Br J Sports Med 42:969-973 doi:10.1136/bjsm.2008.050369
  • Original article

Antibiotic treatment in patients with low-back pain associated with Modic changes Type 1 (bone oedema): a pilot study

  1. H B Albert1,
  2. C Manniche1,
  3. J S Sorensen1,
  4. B W Deleuran2
  1. 1
    The Back Research Center, Funen, University of Southern Denmark, Denmark
  2. 2
    Department of Rheumatology, Aarhus University Hospital and The Institute of Medical Microbiology and Immunology, Aarhus University, Aarhus, Denmark
  1. Hanne Albert, The Back Research Center, Lindevej 5, 5750 Ringe, Denmark; haal{at}
  • Accepted 4 July 2008
  • Published Online First 21 August 2008


Objective: The aim of this study was to assess the clinical effect of antibiotic treatment in a cohort of patients with low-back pain (LBP) and Modic changes Type 1 (bone oedema) following a lumbar herniated disc.

Design: This was a prospective uncontrolled trial of 32 LBP patients who had Modic changes and were treated with Amoxicillin-clavulanate (500 mg/125 mg) 3 × day for 90 days. All patients had previously participated in a randomised controlled trial (RCT) that investigated active conservative treatment for a lumbar herniated disc (n = 166). All patients in that RCT who had Modic changes and LBP at 14 months follow-up (n = 37) were invited to participate in this subsequent antibiotic trial but five did not meet the inclusion criteria.

Results: 29 patients completed the treatment, as three patients dropped out due to severe diarrhoea. At the end of treatment and at long-term follow-up (mean 10.8 months) there was both clinically important and statistically significant (p⩽0.001) improvement in all outcome measures: LBP intensity, number of days with pain, disease-specific and patient-specific function, and global perceived effect.

Conclusions: In this uncontrolled trial, the clinical effect of antibiotic treatment was large in a group of patients with Modic changes suffering from persistent LBP following a disc herniation. These results provide tentative support for a hypothesis that bacterial infection may play a role in LBP with Modic changes and indicate the need for randomised controlled trials to test this hypothesis.

Peer review: fair review articles

The rationale for “peer review: fair review” articles is to ensure that research is not buried simply because it is too challenging and too controversial. There are many examples of papers that were not accepted the first time they were submitted, but were accepted elsewhere and have made a great difference to clinical practice (Khan KM, Stovitz SD, Pluim B, et al. Addressing conflicts of interest and clouding of objectivity: BJSM's "Peer review: fair review" section. Br J Sports Med 2008;42:79). BJSM is committed to encouraging debate and providing a ''safe place'' for ideas that are supported by evidence, but considered ''too radical'' elsewhere.

Modic changes (bone oedema) in vertebrae are an imaging finding recently identified as a prevalent pathoanatomical finding that is commonly associated with low-back pain (LBP).1 2 Modic changes are only visible on magnetic resonance images (MRI)3 and three subtypes have been identified (Types 1 to 3). From histological studies of material harvested during surgery, Modic changes Type 1 involve disruption and fissuring of the endplate with regions of degeneration, regeneration, reactive bone formation, endplate oedema and vascular granulation tissue.3 4 Type 1 are seen on T2-weighted MRI as areas of increased signal intensity and on T1-weighted MRI as low signal intensity extending from the vertebral endplates.

There is an association between Modic changes and LBP.1 2 A recent systematic review of Modic changes and LBP identified 77 study samples from the general, working, and clinical populations. The median prevalence rate for any type of Modic change was 46% in patients with non-specific LBP and 6% in non-clinical populations. A positive association between Modic changes and non-specific LBP was found in 70% of studies with odds ratios ranging from 2.0 to 19.9.5

Infection is one hypothetical cause of Modic changes Type 1.6 Both Van Goethem et al7 and Modic et al8 showed that vertebral endplate signal changes resembling Modic changes Type 1 were a sensitive indicator for spondylodiscitis or disc space infection. Caragee9 observed that about one-third of patients with pyogenic vertebral osteomyelitis were infected with low-virulent bacteria. When antibiotics were given, the majority recovered and became pain- and symptom-free. Similarly, in nuclear tissue removed under sterile conditions during surgery for lumbar herniated discs, 53% of patients were found to be infected with low-virulent anaerobic organisms (Propionibacterium acnes and Corynebacterium propinquum) in contrast to no patients who were operated on for other spinal disorders.10 Stirling et al,10 11 therefore, hypothesised that patients with sciatica sustain a breach in the mechanical integrity of the spinal disc, possibly from minor trauma, which allows access by low-virulent microorganisms.10

The aim of this pilot study was to test antibiotic treatment in a group of patients with Modic changes Type 1 and LBP following herniation of a lumbar disc.


This was a prospective uncontrolled trial of LBP patients who had Modic changes Type 1 (bone oedema). All patients had previously participated in a randomised controlled trial (RCT) comparing two types of active conservative treatment for a lumbar herniated disc (n = 166).12 All patients in the RCT received a MRI scan at the 14 month follow-up.

