Antibiotics for Low Back Pain: What the New Cochrane Review Actually Says

Why This Question Exists at All

The idea that some chronic low back pain might be infectious is not new, but it has lived on the edge of mainstream spine care for more than a decade. The hypothesis goes like this: when MRI shows Modic type 1 changes (a specific signal pattern in the vertebral endplate that suggests local inflammation or edema), a low-grade bacterial colonization of the disc could be driving both the imaging finding and the pain. If true, treating the infection should improve the pain.

The hypothesis got attention because earlier work suggested oral antibiotics could meaningfully reduce pain in this narrowly defined subgroup. The question that mattered for the field was whether the original signal would survive replication, with rigorous methods, in independent populations.

The Cochrane team led by Liu and colleagues set out to answer that question with the most careful synthesis possible of the evidence available through August 2025.

What the Review Did

The authors searched the Cochrane Central Register of Controlled Trials, MEDLINE, Embase, and two trial registries. They included randomized controlled trials of antibiotics versus placebo or another active treatment in adults with low back pain, radicular pain, or both. They split the patient population into four MRI-defined subgroups, with the primary comparison focused on patients with Modic type 1 changes plus evidence of disc herniation.

Risk of bias was assessed independently by two reviewers using the Cochrane RoB 1 tool. Outcomes were synthesized using random-effects meta-analysis. Certainty of evidence across each outcome was rated using the GRADE framework.

Three trials met inclusion criteria. They were conducted in Denmark, Belgium, and Norway. Total enrollment across all three was 402 participants, with mean ages between 44.7 and 51.0 years. Two trials tested amoxicillin (one with clavulanate, one without) for 100 days against placebo, with 12 month follow-up. The third tested minocycline versus amitriptyline versus placebo for 14 days in patients with lumbar radicular pain and disc herniation. The third trial included patients regardless of Modic status.

What the Review Found

For the primary patient subgroup (Modic type 1 changes plus disc herniation), oral amoxicillin (with or without clavulanate) may improve pain slightly and may result in small to moderate improvement in disability at 12 to 14 weeks compared with placebo. The numbers: mean back pain score 50.6 out of 100 in the amoxicillin group versus 59.0 out of 100 in placebo (mean difference -8.4 points, 95% CI -16.2 to -0.7, two trials, 255 participants). Mean disability score 45.2 out of 100 in amoxicillin versus 55.7 in placebo (mean difference -10.5 points, 95% CI -15.9 to -5.1).

The certainty of evidence for both pain and disability is low. The Cochrane team downgraded for imprecision and indirectness.

On adverse events, the picture is muddier. Adverse events of any kind were reported in 76 of every 100 people in the amoxicillin groups compared with 49 of every 100 in placebo (risk ratio 1.79, 95% CI 0.54 to 5.94). Serious adverse events were reported in 3 per 100 versus 2 per 100. The certainty of evidence on harms is very low. The wide confidence intervals mean the true risk difference is not pinned down.

For Modic type 2 changes plus disc herniation, the evidence is uncertain. For lumbar radicular pain plus disc herniation without specified Modic status (the minocycline trial), the evidence is uncertain. For patients with any Modic changes but no disc herniation, no trials exist.

What This Means in Plain Terms

A patient asking "could antibiotics fix my back pain" is asking a question this review can almost always answer with no. The evidence base only addresses a narrow subset: chronic low back pain that is moderate to severe, present for more than six months, with MRI confirmation of both Modic type 1 changes and a herniated disc. Most adults with back pain do not have an MRI to start with, do not have Modic type 1 changes if imaged, and do not have a documented herniation matching the population studied.

For the patient who does fit the studied subgroup, the honest read of the evidence is that amoxicillin may help a little. The pain and disability improvements are real but modest, the certainty is low, and roughly three quarters of patients in the amoxicillin arm reported some kind of adverse event.

Antibiotic stewardship is the second consideration. A 100 day course of broad-spectrum oral antibiotic is not a small intervention. The tradeoff between modest, low-certainty back pain benefit and the population-level cost of antibiotic exposure is a conversation that a primary care provider or spine specialist with full clinical context should lead.

What I Take From This in Practice

I see chronic low back pain in my Tulsa walk-in clinic every day. Patients arrive with pain that has lasted weeks or months, often after self-management failed and before they have any imaging. The first clinical question I am answering is rarely an MRI question. It is a mechanical assessment, a conversation about what the pain pattern looks like, and a decision about whether conservative care is the next reasonable step.

This review does not change my approach to that patient. It refines a specific clinical conversation that most chiropractors will have only occasionally. When a patient walks in with prior imaging that already shows Modic type 1 changes plus a herniated disc, I now have a more informed reference point if they ask whether antibiotics are a real option. The honest answer is that the evidence is preliminary, the benefit is modest, the harms are uncertain, and it is a conversation for a prescribing clinician, not for me.

The broader takeaway is methodological. Cochrane reviews are doing what they should do here, which is collecting the available evidence, synthesizing it transparently, and rating the certainty honestly. A low or very low certainty rating does not mean the finding is wrong. It means the design of the evidence does not support strong claims, and that future trials could meaningfully change the estimate. That nuance matters in patient conversations.

Limitations the Review Itself Names

What This Review Does Not Resolve

It does not establish whether bacterial colonization causes Modic type 1 changes. It does not establish whether antibiotic response, where it occurs, is mediated by killing bacteria, by anti-inflammatory effects of the antibiotics themselves, or by some other mechanism. It does not address whether shorter or longer courses, different agents, or different dosing schedules would change the benefit to harm balance.

It also does not extend to acute low back pain, to chronic low back pain without imaging findings, or to the broader population of patients who present with pain but never undergo MRI. Those patients are outside the scope of the question this review answered.