BMJ

Confidential: For Review Only Efficacy of antibiotic treatment in patients with chronic low back pain and Modic changes (the AIM study): a double- blind, randomised, placebo-controlled, multicentre trial

Journal: BMJ

Manuscript ID BMJ-2019-050262.R1

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the 29-Jun-2019 Author:

Complete List of Authors: Bråten, Lars Christian; Oslo University Hospital Ullevaal, FORMI; University of Oslo Faculty of Medicine Rolfsen, Mads Peder; Oslo University Hospital, Department of Orthopaedic surgery; University of Oslo Faculty of Medicine Espeland, Ansgar; Haukeland University Hospital, Department of Radiology; University of Bergen, Dep. of Clinical Medicine Wigemyr, Monica ; Oslo University Hospital Ullevaal, FORMI Assmus, Jörg; Haukeland University Hospital, Competence Center for Clinical Research Froholdt, Anne; Vestre Viken Hospital Trust, Dep. of Physical Medicine and Rehabilitation Haugen, Anne; Østfold Hospital Trust, Dep. of Rheumatology Marchand, Gunn Hege; Saint Olavs Hospital University Hospital in Trondheim, Dep. of Physical Health and Rehabilitation; Norwegian University of Science and Technology, Department of Neuromedicine and Movement Science Kristoffersen, Per Martin; Haukeland University Hospital, Department of Radiology; University of Bergen, Dep. of Clinical Medicine Lutro, Olav; Stavanger University Hospital Randen, Sigrun; Vestre Viken Hospital Trust, Dep. of Physical Medicine and Rehabilitation Wilhelmsen, Maja; University Hospital of North Norway, Dep. of rehabilitation; UiT Arctic University of Norway, Dep. of Clinical Medicine Winsvold, Bendik; Oslo University Hospital, Department of Research, Innovation and Education Kadar, Thomas; Haukeland University Hospital, Department of Physical Medicine and Rehabilitation Holmgard, Thor Einar; Norwegian Back Pain Association Vigeland, Maria; Oslo University Hospital Ullevaal, FORMI Vetti, Nils; Haukeland University Hospital, Department of Radiology; University of Bergen, Dep. of Clinical Medicine Nygaard, Øystein; Saint Olavs Hospital University Hospital in Trondheim, Department of Neuroscience; Norwegian University of Science and Technology, Department of Neuroscience Lie, Benedicte; Oslo University Hospital, Department of Immunology Hellum, Christian; Oslo University Hospital, Department of Orthopaedic

https://mc.manuscriptcentral.com/bmj Page 1 of 101 BMJ

1 2 3 surgery 4 Anke, Audny; University Hospital of North Norway, Dep. of rehabilitation; 5 UiT Arctic University of Norway, Dep. of Clinical Medicine 6 Grotle, Margreth; Oslo Metropolitan University, Department of 7 physiotherapy; Oslo University Hospital Ullevaal, FORMI 8 Schistad, Elina; Oslo University Hospital, Department of Physical 9 Medicine and Rehabilitation Skouen, Jan; Haukeland University Hospital, Department of Physical 10 Confidential:Medicine and ForRehabilitation; Review University of Bergen, Only Department of Global 11 Public Health and Primary Care 12 Grøvle, Lars; Østfold Hospital Trust, Dep. of Rheumatology 13 Brox, Jens Ivar; Oslo University Hospital, Physical Medicine and 14 rehabilitation; University of Oslo Faculty of Medicine 15 Zwart, John-Anker; Oslo University Hospital, Department of Neurology 16 and FORMI; University of Oslo, Faculty of Medicine Storheim, Kjersti; FORMI, Clinic for Surgery and Neurology; Oslo 17 Metropolitan University, Department of physiotherapy 18 19 Keywords: back pain, Modic changes, antibiotic, randomised trial 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 2 of 101

1 2 3 1 Title page 4 5 6 2 7 8 9 3 Efficacy of antibiotic treatment in patients with chronic low back pain and Modic 10 4 changes (the AIM study): a double-blind, randomised, placebo-controlled, multicentre 11 5 trial 12 Confidential: For Review Only 6 13 7 Authors: 14 8 1. Lars Christian Haugli Bråten, MD (1.FORMI, Oslo University Hospital HF, Ulleval, Bygg 37b, P.O. Box 4956, Nydalen, 0424, 15 9 Oslo, Norway. 2.University of Oslo, Faculty of Medicine, P.O. Box 1072 Blindern, 0316 Oslo, Norway). 16 10 [email protected] 11 2. Mads Peder Rolfsen, MD (1.Department of orthopaedic surgery, Oslo University Hospital, Postbox 4956, Nydalen, 0424, Oslo, 17 12 Norway. 2.University of Oslo, Faculty of Medicine, P.O. Box 1072 Blindern, 0316 Oslo, Norway). [email protected] 18 13 3. Ansgar Espeland, prof (1.Department of Radiology, Haukeland University Hospital, Jonas Liesvei 65, 5021 Bergen, Norway. 2. 19 14 Department of Clinical Medicine, University of Bergen, P.O. Box 7804, 5020 Bergen, Norway). [email protected] 20 15 4. Monica Wigemyr, MSc (FORMI, Oslo University Hospital HF, Ulleval, Bygg 37b, P.O. Box 4956, Nydalen, 0424, Oslo, 16 Norway). [email protected] 21 17 5. Jörg Aßmus, PhD (Centre for Clinical Research, Haukeland University Hospital, Jonas Liesvei 65, 5021 Bergen, Norway). 22 18 [email protected] 23 19 6. Anne Froholdt, PhD (Department of Physical Medicine and Rehabilitation, Drammen Hospital, Vestre Viken Hospital Trust 20 Drammen, P.O. Box 800, 3004 Drammen, Norway). [email protected] 24 21 7. Anne Julsrud Haugen, PhD (Department of Rheumatology, Østfold Hospital Trust, Grålum, Norway. Postal address: P.O. Box 25 22 300, 1714 Grålum, Norway). [email protected] 26 23 8. Gunn Hege Marchand, PhD (1.Department of Physical Medicine and Rehabilitation, St. Olavs Hospital, Trondheim University 24 Hospital, P.O. Box 3250 Torgarden, NO-7006 Trondheim, Norway. 2. Department of Neuromedicine and Movement Science, 27 25 Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology, Norway). 28 26 [email protected] 29 27 9. Per Martin Kristoffersen, MD (1.Department of Radiology, Haukeland University Hospital, Jonas Liesvei 65, 5021 Bergen, 28 Norway. 2. Department of Clinical Medicine, University of Bergen, P.O. Box 7804, 5020 Bergen, Norway). 30 29 [email protected] 31 30 10. Olav Lutro, MD (Stavanger University Hospital, Helse Stavanger HF, Postboks 8100, 4068 Stavanger, Norway). 32 31 [email protected] 33 32 11. Sigrun Randen, MD (Department of Physical Medicine and Rehabilitation, Drammen Hospital, Vestre Viken Hospital Trust 33 Drammen, P.O. Box 800, 3004 Drammen, Norway). [email protected] 34 34 12. Maja Wilhelmsen, PhD (1.Department of Rehabilitation, University Hospital of North Norway, P.O. Box 100, 9038 Tromsø, 35 35 Norway. 2.Faculty of Health Sciences, Department of Clinical Medicine, UiT The Arctic University of Norway, Tromsø, 36 36 Norway). [email protected] 37 13. Bendik Slagsvold Winsvold, PhD (Department of Research, Innovation and Education, Division of Clinical Neuroscience, Oslo 37 38 University Hospital, P.O. Box 4956, Nydalen, 0424, Oslo, Norway). [email protected] 38 39 14. Thomas Istvan Kadar, PhD (Department of Physical Medicine and Rehabilitation, Haukeland University Hospital, Helse Bergen 39 40 HF, P.O. Box 1, 5021 Bergen, Norway). [email protected] 41 15. Thor Einar Holmgard, patient representative (Norwegian Back Pain Association, P.O. Box 9612 Fjellhagen, 3065 Drammen, 40 42 Norway). [email protected] 41 43 16. Maria Dehli Vigeland, MSc (1.FORMI, Oslo University Hospital HF, Ulleval, Bygg 37b, P.O. Box 4956, Nydalen, 0424, Oslo, 42 44 Norway. 2.University of Oslo, Faculty of Medicine, P.O. Box 1072 Blindern, 0316 Oslo, Norway). [email protected] 43 45 17. Nils Vetti, PhD (1.Department of Radiology, Haukeland University Hospital, Jonas Liesvei 65, 5021 Bergen, Norway. 2. 46 Department of Clinical Medicine, University of Bergen, P.O. Box 7804, 5020 Bergen, Norway). [email protected] 44 47 18. Øystein Petter Nygaard, PhD (1.Department of neurosurgery, St. Olavs University Hospital, P.O. Box 3250 Torgarden, NO-7006 45 48 Trondheim, Norway 2. Department of Neuroscience, Norwegian University of Science and Technology (NTNU), Trondheim, 46 49 Norway 3. National Advisory Unit on Spinal Surgery, St. Olavs Hospital, Trondheim, Norway). [email protected] 50 19. Benedicte Alexandra Lie, prof (Department of Medical Genetics, University of Oslo and Oslo University Hospital, P.O. Box 47 51 4956, Nydalen, 0424, Oslo, Norway). [email protected] 48 52 20. Christian Hellum, PhD (Department of orthopaedic surgery, Oslo University Hospital, P.O. Box 4956, Nydalen, 0424, Oslo, 49 53 Norway). [email protected] 54 21. Audny Anke, prof (1.Department of Rehabilitation, University Hospital of North Norway, P.O. Box 100, 9038 Tromsø, Norway 50 55 2. Faculty of Health Sciences, Department of Clinical Medicine, UiT The Arctic University of Norway, Tromsø, Norway). 51 56 [email protected] 52 57 22. Margreth Grotle, prof (1.Oslo Metropolitan University, Department of physiotherapy, P.O. Box 4 St. Olavs plass, NO-0130 Oslo, 58 Norway 2. FORMI, Oslo University Hospital HF, Ulleval, Bygg 37b, P.O. Box 4956, Nydalen, 0424, Oslo, Norway 3. Keele 53 59 University, Visiting Professor). [email protected] 54 60 23. Elina Iordanova Schistad, PhD (Department of Physical Medicine and Rehabilitation, Oslo University Hospital HF, Ulleval, P.O. 55 61 Box 4956, Nydalen, 0424, Oslo, Norway). [email protected] 56 62 24. Jan Sture Skouen, prof (1.Department of Physical Medicine and Rehabilitation, Haukeland University Hospital, Helse Bergen HF, 63 Box 1, 5021 Bergen, Norway 2.University of Bergen, Department of Global Public Health and Primary Care, Kalfarveien 31, 57 64 5018 Bergen, Norway). [email protected] 58 65 25. Lars Grøvle, PhD (Department of Rheumatology, Østfold Hospital Trust, Grålum, Norway. Postal adress: P.O. Box 300, 1714 59 66 Grålum, Norway). [email protected] 60

