Extended note Β· for clinicians and interested readers
Long-standing musculoskeletal pain has repeatedly resolved in my patients during prolonged antibiotic therapy given for gastrointestinal indications. This note sets out the mechanisms that could account for it, and what is established and what is not.
01 Starting point
In treating patients for gastrointestinal microbial overgrowth, I kept encountering the same thing: after the gut was cleared, long-standing pain that had never been treated settled as well. Low back, knee, shoulder, often several sites at once β in patients who had not consulted me about pain at all.
I first took this to be incidental. As it recurred, that became difficult to sustain. Several features have been consistent.
The variance in duration is striking, and it does not apply to pain alone. Gastrointestinal symptoms follow the same slow course β patients constipated their whole lives, patients chronically distended, improve gradually over the same period. One does not resolve first with the other trailing behind; the two move together.
Yet at some point treatment can be stopped without relapse. This is not a state that a short course resolves, but once resolved it holds. Taken together with the parallel course of gut and pain, this points toward something common to both that diminishes only slowly.
02 Two paths
Two broad routes are possible.
Gas and inflammatory products of overgrown bacteria and fungi circulate systemically and lower the threshold at which nerves signal pain. Reduce the source and the threshold rises again.
Organisms persist in joints, bone or soft tissue, sustaining low-grade inflammation. Investigations are unremarkable while pain continues. Clear the organism and the pain recedes.
03 Established entities
This is not a hypothesis. Several such conditions are long established and appear in standard texts. They are uncommon and difficult to diagnose, and are therefore typically mistaken for something else for years before being identified.
Tropheryma whipplei
The classic example of an infection presenting for years as seronegative, non-erosive polyarthritis before diagnosis. In a FrenchβItalian multicentre series the median interval from first symptom to diagnosis was five years, and 24 of 29 patients had received immunosuppressive therapy without benefit in the interim.1
In a French registry, the median duration of treated rheumatic disease before diagnosis was 79 months, and 93% went into remission on antibiotic therapy, most of them able to stop their existing drugs.2
Brucella species
A meta-analysis of 56 studies found osteoarticular involvement in 27β36% of patients with brucellosis.3
Of particular relevance here: unlike spondylitis, sacroiliitis and peripheral arthritis are non-destructive and heal without sequelae. Imaging can therefore be entirely normal while pain persists.
Coxiella burnetii
Rare, but the reason the literature gives for delayed diagnosis is precisely the point at issue: joint involvement progresses slowly over a long period and inflammatory markers remain low, so the diagnosis is easily missed.
Presentation as bilateral extensor tenosynovitis has been reported. Treatment runs 18 months or more.
chiefly low-virulence organisms
Among 73 culture-confirmed prosthetic joint infections, 32% had a normal CRP, and 23% had normal serology, normal examination and normal radiographs, with pain as the sole finding.4
The authors noted that had the standard diagnostic criteria been applied, that 23% would never have been diagnosed at all.
What these four have in common is the substantive point. Bacteria resident in musculoskeletal tissue can produce chronic pain; blood tests and imaging may be normal while they do; and diagnosis is delayed by years. None of this is contested.
04 Treatment
The standard regimens are as follows.
| Condition | Standard therapy | Duration |
|---|---|---|
| Whipple's disease arthropathy | Doxycycline + hydroxychloroquine | Prolonged |
| Osteoarticular brucellosis | Doxycycline-based combination (including quinolone regimens) | 6 weeks or more |
| Chronic Q fever, osteoarticular | Doxycycline + hydroxychloroquine | 18 months or more |
Two features are shared. Each regimen is built around a tetracycline, and each runs for months to years, because organisms residing within cells are cleared only slowly.
And none of them is managed with a single agent, since monotherapy invites relapse and resistance. Neither brief nor single-drug treatment is characteristic of these infections.
05 The other route
Tissue residence accounts for only half the picture. A second route β products generated in the gut lowering the pain threshold systemically β is experimentally established.
Hydrogen sulfide, among the gases produced by gut bacteria, directly activates two channels that carry pain signalling in sensory neurons: the T-type calcium channel Cav3.2 and TRPA1.5
More pertinently, hydrogen sulfide instilled into the colon produces not only visceral pain but referred hyperalgesia at sites remote from the gut β that is, input arising in the bowel can raise pain sensitivity systemically.
