Journal Club: Antibiotics for back pain

6 min read May 9, 2013 at 5:28am on Health, Infection and Science

This is a bit of a departure for me, but this article has been getting a lot of attention in the media, so I think it's worthwhile doing.

First, a disclaimer. No blog or any other page on the internet is a substitute for proper medical advice you get from sitting down and talking to your doctor. Every situation is different - and I think that particularly for this paper, as it deals with a chronic condition, the best person to speak to is the one who knows you best.

The abstract for the article is available here and the full version is here (will need to login to your hospital / university website to get the full version).

Some of the sensationalist media it's received is here or here. ABC Radio did a piece, and a more skeptical report is here.  Michael Vagg has covered the article over at The Conversation and must be a clever guy because he largely agrees with what I've got to say below.

Short summary

A randomised, placebo controlled trial of amoxicillin/clavulanate given for 100 days to people with chronic lower back pain has showed a statistically significant reduction in pain, functional assessment and MRI changes.


Chronic back pain is a common cause of presentation to GPs and accounts for quite a bit of lost working time. By and large, existing treatments are not highly effective, and mainly involve rest early and then gentle mobility with pain relief.  There, unfortunately, is no magic bullet - no matter what people say. Chiropractic is about as effective as standard care - which is to say not very. In clinical studies, most patients with non-specific lower back pain, it will resolve spontaneously - this tends to occur in spite of treatment rather than because of it. Unfortunately, a small proportion of patients will go on to have severe, ongoing symptoms. There is concern in medical circles about the steadily increasing prescription of strong opioid pain killers - morphine, oxycodone tablets and fentanyl / buprenorphine patches. The Australian therapeutic guidelines recommend against chronic use of these agents. The other treatment option is surgery - although for most cases where there is not a substantial anatomical defect, the results are also less than inspiring. Another problem is that surgery is frequently done late in chronic back pain  - after everything else has been tried, and a chronic pain state has been established and which surgery may not fix.

Rationale for the trial

For most patients with back pain without any of the warning signs, no scans or x-rays are generally needed. If an MRI is done, changes called "Modic changes" are commonly seen. These can be thought of as inflammatory changes in the edges of the vertebrae.  I'm not an orthopaedic surgeon, but I am reliably informed that these modic changes are an important predictor of pain - which makes sense.  Inflamed bones are painful.

The next leap is where I have the first big problem with the article. Not all inflammation is due to infection.  I hope this is pretty obvious.  Think of a sprained ankle;  it gets really swollen, but there's no suspicion that it's infected.  So it is (I think) with these modic changes.  

The paper's rationale for using antibiotics is based on a previous work of Sterling and colleagues (doi:10.1016/S0140-6736(00)05109-6) and quote it thus "Stirling found nuclear tissue removed under strict sterile conditions to be infected with low virulent anaerobic organisms, (Propionibacterium acnes and Corynebacterium propinquum) in 53% of patients.  

It's also worth noting that the samples were incubated for a long time in the lab (increasing the risk of contamination by the lab staff) and that this was tissue taken from the intravertebral discs (not the bone end-plates) of people with sciatica and bulging discs (not functional lower back pain).

Now Propionibacterium is a tricky germ to grow - slow growing, and grows better under conditions without oxygen - which is fiddly in the lab.  As well as being the germ behind acne, it is also prone to cause infections in shoulder reconstructions.  Those aside, the vast majority of times we isolate this germ in the lab, we think of it as a contaminant rather than a true infection. It is a ubiquitous organism on the skin, and if not enough care is taken collecting samples, it is very easy to contaminate a sterile specimen with a small amount of this germ.  

Obviously in response to the Stirling letter, was the paper of Carricajo et al (doi:10.1016/j.jhin.2007.04.007) which found that only 2 of 54 patients with herniated discs cultured P. acnes, and that both these patients also cultured the organism from the ligamentum flavum and paraspinal muscles.  As well as that, 18 other patients also cultured the organism from the control sites but not the disc, and four air samples from the operating theatre were also positive for P. acnes.    Their conclusion was that contamination was a far more likely explanation than a low-grade infection.

Statistical Power

The trial was small - around 70 patients in each group, but the sample size was calculated to find a difference in the grading scores, with a power of 0.90 - the short version of this is that there's a 1/10 chance of a Type II error - that is finding a difference where none exists - purely by chance.   You would think that such a large magnitude of difference would make a statistical fluke less likely, but in fact a small sample size could result in a cluster of aberrant results away from the true answer more easily than a larger trial.

Adverse Effects

The adverse effects in this trial were significant (all antibiotic vs placebo):

  • Any adverse effects:  65% vs 23%  (mainly loose bowel motions / flatulence)
  • "Middle grade" effects:  27% vs 11% (defined as lasting > 3 weeks)
  • "Considerable" effects:  21% vs 6%  (not defined at all)

What wasn't considered was secondary effects of the antibiotics - creation of resistant organisms in the patients which may not come up during the duration of follow-up of the trial, and the wider effects of antibiotic resistance on the community.


A well-designed, but small trial has shown marked improvement in outcomes of a condition for which, up until now, treatments have been largely unsatisfactory.  Although there is biologic plausibility, the causal link is weak, and there has been another small study which specifically calls into question the whole premise that underpins the trial.  The treatment is cheap, and certainly a better option than putting people on chronic opioids, but has significant side effects - and further population-level side effects which the trial does not look into (and which are generally poorly appreciated by everyone - public and doctors). Far from being the potential Nobel prize winner that it's been reported as, I think this trial - overall - is a bit of a lemon.   I'd like to see a much larger trial, which also includes some data on rates of trial-acquired resistance.  But it is an important first step - and as the cliche goes, more research is definitely needed.