1 https://mc.manuscriptcentral.com/bmj Page 3 of 101 BMJ

1 2 3 67 26. Jens Ivar Brox, prof (1.Department of Physical Medicine and Rehabilitation, Oslo University Hospital HF, Ulleval, P.O. Box 4 68 4956, Nydalen, 0424, Oslo, Norway 2.University of Oslo, Faculty of Medicine, P.O. Box 1072 Blindern, 0316 Oslo, Norway). 69 [email protected] 5 70 27. John-Anker Zwart, prof (1.FORMI, Oslo University Hospital HF, Ulleval, Bygg 37b, P.O. Box 4956, Nydalen, 0424, Oslo, 6 71 Norway 2.University of Oslo, Faculty of Medicine, P.O. Box 1072 Blindern, 0316 Oslo, Norway). [email protected] 7 72 28. Kjersti Storheim, prof (1.FORMI, Oslo University Hospital HF, Ulleval, Bygg 37b, P.O. Box 4956, Nydalen, 0424, Oslo, Norway 73 2.Oslo Metropolitan University, Department of physiotherapy, PO box 4 St. Olavs plass, NO-0130 Oslo, Norway). 8 74 [email protected] 9 75 29. The AIM-study group (see list of authors at page 7) 10 76 11 77 Correspondence to: dr. Lars Christian Bråten, FORMI, Oslo University Hospital HF, Ullevål, Bygg37b, P.O. Box 12 78 4956, Nydalen,Confidential: 0424, Oslo, Norway. [email protected] For Review. Phone +47 99299034 Only 13 79 14 80 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 https://mc.manuscriptcentral.com/bmj BMJ Page 4 of 101

1 2 3 81 Abstract 4 82 OBJECTIVE: To assess the efficacy of three months of antibiotic treatment compared to placebo in patients with 5 83 chronic low back pain (LBP), prior disc herniation, and vertebral endplate changes (Modic changes (MCs)). 6 84 7 85 DESIGN: Double blind, parallel group, placebo controlled, multicentre trial. 8 86 9 87 SETTING: Hospital outpatient clinics at six hospitals in Norway. 10 88 11 89 PARTICIPANTS: 180 patients with chronic LBP, previous disc herniation, and type I (n=118) or type II (n=62) 12 90 MCs enrolledConfidential: from June 2015 to September 2017. For Review Only 13 91 14 92 INTERVENTIONS: Patients were randomised to three months of oral treatment with either 750 mg amoxicillin 15 93 or placebo three times daily. The allocation sequence was concealed by using a computer-generated number on the prescription. 16 94 95 17 96 MAIN OUTCOME MEASURES: The primary outcome was the Roland-Morris Disability Questionnaire 18 97 (RMDQ) score (range 0-24) at 1-year follow-up in the intention-to-treat population. We predefined the minimal 19 98 clinically important between-group difference in mean RMDQ score as 4. 20 99 21 100 RESULTS: In the primary analysis of the total cohort at 1 year, the difference in the mean RMDQ score between 22 101 the amoxicillin and the placebo group was -1·6 (95% CI, -3·1 to 0·0; P=0·04). In the secondary analysis, the 23 102 difference in the mean RMDQ score between the amoxicillin and the placebo group was -2·3 (95% CI, -4·2 to - 24 103 0·4; P=0·02) in the type I MC group, and -0·1 (95% CI, -2·7 to 2·6; P=0·95) in the type II MC group. Fifty 25 104 patients (56%) in the amoxicillin group experienced ≥1 drug-related adverse events vs 31 (34%) in the placebo 26 105 group. 27 106 28 107 CONCLUSIONS: Amoxicillin did not provide a clinically important benefit and our results do not support the 29 108 use of antibiotic treatment for chronic LBP and MCs. 30 109 31 110 TRIAL REGISTRATION: 32 111 ClinicalTrials.gov NCT02323412 33 34 112 Introduction 35 113 The current management of low back pain (LBP) offers at best low to moderate improvement in pain and disability.1 Researchers therefore attempt to identify subgroups with effect of specific treatment. A suggested 36 114 115 subgroup of chronic LBP are those patients with signal changes in the vertebral bone marrow extending from the 37 116 endplate (Modic changes (MCs)) on magnetic resonance imaging (MRI).2 MCs are classified into type I (oedema 38 117 type), type II (fatty type), and type III (sclerotic type, less common).3 The pathogenesis of MCs is unclear. One 39 118 hypothesis is that MCs and LBP are caused by low-grade bacterial discitis caused by Cutibacterium acnes (C. 40 119 acnes) (formerly known as Propionibacterium acnes), an aerotolerant gram positive anaerobe bacteria and a 41 120 common skin-commensal.4 The proposed port of entry into the disc is via the bloodstream, made possible during 42 121 vascularization related to inflammation caused by a disc herniation.5 Animal models show that C acnes could 43 122 grow in degenerated discs and cause MCs.6,7 However, several microbiological studies of disc biopsies have 44 123 conflicting results.4 45 124 46 125 A systematic review found only one randomized trial assessing the efficacy of antibiotic treatment in patients 47 126 with low back pain.4 That randomized trial reported a substantial effect of three months of antibiotic treatment 48 127 over placebo (between-group difference of 8·3 points on the Roland-Morris Disability Questionnaire (RMDQ) at 49 128 1-year follow-up) in patients with chronic LBP, prior disc herniation, and type I MCs.8 The trial conclusion has 50 129 been questioned based on almost no improvement in the control group9, no evaluation of blinding efficacy9, and 51 130 a high proportion of participants with previous disc surgery and thus a potential risk of bacterial contamination.10 52 131 Current guidelines do not recommend (for or against) antibiotic treatment in patients with persistent LBP and 53 132 MCs.11 About 40-50% of patients with non-specific LBP have MCs, and antibiotic treatment in subgroups of this 54 133 large patient populationmay increase antibiotic resistance.12-14 55 134 56 135 We set out to see if we could replicate the findings in the former randomized trial. The present trial was designed 57 136 as a randomized, double-blind, placebo-controlled trial to evaluate the efficacy and harm of three months of oral 58 137 treatment with amoxicillin at 1-year follow up in patients with chronic LBP and type I or II MCs at the level of a 59 138 previous lumbar disc herniation. 60 139