Patients were invited to participate in the current trial if their follow-up MRI displayed Modic changes Type 1 in a vertebra adjacent to their previous herniated disc and they had LBP at the time of this follow-up examination. Patients were excluded if they had an allergy to antibiotics, had a current infection or declined participation in the antibiotic trial.

Treatment consisted of Amoxicillin-clavulanate (500 mg/125 mg) (Spektramox) three times a day, at 8 hour intervals, for 90 days. Three independent experts in infectious diseases were presented with the bacterial culture results in Stirling’s10 study and requested to suggest the ideal antibiotic. All experts recommended Amoxicillin-clavulanate. Treatment was for 90 days as this is the usual duration of antibiotic treatment for postoperative discitis. During the treatment and follow-up period of this antibiotic trial, the patients received no other treatment except optional mild analgesics.

Outcome measures

All patients had a clinical examination and filled in questionnaires at baseline, at end of treatment and at a follow-up approximately 11 months later. Serum analysis was performed at baseline and at the end of treatment.

The outcome measures were: Global perceived effect, Roland Morris Questionnaire (RMQ),13 the Self-Perceived Function Scale, days with LBP, the LBP Rating Scale14 and serum analysis (table 1).

Table 1 Outcome measures


To test for differences between baseline, end of treatment and the long-term follow-up measures, Wilcoxon signed-rank tests were performed (p<0.5) using SPSS version 13.0 (SPSS, Chicago, Illinois, USA).


The study was approved by The Danish Medicine Agency and The Regional Scientific Ethical Committee no. VF20030212, and written informed consent was obtained from all participants prior to participation.


Of the 166 patients in the previous RCT, 37 met the inclusion criteria for the current antibiotic trial (a further six patients had Modic changes but no LBP at 14 months review). The distribution of patients from the two treatment groups in the previous RCT was similar (16% and 18%). Five patients were excluded from the current antibiotic trial, due to: not wanting to take antibiotics for such a long period (n = 3), spontaneous recovery in the waiting period before the study started (n = 1), and loss of contact (n = 1). Therefore, 32 patients enrolled in the study but three dropped out due to severe diarrhoea, leaving 29 patients who completed the antibiotic treatment regime.

Of the 29 patients who completed the antibiotic treatment, 10 (34%) were female, mean age 45.7 (SD 11.1) years and 19 (66%) were male, mean age 47.7 (8.2) years. The original disc lesions were as follows: four (14%) patients had a bulge, 12 (41%) a focal protrusion, five (17%) a broad-based protrusion, seven (24%) an extrusion, and one (3%) a sequestrated disc, according to the Fardon and Milette nomenclature.15

The outcome measurements at baseline, at end of treatment, and at follow-up are described in table 2.

Table 2 The scores of the outcome measures; level of LPB, days with LBP, and physical function

All outcome measures (disease-specific function, patient-specific function, global perceived health, and LBP) showed statistically significant improvements both at the end of the treatment period and at the long-term follow-up. Using the measure of global perceived health, at the end of treatment, 15 (52%) of the patients reported that they were much better or cured, seven (24%) were moderately better, and seven (24%) reported they were unchanged. None experienced a worsening of symptoms.

The patients’ improvement of disease-specific function was measured using the Roland Morris Questionnaire (RMQ). A reduction in RMQ score is regarded as clinically important if it exceeds 30% of the patient’s original score.16 In our cohort, approximately two-thirds of the patients reduced their RMQ scores more than 30% (table 3). Furthermore, a 30% reduction in the average baseline RMQ score of this cohort would have been 2.6 RMQ points and the observed average reduction of 3.1 RMQ points exceeded that estimate.

Table 3 The percentage of patients whose Roland Morris Questionnaire (RMQ) scores were clinically improved, unchanged or worse

The clinical importance of this improvement in function is reinforced by the observation that, prior to commencement of antibiotic treatment, the RMQ scores of people in this cohort were worsening over time. Modic changes were first observed in participants when they were reviewed at the 14 month follow-up of the previous RCT. At this time the patients had a mean RMQ of 5.3. Patients then waited for varying time periods for the commencement of the current antibiotic study. The mean waiting period was 9 months and during this period the average functional level measured on this scale worsened from a mean of 5.3 to 8.9. However, the mean functional level at the end of the antibiotic treatment and at long-term follow-up showed both clinically important and statistically significant improvement (fig 1).

Figure 1 Each line represents the Roland Morris Questionnaire (RMQ) scores for each patient measured at four different points in time: the first point when eligibility for inclusion in the antibiotic trial was identified, the second point at the start of antibiotic treatment, the third point at the end of antibiotic treatment, and the fourth point at the follow-up (mean 10.8 months) after antibiotic treatment.The thick black line from end of antibiotic treatment to long-term follow-up represents many patients with the value 0.