In a study of fibromyalgia, all 42 patients had an abnormal breath test against 20% of controls, and the severity of somatic pain correlated significantly with breath hydrogen. No such correlation was present in the irritable bowel group.6
The study is old and small, but as a direct human observation that gut gas production and systemic pain move together, it remains among the very few.
It should be added that pain is not the only thing this route produces. The same gases and metabolites present as bloating, halitosis, brain fog and skin complaints. That side of it β organised around breath gas measurement and symptom patterns β is set out at Neo-Skepticism, the same author's work.
The two routes are not mutually exclusive and may coexist in the same patient. The agents used here act on both β they are absorbed systemically and reach tissue, while also substantially altering the gut flora. Clinical response alone therefore cannot distinguish which route was operating.
06 What remains unknown
What is observed clinically is the resolution of pain. Whether it originated in tendon, ligament, joint, or not in tissue at all but in the gut, cannot presently be determined. In particular, no human study has tested the proposition that bacterial colonisation of tendon or ligament causes pain. Not even controlled investigation of bacteria in common chronic tendon disorders β Achilles tendinopathy, lateral epicondylalgia β has been reported. An explanation of the form "there were bacteria in the tendon" therefore does not stand.
Antibiotic therapy has been tested in three randomised trials in patients with lumbar disc pathology and adjacent endplate change. A Danish single-centre trial in 2013 reported a marked effect; subsequent multicentre trials in Norway (2019) and Australia (2026) did not reproduce that effect size.
All three, however, tested an aminopenicillin, confined to the lumbar spine, in patients selected by a specific imaging finding. The agent and the target differ from what is described here. It has also long been argued that the intervertebral disc, being avascular, does not achieve adequate oral antibiotic concentrations β and a small trial delivering antibiotic directly into the disc has since reported a positive signal. That constraint does not apply to vascularised periarticular soft tissue or synovium.
Reports of bacteria detected in tissue from people without pain have accumulated β in normal intervertebral discs, and in the deep shoulder tissue of asymptomatic individuals, at comparable frequency. The detection of organisms at a site therefore cannot by itself identify them as the cause of pain. This remains unresolved across the field.
07 Investigation and use
Each of the conditions in section 03 has an identifiable organism. Where chronic musculoskeletal pain persists and has not responded to conventional treatment, the following may be worth considering.
Interpretation requires care. T. whipplei in particular is carried asymptomatically in the gastrointestinal tract by 2β11% of the general population, so a positive PCR does not establish disease. It must be read against the clinical picture.
Gut-derived input and tissue-resident organisms are difficult to distinguish clinically. A few observations may indicate direction.
If the timing of breath-test normalisation and the timing of pain improvement diverge, gut gas is less likely to be what was sustaining the pain. Conversely, if pain shifts with dietary or non-antibiotic measures alone, that favours the gut. Whether the same result can be reproduced with a non-absorbed agent bears on the same question.
Over prolonged courses it matters not to rely on a single class. Repeated use of one class allows organisms outside its spectrum to survive and occupy the space vacated, producing a further overgrowth. Coverage gaps differ by class β quinolones, for instance, cover anaerobes poorly β so several classes should be combined to give broad and balanced coverage.
Fluoroquinolones carry a recognised risk of tendinopathy, tendon rupture and peripheral neuropathy, and the neurological effects are described in the literature as potentially persistent.
The timing of these events, however, is fairly consistent. Tendon-related events occur largely within the first one to two months; patients who pass that window frequently continue for long periods without difficulty. Regular review, with prompt discontinuation or a change of class at the first sign of a problem, is the substance of managing this.
Managed in this way, I have not encountered a patient left with lasting sequelae. Where neuropathic symptoms did appear, they resolved in every case with adjustment of therapy and repletion of vitamin B12 and B1. With clinical attention, the adverse effects have been controllable. Whether and for how long to treat remains a judgement made on investigation and clinical course.
References
PMID 24187107PMID 38141787PMID 30657765PMID 28321490PMID 22785020PMID 15020342