3 https://mc.manuscriptcentral.com/bmj Page 5 of 101 BMJ

1 2 3 140 Methods 4 141 5 6 142 Study design 7 143 This was a multicentre, randomised, double-blind, placebo-controlled, parallel-group trial with a treatment phase 8 144 (three months) and a follow-up phase (nine months). 9 145 10 146 The trial was performed in accordance with the Helsinki Declarationand the ICH-GCP (Good Clinical Practice), 15 11 147 reported according to the CONSORT guidelines, and registered at ClinicalTrials.gov in December 2014 under 12 148 the identifier:Confidential: NCT02323412. The trial was approved For by the ReviewRegional Committees Onlyfor Medical Research Ethics - South East Norway (2014/158/REK sør-øst) and the Norwegian Medicines Agency (SLV; reference no 13 149 150 14/01368-11; EUDRACT NO 2013-004505-14) before commencing, and monitored by the Clinical Trial Unit at 14 151 Oslo University Hospital. All participants gave written informed consent. Methods were unchanged after trial 15 152 commencement. Funding was granted by a governmental organisations (Helse Sør-Øst and Helse Vest), which 16 153 had no part in the planning, performing, or reporting of the trial. The trial protocol is available at 17 154 www.clinicaltrials.gov. 18 155 19 156 Participants 20 157 Patients were recruited from outpatient clinics at 6 hospitals in Norway from June 2015 to September 2017. 21 158 Inclusion criteria were age 18 to 65 years, LBP for > 6 months with intensity  5 on a 0-10 Numerical Rating 22 159 Scale (NRS, mean of: current, worst within the preceding two weeks, and usual/mean within the preceding two 23 160 weeks), lumbar disc herniation on MRI in the preceding two years, and type I and/or type II MCs (with height ≥ 24 161 10% of vertebral height and diameter > 5 mm) at the herniated disc level. In contrast to the trial we set out to 25 162 reassess, we chose to also include patients with type II MCs, as differentiating between type I and type II MCs is 26 163 of uncertain relevance and may depend on the magnetic field strength of the MRI scanner used.16,17 Surgery for 27 164 disc herniation in the last year and antibiotic treatment in the last month led to exclusion. Figure S1 in the 28 165 Appendix contains all eligibility criteria.18 29 166 30 31 167 Randomisation, masking and procedures 32 168 Patients were randomized at median 13 days after inclusion into either three months oral treatment with 33 169 amoxicillin 750 mg three times daily or placebo (maize starch). The tablets were encapsulated (Kragerø Tablettproduksjon AS), with identical capsules, containers, and labels for both treatment groups. Randomisation 34 170 lists were created by a third party statistician using STATA 13 (StataCorp LP, College Station, TX, USA) and 35 171 172 stratified by MC type (I/II) and previous disc herniation surgery or not, with a 1:1:1:1 allocation and random 36 173 block sizes of 4 and 6. This ensured that a similar number of patients received antibiotics or placebo within each 37 174 stratum. The allocation sequence was concealed. Care providers gave each patient a prescription with a computer 38 175 generated allocation number to be used at the dedicated hospital pharmacies. All care providers, research staff, 39 176 statisticians, and patients were unaware of the assignment group during the data collection. Care providers, 40 177 research staff and statisticians were also blinded during primary and secondary analyses and first draft of this 41 178 manuscript. Included patients were recommended not to start additional treatments for back pain but were 42 179 allowed to continue ongoing therapy. They were encouraged not to use NSAIDs during the intervention period. 43 180 44 181 To confirm MCs seen on a clinical MRI available at screening, baseline MRI was performed at median 22 45 182 (interquartile range 15 to 30) days before treatment initiation using identical 1.5 T protocols at each study centre. 46 183 Patients with primary (most extensive) or secondary type I MCs (hypointense on T1 images, hyperintens on T2) 47 184 were allocated to the type I MC group. Patients with type II MCs (hyperintense on T1, iso- or hyperintense on 48 185 T2) but not type I MCs were included in the type II MC group. Borderline type I vs type II MCs (near iso-intense 49 186 on T1) were rated as type II. Two experienced radiologists independently evaluated MCs on baseline MRI, disc 50 187 herniation on MRI from the previous two years, and eligibility for the trial in the type I vs type II MC group vs 51 188 ineligible (kappa = 0·62). They discussed and solved all disagreements on eligibility. 52 189 53 Outcomes and Data Collection 54 190 The primary outcome was the score on the validated Norwegian version of the RMDQ at 1-year follow-up. 55 191 192 RMDQ scores range from 0 to 24. Higher scores indicate more severe pain and disability. Secondary outcomes 56 193 included pain related disability (Oswestry Disability Index (ODI) 2·0), LBP intensity (0-10 NRS), and health- 57 194 related quality of life (EQ5D-5L, version 2·0). Trial measurements were unchanged after trial start and were 58 195 reported at several time points (Appendix, Table S1) using a web-based data capture system (Viedoc™) or, in a 59 196 few cases, using a paper version. Trial care providers (physicians or physiotherapists) performed active 60 197 surveillance of side effects/adverse events (clinical and biochemical) from baseline to 1 year using CTCAE v4.0 4 https://mc.manuscriptcentral.com/bmj BMJ Page 6 of 101

1 2 3 198 in accordance with medDRA-coding.19 Adverse events were monitored and cross-checked against clinical 4 199 patient notes, and all are reported. Compliance was monitored by weekly patient reported questionnaires and by 5 200 counting the returned study medication capsules at the end-of-treatment visit (Table S1 in the Supplementary 6 201 Appendix) 7 202 8 9 203 Statistical Analysis 20 10 204 We specified all main statistical analyses in a Statistical Analysis Plan in advance of database locking. Post-hoc 11 205 analyses not described in the Statistical Analysis Plan are marked in the manuscript and the tables. 12 206 Confidential: For Review Only In the primary analysis we compared mean RMDQ scores between the two treatment groups in the whole 13 207 208 intention-to-treat population at 1 year using ANCOVA, adjusted for baseline RMDQ score and the stratification 14 209 variables. Missing RMDQ values were imputed with a multiple imputation model.20 In the secondary analyses, 15 210 we repeated this comparison in each MC type group. We also performed analyses of key supportive objectives 16 211 (secondary outcomes: ODI, LBP intensity, and EQ5D), per-protocol analysis, and other sensitivity analyses, 17 212 responder analyses (>30%, >50%, and >75% improvement from baseline, excluding patients with >30% 18 213 missing RMDQ items), and Linear Mixed-Effects (LME) models as described in the Statistical Analysis Plan.20 19 214 An exploratory objective was to report the incidence of adverse events occurring from randomisation to 1-year 20 215 follow-up. Adverse events were presented descriptively for the safety population according to intervention 21 216 group, without statistical testing. 22 217 23 218 We predefined a clinically important between-group mean difference of 4 RMDQ points. We considered smaller 24 219 differences not clinically relevant in this trial (despite some have regarded a difference of 2·5 as relevant21), 25 220 since we were reassessing the potential curative effect of antibiotic treatment, and a much higher difference (8·3) 26 221 was reported in the former trial8. In comparison, a change of 2-3 RMDQ points in individual patients over time 27 222 may represent measurement error22 and the minimal detectable change of the RMDQ was 4 in a Norwegian 28 223 study of chronic LBP patients23. In each MC type group, we needed 66 analysed patients to detect (β=0·1, 29 224 α=0·05) a difference of 4 (SD 5) in mean RMDQ score between the two treatment groups. In the total sample, 30 225 132 analysed patients allowed detection (β=0·1, α=0·05) of a difference of 2.8 (SD 5) in mean RMDQ score. We 31 226 added 20% to allow for drop-outs and planned to include 80 subjects in each group. 32 227 33 228 When half of the study participants had completed their 1-year assessment, an independent statistician blinded to 34 229 treatment groups performed a pre-specified interim analysis to ensure that continuation of the study was ethical. 35 230 If mean RMDQ score differed >7 points between the two treatment groups, an independent Data Monitoring Committee could recommend termination of the study. No results from the interim analysis were communicated 36 231 232 to the trial staff. We did not adjust the significance level due to the interim analysis as the trial did not have a 37 233 group-sequential design and only one primary analysis. 38 234 39 235 We calculated Bang blinding index in each treatment group based on patients’ reports at 1-year follow-up 40 236 regarding which study medicine they thought they had received (antibiotics, placebo, or unsure). Bang blinding 41 237 index ranges from -1 (all patients report incorrect treatment) to 1 (all patients report correct treatment); 0 42 238 indicates random reporting of treatment group.24 43 239 44 240 Primary, secondary, and key supportive objectives analyses were carried out independently by a senior 45 241 statistician using software package R version 3.4.4 and a PhD student using software package Stata version 15, 46 242 both blinded to treatment group. 47 243 48 49 244 Patient involvement 50 51 245 A patient representative was a member of the Scientific Board of the study, which effected all the major decision 52 246 from planning and design of the study, to the dissemination of the study results. The patient representative assessed the burden of the study medication and the time and efforts required to participate in the trial, and took 53 247 248 part in writing this manuscript. 54 249 55 56 250 Results 57 251 58 59 60