The serum analysis revealed that only a few patients exceeded the reference values, indicating that no aggressive infection was present. There were only modest changes from baseline to the end of treatment. Unfortunately, during the study period the laboratory changed the methods of analysis for lactate dehydrogenase and also alkaline phosphatase; therefore values of the first method are presented for the respective number of patients, likewise with the second method (table 4).

Table 4 The serum analyses at baseline and end of treatment


Modic changes (bone oedema) observed on MRI are associated with LBP,1 2 17 18 but the pathogenesis of Modic changes remains unclear. Both mechanical stress and infection secondary to disc herniation have been hypothesised as pathological pathways.6 In this pilot study of the infection hypothesis, antibiotic treatment appeared to be effective in patients with LBP and Modic changes Type 1. We included patients whose only known illness was a previous disc herniation with present Modic changes and LBP. Patients’ health status improved on all outcome measures: global perceived health, disease-specific and patient-specific disability, pain in the lumbar area and number of days with pain.

It is often reported that LBP fluctuates, and, by nature, patients normally seek medical assistance when the pain is at its worst. Therefore, the positive results observed in many cohort studies may be due to spontaneous recovery and/or regression towards the mean. In order to minimise this, we only recruited patients at a prescheduled follow-up examination 14 months after experiencing a herniated lumbar disc. No patients were recruited when seeking medical assistance for an acute exacerbation of their condition, and all patients were observed for a further mean of 9 months before the antibiotic treatment was initiated. During this period only one patient experienced a spontaneous recovery. Furthermore, the observation that, on average, patients experienced a worsening of their condition during the 9 months observation period but demonstrated improvement after antibiotic treatment would suggest spontaneous recovery was highly unlikely.

The improvement could be a result of the natural cause of Modic changes. Unfortunately the knowledge about this is scarce. A few longitudinal studies 2 4 17 19 describe over a period of 14–72 months that 48% to 86% of the Modic changes remain stable and 14% to 52% convert into another type of Modic change. This very slow natural cause makes it is less likely that the changes in this study are attributed to the natural cause.

The positive results of the antibiotic treatment support the hypothesis that some Modic changes may result from a low-virulent bacterial infection.6 Modic identified five patients in 1984 with suspected infection in the disc and in the vertebral end plates.8 They were treated with antibiotics for 4 weeks and rescanned. A change in the signal in the central area within the disc was observed suggesting healing and degeneration. Stirling et al10 cultured nuclear tissue removed under sterile conditions during surgery for lumbar herniated discs. Of these, 53% were infected with low-virulent anaerobic organisms (Propionibacterium acnes and Corynebacterium propinquum) in contrast to none of those patients who underwent surgery for other spinal disorders. These bacteria are found in all individuals on their skin and in their oral cavities. They frequently invade the circulatory system (for example during tooth brushing), where they do not present any health risks due to the aerobic environment in the bloodstream.2023 They can, however, impose a health risk in patients with a special environment, for example, in individuals with a prosthetic device.

In a herniated disc, nuclear material may migrate into the spinal canal. Within a short time, new capillarisation takes place in and around the extruded nuclear material2427 and inflammation occurs with an increased presence of macrophages.2628 In this particular environment we hypothesise that anaerobic bacteria may enter the disc, resulting in a low-virulent and slowly developing infection. As intervertebral discs are avascular, they are an ideal environment for the growth of anaerobic bacteria and local inflammation in adjacent bone may result due to production of cytokines. Due to the low virulence of these bacteria, tissue reactions are slow and therefore poorly illuminated on MRI. Also, the infection will not spread to aerobic tissues, which may explain why Modic changes are usually observed in close proximity to the site of a disc herniation. Adding credibility to this pathoanatomical model is the observation that patients with a normal disc contour do not develop Modic changes and that larger disc lesions result in more frequent Modic changes.2

Many antibiotics, especially tetracycline derivates, have an anti-inflammatory effect, via TNFα inhibition, and are therefore able to reduce pain. The positive effect observed in the current trial could be attributed to the anti-inflammatory effect rather than to an antibiotic effect. However, Amoxicillin-clavulanate (Spektramox) was chosen in part due to its extremely low anti-inflammatory effect29 30 and therefore we believe that the antibiotic effect is more likely to be responsible for the positive outcomes observed in this study.

We acknowledge that bacteria may not be the only cause of Modic changes, and that Modic changes may be the end stage of a number of pathological pathways. For example, in some cases there may be a biomechanical aspect, where a degenerated disc with diminished resistance to shear forces may result in Modic changes. However, the results of the current study tentatively suggest that antibiotic treatment could benefit some patients suffering from LBP and Modic changes. Clearly, this study is an observation of a highly selected patient group, and these results need to be tested in randomised controlled trials using antibiotics and placebo treatment. Further insight could be gained from biopsy studies of discs to identify any bacteria associated with Modic changes and provide greater precision in determining ideal antibiotic treatment.


  • Competing interests: None.