability in which symptoms were relieved by LDN. In line with the IMMPACT guidelines [34], we therefore recommend that future studies on the effect of LDN on FM include phenotyping of the subjects to account for possible variations in pathophysiology and to include outcome measures of key FM symptoms other than pain, as well as measures of changes in function.

Limitations

Our results might be biased by several factors. Given the sequential method, we had to evaluate the effect of the treatment after a relatively short period of time, and two weeks was chosen based on time–response curves from previous trials. This period of time might not have been sufficient to give a positive effect in all patients, as the effect of LDN is thought to be mediated through opioid receptor upregulation and attenuation of inflammatory pathways, which might require more time. We found that three patients reported positive effect after three weeks but not after two weeks. During the trial, we did not find these three cases sufficient to change the evaluation of response at three weeks, but if the evaluation had been made after three weeks, it might have given rise to a lower estimated ED50 and ED95. The test doses in the study were chosen based on our clinical experience that many FM patients benefit from lower doses than 4.5 mg, and as we did not have any experience using doses higher than 4.5 mg, we were reluctant to choose doses higher than 6 mg. During the study, it became clear that the subjects tolerated doses up to 6 mg well. If we had chosen to include test doses above 6 mg, we might have found a higher estimated ED50 and ED95. Another limitation to the trial is that some patients might have experienced a placebo effect after only two weeks of treatment. It should be noted, though, that all responders reported a positive effect after both two and three weeks of treatment, and all responders had a minimum 30% decrease of several common FM symptoms, with the average number of domains improved after two and three weeks being 3.5 and 4.2, respectively. The subjects in this study were drawn from a university hospital, and the subjects had typically failed to benefit from traditional therapies. These subjects belong to the severe end of the spectrum of FM and might benefit less from treatment with LDN than FM patients with a milder level of the disease. The study was only single-blinded. Double-blinding would have been preferable but was not feasible due to limited funding. To minimize bias, subjects completed their questionnaires electronically before the visit, thereby minimizing investigator influence on response. This study was designed as a dose–response study, and evaluation of effect parameters must be made with caution.

Conclusions

In this dose–response study investigating the effect of LDN on FM using the “up-and-down” method, we have estimated the ED50 to be 3.88 mg and the ED95 to be 5.40 mg. Larger randomized controlled trials comparing LDN with placebo are needed in the future, allowing for large-sample traditional statistics like chi-square-based tests for effects. Given that the “up-and-down” sample method does not support sample size or power calculations for such a trial, a pilot trial might conveniently precede. Choosing a relevant test dose is crucial, as a test dose that is too low might lead to low response rates and a test dose that is too high might lead to high dropout rates because of side effects. Based on our current findings, we conclude that 4.5 mg, which has been used in previous trials, seems to be a good choice, as it lies between our estimates for ED50 and ED95.

We suggest that future clinical trials exploring the effect of LDN on FM incorporate measures of “tenderness,” valid and reliable measures of hyperalgesia, and measures of physical functioning to explore the hypothesis that LDN primarily influences hyperalgesia, fatigue, and sleep and secondarily influences pain.

Funding sources: KBP received funding for the study from the Danish Rheumatism Association (R155-A4821-B1326).

Conflicts of interest: The authors have no conflicts of interest to declare.