5 https://mc.manuscriptcentral.com/bmj Page 7 of 101 BMJ

1 2 3 252 Patients 4 253 The trial-group assignments, loss to follow-up, treatment completion, and protocol deviations are summarized in 5 254 Figure 1. Of the 580 patients assessed for eligibility, 180 underwent randomisation, 118 with type I MCs and 62 6 255 with type II MCs. As inclusion of type I MCs was faster, enrolment was closed before the goal for type II MCs 7 256 patients had been met (in September 2017). We prospectively presumed at the time of enrolment closure that a 8 257 power of 88% (using the same specifications as above for 62 patients) for the type II MC group was sufficient. 9 258 One-year follow-up data were obtained by the end of September 2018. 10 259 11 260 Table 1 shows baseline characteristics; baseline values of the outcomes are included in Table 2. The placebo 12 261 group had Confidential:a somewhat higher percentage of participants For with Review jobs requiring physical Only workload. We did not 13 262 consider there to be any other relevant differences between the treatment groups. In total, 38 patients (21%) had 14 263 former surgery for disc herniation. All 180 randomized patients started on the study medication and were 15 264 included in the intention-to-treat analysis. Of these, 155 patients were included in the per-protocol analysis. 16 265 Outcomes 17 266 At 1 year, the mean RMDQ score was reduced since baseline in both treatment groups (-3·7 points in the 18 267 amoxicillin group and -2·1 points in the placebo group) (Table 2 and Figure 2). The adjusted mean difference in 19 268 RMDQ score between the amoxicillin group and the placebo group at 1 year was -1·6 points (95% CI, -3·1 to 20 269 0·0; P=0·044) (Table 2). The adjusted between-group difference of the mean RMDQ score in patients with type I 21 270 MCs was -2·3 (95% CI, -4·2 to -0·4; P=0·02), and -0·1 (95% CI, -2·7 to 2·6; P=0·95) for patients with type II 22 271 MCs. Analyses of the secondary outcomes are summarized in Table 2 and Table 3. The responder analyses 23 272 (Table S2), and sensitivity analyses (Table S5) are summarized in the Supplementary Appendix. 24 25 273 Adverse Events 26 274 In the amoxicillin group, 50 patients (56%) had ≥1 drug-related adverse events (possibly/probably/definitely 27 275 related to study medication) compared to 31 patients (34%) in the placebo group. One or more serious adverse 28 276 events occurred in 6 patients (7%) in the amoxicillin group and 2 patients (2%) in the placebo group, none were 29 277 related to the study medication (Table 4). In the amoxicillin group, 11 patients (12%) discontinued or paused the 30 278 study medication due to adverse events compared to 2 patients (2%) in the placebo group. No deaths occurred 31 279 during the trial. 32 280 Blinding 33 281 167 patients responded to the blinding question at 1 year. The Bang blinding index was -0·16 (95% CI -0·31 to – 34 282 0·01; P=0·96) in the amoxicillin group and 0·52 (95% CI 0·40 to 0·64; P<0·001) in the placebo group (Table S3 35 283 and Figure S4 in the Supplementary Appendix). 36 284 37 38 285 Discussion 39 286 In this study on patients with chronic LBP and MCs at the level of a previous disc herniation, three months of 40 287 treatment with amoxicillin did not provide a clinically important benefit compared to placebo. Secondary 41 288 analyses and sensitivity analyses supported this finding. 42 289 The analyses of secondary outcomes also found considerably smaller differences than the recommended 43 290 thresholds for a clinically important change within groups (ranges of recommended thresholds: ODI 13-20; LBP 44 291 intensity NRS 2-3; EQ5D 0·11-0·30).27-29 Fifty patients (56%) in the amoxicillin group vs 31 (34%) in the 45 292 placebo group experienced ≥1 drug-related adverse events. 46 293 47 294 Our results contrast with those of the trial we are reassessing, which showed a substantial effect (8·3 points 48 295 between-group difference on the RMDQ-score) of 3-month combination therapy of amoxicillin and clavulanic 8 49 296 acid. In the present study, amoxicillin was used without clavulanic acid. Since there is little or no resistance to penicillin among C. acnes in vitro30, and clavulanic acid is known to penetrate discs poorly31,32, we consider the 50 297 298 difference in treatment regimens to be an unlikely explanation for the discrepant trial results. The trial we are 51 299 reassessing reported a non-significant difference in effect between amoxicillin/clavulanic acid 500/125 mg vs 52 300 1000/250 mg three times daily.8 However, the pharmacodynamic property predicting treatment success with 53 301 amoxicillin is time above the minimum inhibitory concentration. Further increases in amoxicillin concentration 54 302 do not increase bactericidal activity, suggesting there should be a minimal dose-response relationship. Thus, we 55 303 consider the dosage of amoxicillin used in the present trial, 750 mg three times daily, to be sufficient. The former 56 304 trial included a higher percentage of patients with former disc surgery (52% in the antibiotic group) compared to 57 305 the present study (20% in the antibiotic group) and had relatively fewer patients at follow up in the antibiotic 58 306 treatment group (86% vs 96% in the present trial). We cannot rule out that this may have influenced the 59 307 differences in results. The improvement in RMDQ score from baseline to 1-year follow up in the placebo group 60 308 was smaller in the prior trial (from 15.0 to 14.0 vs 12.8 to 10.7 in the present trial). It is difficult to evaluate

6 https://mc.manuscriptcentral.com/bmj BMJ Page 8 of 101

1 2 3 309 whether this may have been due to poorer blinding in the previous study, since data on blinding efficacy were 4 310 lacking. 5 311 6 312 7 313 Our predefined minimal clinically important between-group difference of 4 points on the RMDQ is somewhat 8 314 larger than that used in some other randomised trials of patients with LBP. However, it can be considered 9 315 conservative given the results of the trial we were reassessing and the proposed rationale for the treatment that an 10 316 infection leads to Modic changes and LBP. If the symptoms were mainly due to an infection with C. acnes, we 11 317 should expect a large symptom improvement with effective antibiotic treatment. Perhaps also the nature of the 12 318 intervention,Confidential: with its associated risk of complications For and contribution Review to antibiotic Only resistance, suggest we should 13 319 demand a larger effect size than for many other suggested treatment options for LBP. Patients require a 30% 33 14 320 extra improvement if taking NSAIDs compared to no intervention, which equates approximately to our 15 321 predefined level of difference. We observed a small, statistically significant treatment effect for the primary outcome, two secondary outcomes (ODI and EQ5D), and among patients with type I MCs, all in favour of the 16 322 323 amoxicillin group. However, the study was overpowered for all these analyses. The statistically significant effect 17 324 sizes are small and most likely not clinically relevant for the whole study population. In addition, patients’ 18 325 expectation of treatment effect might partly explain the effect (Table S5). We cannot exclude a clinically 19 326 relevant effect for some patients with type I MCs as the 95% CI for RMDQ (-4·2 to -0·4) just included the 20 327 difference of 4 that the study was designed to detect. Nevertheless, the true difference is most likely < 4. The 21 328 results for the secondary outcomes in the type I MC group were similar and indicated no clinically relevant 22 329 between-group differences. 23 330 24 331 This study has some limitations. The main limitation was the initiation of antibiotic treatment without first 25 332 documenting infection in tissue samples taken from individual patients. Such microbiological investigations 26 333 were not performed due to risk of complications and suspected low diagnostic yield from samples with low- 27 334 grade infection obtained with needle biopsies.34 A second, related limitation is the heterogeneity of the treated 28 335 sample. We cannot exclude the possibility of treatment effect in subgroups. Third, a Bang blinding index of 0·52 29 336 (95% CI 0·40 to 0·64) in the placebo group indicates there was some degree of unblinding in this group, 30 337 potentially reducing the placebo effect in the placebo group, and causing a falsely high between-group treatment 31 338 effect. However, blinding seemed to be maintained in the amoxicillin group, reducing the impact of possible 32 339 biases. The main determinants of reporting antibiotic treatment were improvement of symptoms and adverse 33 340 events (Table S6, Supplementary Appendix). Finally, we cannot rule out the possibility that some patients in the 34 341 placebo group took amoxicillin without reporting to us, which could influence the difference in outcome 35 342 between the treatment groups. To reduce this problem we meticulously registered all medications. External 36 343 monitors cross-checked all registered medications with patients’ notes. 37 344 38 345 The inclusion- and exclusion criteria of this study, for example regarding herniation, pain score, age, and other diseases may reduce its generalizability. Particularly adverse events are more likely to occur in older patients.35 39 346 347 We are not aware of any biological mechanism that could make a treatment effect of amoxicillin more likely in 40 348 the excluded groups. 41 349 42 350 The proportion of patients with drug-related adverse events in this study was higher than reported in the 43 351 literature for long-term therapy with penicillins,36 and was high also in the placebo group (34%). Possible 44 352 explanations for this are that common complaints in the general population mimick adverse events, the 45 353 meticulous registration of adverse events in our trial, and some nocebo effects. Our finding that 12% in the 46 354 amoxicillin group vs 2% in the placebo group discontinued or paused study medication due to adverse events, 47 355 and that no patients experienced drug-related serious adverse events, is consistent with previous studies.36 48 356 49 357 The responder analyses suggest that between 5 and 18 patients need to be treated for one to improve; numbers 50 358 and statistical significance vary depending on cut-off values for definition of improvement. The nature of our 51 359 intervention suggests that we should require high cut-off values. It is our opinion that the numbers do not justify 52 360 three months of treatment with antibiotics, considering the increase in adverse events and the context of 53 361 increasing antibiotic resistance worldwide. In addition, differences between the treatment groups were not 54 362 consistent across all patient reported outcomes, further questioning the relevance of the small between-group 55 363 difference in the primary analysis. 56 364 57 365 In conclusion, we were not able to replicate the findings of the previous randomized trial. Our study did not 58 366 show any clinically important effect of three months oral antibiotic treatment in patients with chronic LBP, MCs, 59 367 and a former herniated disc. The present results do not support the use of antibiotic treatment for chronic LBP 60 368 and MCs.

7 https://mc.manuscriptcentral.com/bmj Page 9 of 101 BMJ

1 2 3 369 4 370 5 6 What is already known on this topic -A systematic review from 2015 (see main text) searched the MEDLINE and EMBASE, 7 included studies that examined the relationship between bacteria, back pain, and Modic changes, 8 and reported only one randomised, controlled trial which tested antibiotic treatment for patients 9 with chronic back pain and Modic changes. 10 -That trial reported a large treatment effect (difference between groups of 8·3 points on the 11 RMDQ). 12 What thisConfidential: study adds For Review Only 13 - Our largest observed difference between the treatment groups (2·3 points on the RMDQ) for 14 patients with type I MCs was substantially smaller and below the predefined minimal clinically 15 relevant difference. 16 -There is not sufficient evidence to support the use of antibiotic treatment for chronic low back 17 pain and Modic changes. 18 19 20 21 22 23 24 25 26 27 28 371 29 372 30 373 31 374 32 375 Contributors: 33 376 All authors critically revised the manuscript for important intellectual content. Trial care providers: MPR, AF, 34 377 AJH, GHM, SR, MWh, BSW, TK, AA and LCB. AE, PMK and NV evaluated MRIs. MW was the national 35 378 coordinator of the study and collected biological material. JA and LCB did the statistical analyses. AF, AJH, 36 379 ØPN, AA, JSS, LG, JIB were Principal Investigators and took part in planning the study with OL, THE, MG, 37 380 EIS, JAZ and KS. TEH was the patient representative. JAZ was the Coordinating Investigator and KS was the 38 381 project manager. LCB wrote the first draft with input from KS, and is guarantor. 39 382 The AIM-study group: 40 383 University Hospital North Norway, Tromsø (four patients):Terese Fors, Guro Kjos, Ida Beate Østhus 41 384 (department of rehabilitation). Trondheim University Hospital, Trondheim (21 patients): Britt Elin Lurud, Fredrik Granvigen (department of 42 385 physical medicine and rehabilitation), Hege Andersen (National Advisory Unit of Spinal Surgery), Vidar Rao 43 386 387 (department of neurosurgery). 44 388 Haukeland University Hospital, Bergen (37 patients): Siv Krüger Claussen, Erling Andersen (department of 45 389 clinical engineering). 46 390 Vestre Viken Hospital, Drammen (38 patients): Hilde Presberg (department of neurology). 47 391 Oslo University Hospital, Oslo (50 patients): Linda Margareth Pedersen, Karianne Wiger Gammelsrud 48 392 (department of microbiology), Siri Tennebø Flåm, Magnus Dehli Vigeland (department of medical genetics). 49 393 Østfold Hospital Trust (30 patients): Marianne Thorsø, Knut Morten Huneide, Veronica Sørensen (department of 50 394 physical medicine and rehabilitation). 51 395 52 396 Acknowledgement 53 397 We thank Helse Sør-Øst (grant no: 2015090) and Helse Vest (grant no: 911938 and 911891) for funding the 54 398 AIM-study, all patients participating in the study, and Eira Kathleen Ebbs for proofreading English language of 55 399 the manuscript. 56 400 57 401 Competing interests: 58 402 All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf. Kjersti 59 403 Storheim reports that funding was granted by governmental organisations (Helse Sør-Øst and Helse Vest) during 60 404 the conduct of the study. All other authors declare: no support from any organisation for the submitted work; no

8 https://mc.manuscriptcentral.com/bmj BMJ Page 10 of 101

1 2 3 405 financial relationships with any organisations that might have an interest in the submitted work in the previous 4 406 three years; no other relationships or activities that could appear to have influenced the submitted work. 5 407 6 408 Transparency 7 409 The lead author (LCB) affirms that the manuscript is an honest, accurate, and transparent account of the study 8 410 being reported; that no important aspects of the study have been omitted; and that any discrepancies from the 9 411 study as planned (and, if relevant, registered) have been explained. 10 412 11 413 Dissemination declaration 12 414 We plan toConfidential: disseminate the results to study participants For and theReview patient organisation Only (Norwegian Back Pain 13 415 Association) in advance of publication. 14 416 15 417 Data sharing statement Requests to access data should be addressed to [email protected]. De-identified individual 16 418 419 participant data (including data citionary) will be available to medical researchers by request in accordance with 17 420 local registration and ethical approval, when the article has been published until 1th of July, 2029. All proposals 18 421 requesting data access will need to specify an analysis plan and will need approval of the scientific board before 19 422 any data can be released. 20 423 21 424 I, Lars Christian Haugli Bråten, The Corresponding Author of this article contained within the original 22 425 manuscript which includes any diagrams & photographs within and any related or stand alone film submitted 23 426 (the Contribution”) has the right to grant on behalf of all authors and does grant on behalf of all authors, a 24 427 licence to the BMJ Publishing Group Ltd and its licencees, to permit this Contribution (if accepted) to be 25 428 published in the BMJ and any other BMJ Group products and to exploit all subsidiary rights, as set out in our 26 429 licence set out at: http://www.bmj.com/about-bmj/resources-authors/forms-policies-and-checklists/copyright- 27 430 open-access-and-permission-reuse. 28 431 29 432  I am the sole author of the Contribution. 30 433  I am one author signing on behalf of all co-owners of the Contribution. 31 434  The Contribution has been made in the course of my employment and I am signing as authorised 32 435 by my employer. 33 436  I am a US Federal Government employee acting in the course of my employment. 34 437  I am not a US Federal Government employee, but some or all of my co-authors are. 35 438  I am an employee of the UK Crown* acting in the course of my employment 36 439  I am a US Federal Government employee acting in the course of my employment. 37 440  I am not a US Federal Government employee, but some or all of my co-authors are. 38 441  I am an employee of the UK Crown acting in the course of my employment 39 442  I am not an employee of the UK Crown acting in the course of my employment but some/all of my co-authors are. 40 443 41 444 42 43 445 44 446 45 447 46 448 47 449 48 450 49 451 50 452 51 453 52 454 53 54 55 56 57 58 59 60

9 https://mc.manuscriptcentral.com/bmj Page 11 of 101 BMJ

1 2 3 455 References: 4 456 1. Maher C, Underwood M, Buchbinder R. Non-specific low back pain. The Lancet 5 457 2017;389:736-47. 6 458 2. Dudli S, Fields AJ, Samartzis D, Karppinen J, Lotz JC. Pathobiology of Modic changes. European 7 459 Spine Journal 2016:1-12. 8 9 460 3. Modic MT, Steinberg PM, Ross JS, Masaryk TJ, Carter JR. Degenerative disk disease: 10 461 assessment of changes in vertebral body marrow with MR imaging. Radiology 1988;166:193. 11 462 4. Urquhart DM, Zheng Y, Cheng AC, et al. Could low grade bacterial infection contribute to low 12 463 back pain?Confidential: A systematic review. BMC Medicine For 2015;13:13. Review Only 13 464 5. Stirling A, Worthington T, Rafiq M, Lambert PA, Elliott TSJ. Association between sciatica and 14 465 Propionibacterium acnes. The Lancet 2001;357:2024-5. 15 466 6. Chen Z, Zheng Y, Yuan Y, et al. Modic Changes and Disc Degeneration Caused by Inoculation 16 467 of Propionibacterium acnes inside Intervertebral Discs of Rabbits: A Pilot Study. BioMed research 17 18 468 international 2016;2016:9612437. 19 469 7. Li B, Dong Z, Wu Y, et al. Association Between Lumbar Disc Degeneration and 20 470 Propionibacterium acnes Infection: Clinical Research and Preliminary Exploration of Animal 21 471 Experiment. Spine 2016;41:E764. 22 472 8. Albert HB, Sorensen JS, Christensen BS, Manniche C. Antibiotic treatment in patients with 23 473 chronic low back pain and vertebral bone edema (Modic type 1 changes): a double-blind randomized 24 474 clinical controlled trial of efficacy. European Spine Journal 2013;22:697-707. 25 475 9. Sotto A, Dupeyron A. Letter to the editor concerning: “Antibiotic treatment in patients with 26 27 476 chronic low back pain and vertebral bone edema (Modic type 1 changes): a double-blind randomized 28 477 controlled trial of efficacy” by Albert HB et al. Eur Spine J (2013) 22:697–707. European Spine Journal 29 478 2013;22:1704-5. 30 479 10. Dean BJF. Do these results apply to the 'intervention naive' patient? European Spine Journal 31 480 2013;22:1702-. 32 481 11. Low back pain and sciatica in over 16s: assessment and management. NICE guideline [NG59] 33 482 Published date: November 2016. 2016. (Accessed April 9,2019, at 34 483 https://www.nice.org.uk/guidance/ng59/resources.) 35 484 12. Blaser M. The microbiome revolution. J Clin Invest2014:4162-5. 36 37 485 13. Herlin C, Kjaer P, Espeland A, et al. Modic changes-Their associations with low back pain and 38 486 activity limitation: A systematic literature review and meta-analysis. PLoS One 2018;13:e0200677. 39 487 14. Jensen T, Karppinen J, Sorensen J, Niinimaki J, Leboeuf-Yde C. Vertebral endplate signal 40 488 changes (Modic change): a systematic literature review of prevalence and association with non- 41 489 specific low back pain. Eur Spine J 2008;17:1407 - 22. 42 490 15. Moher D, Hopewell S, Schulz K, et al. CONSORT 2010 explanation and elaboration: updated 43 491 guidelines for reporting parallel group randomised trials. BMJ 2010. 44 492 16. Rahme R, Moussa R. The Modic Vertebral Endplate and Marrow Changes: Pathologic 45 46 493 Significance and Relation to Low Back Pain and Segmental Instability of the Lumbar Spine. American 47 494 Journal of Neuroradiology 2008;29:838-42. 48 495 17. Bendix T, Sorensen JS, Henriksson GAC, Bolstad JE, Narvestad EK, Jensen TS. Lumbar modic 49 496 changes-a comparison between findings at low- and high-field magnetic resonance imaging. Spine 50 497 2012;37:1756. 51 498 18. Storheim K, Espeland A, Grøvle L, et al. Antibiotic treatment In patients with chronic low back 52 499 pain and Modic changes (the AIM study): study protocol for a randomised controlled trial. Trials 53 500 2017;18. 54 501 19. Common Terminology Criteria for Adverse Events (CTCAE) v4.0. 2010. (Accessed April 9,2019, 55 56 502 at http://evs.nci.nih.gov/ftp1/CTCAE/About.html.) 57 503 20. Bråten LC RM, Espeland A, Storheim K, Zwart JA, Hellum C, Aßmus J. Statistical analysis plan 58 504 (SAP) for clinical outcomes in the AIM-study: A randomized trial of antibiotic treatment in patients 59 505 with chronic low back pain and Modic Changes (the AIM-study). ClinicalTrials.gov2018. 60

10 https://mc.manuscriptcentral.com/bmj BMJ Page 12 of 101

1 2 3 506 21. United Kingdom back pain exercise and manipulation (UK BEAM) randomised trial: 4 507 effectiveness of physical treatments for back pain in primary care. BMJ 2004;329:1377. 5 508 22. Brouwer S, Kuijer W, Dijkstra PU, Göeken LNH, Groothoff JW, Geertzen JHB. Reliability and 6 7 509 stability of the Roland Morris Disability Questionnaire: intra class correlation and limits of agreement. 8 510 Disability and Rehabilitation 2004;26:162-5. 9 511 23. Grotle M, Vøllestad N, Brox J. Cross-cultural adaptation of the Norwegian versions of the 10 512 Roland-Morris Disability Questionnaire and the Oswestry Disability Index. Journal of Rehabilitation 11 513 Medicine 2003;35:241-7. 12 514 24. BangConfidential: H, Ni L, Davis CE. Assessment ofFor blinding Reviewin clinical trials. Controlled Only Clinical Trials 13 515 2004;25:143-56. 14 516 25. Groll DL, To T, Bombardier C, Wright JG. The development of a comorbidity index with 15 517 physical function as the outcome. Journal of Clinical Epidemiology 2005;58:595-602. 16 17 518 26. Sandanger I, Moum T, Ingebrigtsen G, Dalgard OS, Sorensen T, Bruusgaard D. Concordance 18 519 between symptom screening and diagnostic procedure: the Hopkins Symptom Checklist-25 and the 19 520 Composite International Diagnostic Interview I. Social psychiatry and psychiatric epidemiology 20 521 1998;33:345-54. 21 522 27. Ostelo RWJG, Deyo RA, Stratford P, et al. Interpreting change scores for pain and functional 22 523 status in low back pain: towards international consensus regarding minimal important change. Spine 23 524 2008;33:90. 24 525 28. Solberg T, Johnsen LG, Nygaard OP, Grotle M. Can we define success criteria for lumbar disc 25 26 526 surgery? : estimates for a substantial amount of improvement in core outcome measures. Acta 27 527 Orthop 2013;84:196-201. 28 528 29. Austevoll IM, Gjestad R, Grotle M, et al. Follow-up score, change score or percentage change 29 529 score for determining clinical important outcome following surgery? An observational study from the 30 530 Norwegian registry for Spine surgery evaluating patient reported outcome measures in lumbar spinal 31 531 stenosis and lumbar degenerative spondylolisthesis. BMC Musculoskeletal Disorders 2019;20:31. 32 532 30. Nord CE, Oprica C. Antibiotic resistance in Propionibacterium acnes. Microbiological and 33 533 clinical aspects. Anaerobe 2006;12:207-10. 34 35 534 31. Thomas Rde W, Batten J, Want S, McCarthy I, Brown M, Hughes S. A new in-vitro model to 36 535 investigate antibiotic penetration of the intervertebral disc. Bone & Joint Journal 1995;77-B:967-70. 37 536 32. Housden P, Sullivan M. Do Augmentin or cefuroxime reach effective levels in lumbar 38 537 vertebral discs when used prophylactically for discectomy? European Spine Journal 1993;2:145-8. 39 538 33. Ferreira ML, Herbert RD, Ferreira PH, et al. The smallest worthwhile effect of nonsteroidal 40 539 anti-inflammatory drugs and physiotherapy for chronic low back pain: a benefit-harm trade-off study. 41 540 J Clin Epidemiol 2013;66:1397-404. 42 541 34. McNamara AL, Dickerson EC, Gomez-Hassan DM, Cinti SK, Srinivasan A. Yield of Image- 43 542 Guided Needle Biopsy for Infectious Discitis: A Systematic Review and Meta-Analysis. AJNR American 44 45 543 journal of neuroradiology 2017;38:2021-7. 46 544 35. Giarratano A, Green SE, Nicolau DP. Review of antimicrobial use and considerations in the 47 545 elderly population. Clinical interventions in aging 2018;13:657-67. 48 546 36. Safety of Long Term Therapy with Penicillin and Penicillin Derivatives. 2001. 49 547 50 548 51 52 549 53 550 54 55 56 57 58 59 60

11 https://mc.manuscriptcentral.com/bmj Page 13 of 101 BMJ

1 2 3 551 Table 1 -Baseline characteristics 4 5 Amoxicillin (N=89) Placebo (N=91) 6 Age 44·7 (9·0), n=89 45·2 (9·0), n=91 7 Women 52/89 (58%) 52/91 (57%) 8 9 BMI 26·1 (4·1), n=89 25·9 (4·0) , n=90 10 Smoking, yes 25/89 (28%) 21/89 (24%) 11 Educational level 12 Confidential: For Review Only Primary school (9 years) 10/88 (11%) 9/89 (10%) 13 14 High school (12 years) 36/88 (41%) 42/89 (47%) 15 College or university (<4 years) 27/88 (31%) 18/89 (20%) 16 University (≥4 years) 15/88 (17%) 20/89 (22%) 17 Comorbidity* 18 19 Score 1 (back pain only) 62/89 (70%) 60/91 (66%) 20 Score 2 21/89 (24%) 27/91 (30%) 21 Score >2 6/89 (7%) 4/91 (4%) 22 Former disc surgery ‡ 18/89 (20%) 20/91 (22%) 23 24 Emotional distress (HSCL-25 ≥ 1.75)§ 24/88 (27%) 23/91 (25%) 25 FABQ physical activity (0-24) ¶ 11·2 (5·9), n=89 12·6 (5·8) , n=90 26 FABQ work (0-42) ¶ 17·0 (11·7), n=87 18·9 (12·0) , n=89 27 Duration of back pain in years† 3·0 (1·5-5·6), n=89 3·4 (1·7-7) , n=90 28 29 Physical work load 30 Mostly sitting 37/77 (48%) 26/74 (35%) 31 Job requires a lot of walking 20/77 (26%) 20/74 (27%) 32 Job requires a lot of walking and lifting 17/77 (22%) 24/74 (32%) 33 34 Job requires physically heavy work 3/77 (4%) 4/74 (5%) 35 Employment status 36 Working full time 46/89 (52%) 43/91 (47%) 37 Partial sick leave 14/89 (16%) 20/91 (22%) 38 39 Complete sick leave 22/89 (25%) 16/91 (18%) 40 Disability pension 3/89 (3%) 7/91 (8%) 41 Unemployed 2/89 (2%) 3/91 (3%) 42 Student/other/unknown 2/89 (2%) 2/91 (2%) 43 44 Modic type I group 58/89 (65%) 60/91 (66%) 45 Level of Modic change and previous disc herniation 46 L1/L2 0/89 0/91 47 L2/L3 2/89 (2%) 2/91 (2%) 48 49 L3/L4 7/89 (8%) 5/91 (5%) 50 L4/L5 48/89 (54%) 29/91 (32%) 51 L5/S1 58/89 (65%) 74/91 (81%) 52 552 Data are mean (SD), median (IQR), or no./N(%). N refers to the number of patients providing data. 53 553 * Functional Comorbidity Index25 – Score increased by 1 for each of 18 diagnoses associated with decreased 54 554 physical function. 55 555 § Emotional distress (Hopkins Symptom Checklist–25), values ≥1.75 associated with psychiatric diagnosis.26 56 556 ¶ Fear-Avoidance Beliefs Questionnaire), higher values indicating more fear-avoidance beliefs. 57 557 58 59 60 558

12 https://mc.manuscriptcentral.com/bmj BMJ Page 14 of 101

1 2 3 4 559 Table 2 - Primary and secondary outcomes, all time periods 5 560 6 561 Treatment groups Treatment comparison (adjusted mean difference) 7 562 Amoxicillin Placebo ANCOVA LME ‖ 8 563 Outcome N* Confidential:Mean±SD N Mean±SD N** ForMean (95%CI) Reviewp-value Mean (95%CI)Onlyp-value 9 564 RMDQ (0-24) 10 565 Baseline 88 12·7±4·7 90 12·8±3·7 - - - - - 11 566 3 months 85 10·3±5·8 87 12·4±4·4 180 -1·9 (-3·2 to -0·7) 0·003 -1·9 (-3·3 to -0·5) 0·006 12 567 6 months 83 9·9±5·9 78 11·7±5·2 179‡ -1·5 (-2·9 to 0·0)¶ 0·047 -1·6 (-3·2 to -0·1) 0·04 13 568 569 9 months 77 9·7±6·4 79 11·1±5·4 179‡ -1·6 (-3·1 to 0·0)¶ 0·049 -1·7 (-3·3 to -0·1) 0·04 14 570 12 months 85 9·0±6·2 84 10·7±5·6 180 -1·6 (-3·1 to 0·0) 0·04 -1·7 (-3·3 to -0·1) 0·04 15 571 16 ODI (0-100) 572 Baseline 88 31·9±11·4 89 31·8±10·3 - - - - - 17 573 18 574 3 months 86 26·6±14·7 85 30·4±10·7 179‡ -3·8 (-6·7 to -0·9)§ 0·010 -4·1 (-7·0 to -1·2) 0·006 19 575 12 months 85 24·4±15·0 84 28·9±14·0 179‡ -4·8 (-8·3 to -1·4)§ 0·007 -5·1 (-8·5 to -1·6) 0·004 20 576 Back pain intensity (0-10) 21 577 Baseline 88 6·4±1·2 90 6·3±1·5 - - - - - 22 578 23 579 3 months 85 5·2±2·3 85 5·4±1·9 179‡ -0·3 (-0·8 to 0·3)§ 0·33 -0·4 (-0·8 to 0·1) 0·13 24 580 6 months 83 5·1±2·2 77 5·5±2·2 178‡ -0·5 (-1·1 to 0·1)§¶ 0·11 -0·5 (-1·0 to 0·0) 0·03 25 581 9 months 78 5·2±2·4 77 5·0±2·3 178‡ 0·0 (-0·7 to 0·6)§¶ 0·97 0·0 (-0·5 to 0·5) 0·88 26 582 12 months 85 4·7±2·3 84 5·2±2·3 179‡ -0·6 (-1·3 to 0·0)§ 0·06 -0·7 (-1·2 to -0·2) 0·005 27 583 EQ-5D (-0.59 - 1) 28 584 Baseline 89 0·55±0·19 91 0·54±0·18 - - - - - 29 585 30 586 3 months 85 0·60±0·22 83 0·54±0·21 180 0·06 (0·00 to 0·11)§ 0·04 0·06 (0·00 to 0·11) 0·05 31 587 12 months 84 0·65±0·22 83 0·58±0·22 180 0·07 (0·02 to 0·12)§ 0·012 0·07 (0·01 to 0·13) 0·015 32 588 589 Leg pain intensity (0-10) 33 Baseline 89 3·2±2·6 90 3·2±2·6 - - - - - 34 590 3 months 85 3·1±2·8 84 3·4±2·6 168 -0·3 (-0·9 to 0·4)§¶ 0·42 35 12 months 85 2·8±2·7 82 3·5±2·8 166 -0·8 (-1·4 to -0·1)§¶ 0·03 - - 36 591 37 38 39 592 RMDQ Roland-Morris Disability Questionnaire. Score from 0 to 24. Higher scores indicate more severe pain and disability. 40 593 ODI Oswestry Disability Index. Score from 0 to 100. Higher scores indicate more severe pain and disability. 41 42 13 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 15 of 101 BMJ

1 2 3 594 EQ5D EuroQol’s health-related quality of life. Score from -0·59 to 1. Higher scores indicating better quality of life. 4 595 * Number of answered questionnaires of outcome 5 596 † Imteraction between time and treatment at 12 months with simple contrasts 6 597 ‡ Some patients excluded from analysis as the imputation model did not manage to impute all missing variables 7 598 § Estimates smaller than the recommended thresholds for clinical important change within groups (ODI 13 to 20; LBP intensity NRS 2 to 3; 8 599 Leg pain intensity NRS Confidential:2 to 3·5; EQ5D 0·105 to 0·3) For Review Only 9 600 ¶ Post-hoc analyses not described in the registry 10 601 ‖ LME model including all timepoints (incl week 0-13) 11 602 ** Number of cases included in the analyses after multiple imputations 12 603 13 604 14 605 15 606 16 607 608 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 14 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 16 of 101

1 2 3 609 Table 3 - Work, global perceived effect, patient satisfaction and medication use 4 610 5 Treatment Treatment groups 6 comparison 7 8 Confidential:Amoxicillin For ReviewPlacebo Only 9 Outcome P-value 10 11 Working (including part time sick leave) 12 Baseline –no./total no. (%) 60/89 (67%) 63/91 (69%) 13 14 1 year –no./total no. (%) 56/85 (63%) 53/84 (58%) 0·62 15 Global perceived effect at 1 year † 0·39 16 Improved –no./total no. (%) 24/85 (28%) 18/84 (21%) 17 18 No change –no./total no. (%) 58/85 (68%) 60/84 (71%) 19 Worse –no./total no. (%) 3/85 (4%) 6/84 (7%) 20 21 Patient satisfaction at 1 year 0·52 22 Satisfied –no./total no. (%) 35/85 (41%) 28/84 (33%) 23 24 Somewhat satisfied –no./total no. (%) 8/85 (9%) 9/84 (11%) 25 Neither satisfied nor dissatisfied –no./total no. (%) 34/85 (40%) 35/84 (42%) 26 Somewhat dissatisfied –no./total no. (%) 5/85 (6%) 4/84 (5%) 27 28 Dissatisfied –no./total no. (%) 3/85 (4%) 8/84 (10%) 29 Concomitant medication use 30 31 Analgesics, any 32 Baseline –no./total no. (%)‡ 62/89 (70%) 61/91 (67%) 33 1 year –no./total no. (%) 61/89 (69%) 67/91 (74%) 0·45 34 35 NSAIDs 36 Baseline –no./total no. (%)¶ 38/89 (43%) 36/91 (40%) 37 38 1 year –no./total no. (%) 39/89 (44%) 40/91 (44%) 0·99 39 Opioids 40 41 42 15 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 17 of 101 BMJ

1 2 3 Baseline –no./total no. (%)§ 28/89 (31%) 27/91 (30%) 4 1 year –no./total no. (%) 28/89 (31%) 35/91 (38%) 0·33 5 6 611 † 7-point likert scale recoded to three categories (Improved= Completely recovered, much better. No change=somewhat better, no change, somewhat worse. 7 612 Worse=much worse, worse than ever) 8 613 ‡ 5 patients in the amoxicillinConfidential: group and 4 patients in the placebo group reported For taking analgesics Review for other reasons than LBP Only only 9 614 § No patient took strong opioids. 1 patient in the placebo group reported taking opioids for other reasons than LBP only 10 615 ¶ 2 patients in the amoxicillin group reported taking NSAIDs for other reasons than LBP only 11 12 616 13 617 14 15 16 618 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 16 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 18 of 101

1 2 3 619 Table 4 -Adverse events 4 5 Amoxicillin (n=89) Placebo (n=91) 6 Any adverse event 75 (84%) 76 (84%) 7 Unrelated/unlikely only 25 (28%) 45 (49%) 8 a 9 Drug related, any 50 (56%) 31 (34%) 10 Severity grade ≥ 2 39 (44%) 37 (41%) 11 12 Unrelated/unlikelyConfidential: only For17 (19%) Review32 (35%)Only 13 Drug related, anya 22 (25%) 5 (5%) 14 15 Any serious adverse event 6 (7%) 2 (2%) 16 Unrelated/unlikely only 6 (7%) 2 (2%) 17 Drug related, anya 0 0 18 19 Reported symptoms or eventsb 20 Abdominal pain 11 (12%) 3 (3%) 21 Diarrhea 17 (19%) 10 (11%) 22 23 Rash 19 (21%) 5 (5%) 24 Vaginal Candida infectionc 11 (21%) 0 25 Oral Candida infection 1 (1%) 0 26 620 a Defined as possible /probable/definite attribution to study medication by care providers. 27 621 Any patient with both 'unrelated/unlikely' and ‘drug related' adverse events will be counted in the group ‘drug 28 622 related’. 'unrelated/unlikely' and ‘drug related’ are summed up in their respective title lines above. 29 623 b Terms as defined by caregivers (not necessarily identical to CTCAE) 30 624 c Analyzed with only women as denominator (n=52 in each treatment group) 31 625 32 626 33 627 34 628 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

17 https://mc.manuscriptcentral.com/bmj Figure 2- Flowchart:

Page 19 of 101 BMJ

1 2 3 Figure 1 - Flowchart 4 5 6 582 Assessed for eligibility 7 310 Failed to meet inclusion/exclusion criteria 8 52 No disc herniation within the last two years 9 68 Pain intensity score too low 10 16 Co-morbidity (other LBP diagnosis or previous surgery 11 other than for disc herniation) 174 Other inclusion/exclusion criteria not fulfilled 12 Confidential: For Review Only 180 Underwent randomisation 92 Unknown reasons 13 14 15 16 17 18 19 89 Allocated to amoxicillin and started study 91 Allocated to placebo and started study 20 medication medication 21 88 Reported baseline RMDQ 90 Reported baseline RMDQ 22 23 End of study during treatment phase 24 End of study during treatment phase (0-3 months): (0-3 months): 25 1 lost to follow-up 1 lost to follow-up 26 2 voluntarily discontinued study 1 voluntarily discontinued study 27 28 29 Treatment non-completion: 30 Treatment non-completion: 1 stopped study medication due to an adverse event 1 stopped study medication due to an adverse event 31 1 related to protocol deviation1 related to protocol deviation 4 32 3 stopped study medication due to adverse events 2 voluntarily discontinued study medication 33 related to study medication2 5 non-compliers 34 1 was a non-complier and later stopped study 35 medication due to an adverse event3 36 5 non-compliers 37 38 39 85 Patients evaluated at 1-year follow-up 89 Patients evaluated at 1-year follow-up 40 85 Reported RMDQ at 1 year 84 Reported RMDQ at 1 year 41 77 Completed study without major protocol 78 Completed study without major protocol 42 deviations (per-protocol population) deviations (per-protocol population) 43 End of Study before 3 months (see box above)(n=3) End of Study before 3 months (see box above) (n=2) Treatment non-completion, not including2 (n=7) Treatment non-completion (n=8) 44 Operation for disc herniation (n=1) Diagnosed spondyloarthritis (n=1) 45 Incorrect enrollment (n=1)5 Given both amoxicillin+placebo (n=1)6 46 89 Included in the intention-to-treat population Incorrect enrollment (n=1)7 47 91 Included in the intention-to-treat population 48 49 1 One patient in the amoxicillin group and one patient in the placebo group became pregnant (protocol deviation as all patients were 50 instructed to use contraception), not included in per-protocol population 2 Three patients in the amoxicillin group stopped study medication due to adverse events and were included in the per-protocol 51 population 52 3 One patient in the amoxicillin group stopped study medication due to adverse events but were still not included in the per-protocol 53 population due to poor compliance before stopping study medication 4 Two patients in the placebo group discontinued because they started three month treatment with amoxicillinc/clavulanic acid 54 5 Treated with apocillin 7 days prior to randomisation 55 6 Due to a mistake at pharmacy, the patient was given a mix of bottles containing amoxicillin and placebo 56 7 Treated with cephalexin 7 days prior to randomisation 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 20 of 101

1 2 3 Figure 2 -RMDQ and LBP intensity from baseline to 1 year 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 Both graphs are based on the LME analyses 37 RMDQ -Roland-Morris Disability Questionnaire. Score from 0 to 24. Higher scores indicate more severe 38 pain and disability. 39 LBP intensity -Low back pain intensity (Numerical Rating Scale 0-10) 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 21 of 101 BMJ

1 2 3 1 Supplementary Appendix 4 5 2 6 7 8 3 Table of Contents 9 4 Methods...... 2 10 5 Statistical Analysis ...... 2 11 12 6 FigureConfidential: S1 - List of inclusion and exclusion For criteria Review...... 3 Only 13 14 7 Table S1 - List of trial measurements with timing ...... 4 15 8 Results ...... 6 16 17 9 Table S2 - Responder analyses...... 6 18 10 Figure S2 - Cumulative distribution function of responders...... 8 19 20 11 Table S3 - Response to blinding question at one year ...... 9 21 22 12 Figure S3 - Kernel density estimates...... 10 23 13 Table S4 – Response to blinding question at one year follow-up...... 12 24 25 14 Table S5 - Sensitivity analyses...... 13 26 27 15 Table S6 - Primary and secondary outcomes for the separate MC types...... 14 28 16 Figure S4 - Intensity grade 1 adverse events, frequency distribution ...... 15 29 30 17 Figure S5 - Intensity grade 2 adverse events, frequency distribution ...... 16 31 18 Figure S6 - Serious adverse events or intensity grade 3-4 adverse events, frequency distribution17 32 33 19 Table S7 - Most frequent adverse event types by attribution group and treatment group ...... 18 34 35 20 References:...... 20 36 21 37 38 22 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

1 https://mc.manuscriptcentral.com/bmj BMJ Page 22 of 101

1 2 3 23 Methods 4 24 5 6 25 Statistical Analysis 7 26 In all analyses and descriptive statistics of the primary endpoints (RMDQ consisting of 24 items), we performed 8 27 a mean-imputation in cases with less than 30% of the items missing at any time point. We then performed a 9 28 multiple imputation in those cases where more than 30% of the questions were missing at any time point for the 10 29 analyses of treatment comparison. The imputation model included a broad group of variables associated with the 1 11 30 primary outcome, as described in the Statistical analysis plan. We defined drug effect as the difference between 12 31 the mean effectConfidential: of amoxicillin and the mean effect For of placebo. Review Only In the primary analysis we compared mean RMDQ scores at one year between the two treatment groups in the 13 32 33 whole intention-to-treat population using ANCOVA, adjusted for baseline RMDQ score and the stratification 14 34 variables (MCs type and previous disc surgery). 15 35 We also performed analyses of secondary outcome measures (ODI, LBP intensity and Health-related quality of 16 36 life) in a similar manner as the primary outcome, using multiple imputation models with significance level 17 37 adjusted to 0· 0167 as described in the Statistical Analysis Plan.2 18 38 We performed a per-protocol analysis for the primary outcome, where all patients with a major protocol 19 39 deviation were excluded. Major protocol deviation is defined as noncompliance (taking less than 80% of the 20 40 prescribed pills), incorrect enrollment, back surgery during the trial period, pause of the study medication (both 21 41 treatment groups) or intake of antibiotics (in the placebo group) as reported in Table 5 in the Statistical Analysis 22 42 Plan.2 However, patients who stopped the study medication due to clinical reasons (i.e. adverse events not 23 43 related to protocol deviation) were included in the per-protocol population.3 Compliance was defined as the 24 44 percent of pills that the patient has taken out of the planned number of pills. It was calculated based on data on 25 45 counts of all the pills that the patients returned at end of treatment (Table S1). 26 46 Minor protocol deviation is defined as compliance less than 95% but more than 80%, any treatment for back pain 27 47 (initiated at any time point between baseline and one year follow up), pause of study medication (both treatment 28 48 groups) or intake of antibiotics (in the placebo group) as reported in Table 5 in the Statistical Analysis Plan.2 In 29 49 cases where data for compliance was missing, we relied on the weekly patient reported compliance questionnaire 30 50 (Table S1). Out of 28 patients with missing information on counts of returned pill, 23 patients had reported 31 51 taking tablets less than 80% of the days (or had missing information about these reported numbers) and were 32 52 defined as noncompliance, and hence also as having a major protocol deviation. Treatment non-completion 33 53 (reported in figure 1 in the main manuscript) is defined as either stopping study medication due to adverse events 34 54 or non-compliers (the distinction between these two is relevant for deciding if the patient were included in the 35 55 per-protocol population or not, as described above).

36 56 57 37 58 We calculated, separately for each treatment group, the number and percentage of patients with 1) one or more 38 59 adverse events, 2) one or more grade 2 adverse events, 3) serious adverse events or grade 3-4 toxicity and 4) 39 60 symptoms of diarrhea, abdominal pain, rash, candida infection. For each of these four categories of 40 61 events/symptoms, we also report the number of drug related adverse events (defined as 41 62 possible/probable/definite relationship to treatment) and adverse events with unrelated/unlikely relationship to 42 63 treatment. For all these calculations we present events/symptoms descriptively without any formal statistical 43 64 testing. 44 65 45 66 Further analyses included responder analyses (improved >30, >50 and > 75% compared to baseline value, 46 67 excluding patients with >30% items missing of the RMDQ) and Linear mixed-effects (LME) models as 47 68 described in the Statistical Analysis Plan.2 48 69 49 70 Unweighted kappa was calculated for agreement between the study radiologists on trial eligibility based on MRI 50 71 findings of disc herniation and MCs, using the categories eligible for the MCs type I group, eligible for the MCs 51 72 type II group, or not eligible. 52 73 53 74 The graphics were derived using Matlab 9.0 (Natick, MA). 54 75 55 56 57 58 59 60

2 https://mc.manuscriptcentral.com/bmj Page 23 of 101 BMJ

1 2 3 76 4 Figure S1 - List of inclusion and exclusion criteria 5 6 INCLUSION CRITERIA 7  Age between 18 and 65 years 8  LBP of > 6 months duration in the area below the 12th rib and above the gluteal folds with a Numerical 9 Rating Scale (NRS) pain intensity score of  5 (mean of three 0-10 NRSs; current LBP, the worst LBP 10 within the last 2 weeks, and usual/mean LBP within the last 2 weeks). 11  MRI-confirmed lumbar disc herniation within the preceding 2 years. 12  Type Confidential:I and/or type II MCs in the vertebral body For marrow Review at the same level as the Only previously herniated disc. 13 For patients with previous surgery for disc herniation, the MCs has to be located at an operated level. 14  Written informed consent 15 16 EXCLUSION CRITERIA 17  Allergy to penicillin or cefalosporins 18  Allergy/hypersensitivity to any of the excipients of the study drug 19  Current pregnancy or lactation 20  Kidney (creatinine) or hepatic (ALAT/ASAT) laboratory values above the normal range 21  Phenylketonuria (Følling’s disease) 22  Mononucleosis or leukaemia 23  Any specific diagnosis that may explain patient’s low back symptoms (e.g. tumor, fracture, 24 spondyloarthritis, infection, spinal stenosis). 25  Previous low back surgery (L1 – S1) for reasons other than disc herniation (e.g fusion, decompression, disc 26 prosthesis). 27  Surgery for disc herniation within the last 12 months 28  Previous surgery for disc herniation, but MCs located at level(s) that has/have not been operated on only. 29  Reservation about the intake of gelatine (the capsules contains gelatine, which among other things is 30 produced by ingredients derived from pigs) 31  Regular use of glucocorticoids 32  Regular use of opioids with the exception of codeine and tramadol 33  Not understanding Norwegian language 34  Unlikely to adhere to treatment and/ or complete follow-up (e.g serious ongoing psychiatric disease, drug 35 abuse, plans to move) 36  Antibiotic treatment within the preceding one month before treatment start 37  Contraindications to MRI (e.g. cardiac pacemaker electrodes, metal implant in eye or brain, 38 claustrophobia). 39  Unwilling to participate 40 41 42 77 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3 https://mc.manuscriptcentral.com/bmj BMJ Page 24 of 101

1 2 3 78 Table S1 - List of trial measurements with timing 4 Trial measurements Timeline 5 6 Primary outcome 7  Roland-Morris Disability Questionnaire (RMDQ), Norwegian Reported at 0, 3, 6, 9 and 12 8 validated version, range 0-24, with higher score indicating more months 9 severe pain and disability, measured at one year follow-up (Primary 10 endpoint). 4,5 11 12 Confidential: SecondaryFor outcomesReview Only 13  Oswestry Disability Index (ODI) 2.0, range 0-100. 6 (Key supportive Reported at 0, 3 and 12 14 endpoint) months 15 16  Low back pain intensity (mean of three Numeric Rating Scales Reported every week during 17 (NRSs, range 0-10); current LBP, the worst LBP within the last 2 treatment period and at 6, 9 18 weeks, and usual/mean LBP within the last 2 weeks (for weekly and 12 months 19 reports during the intervention period; the wording "last 2 weeks" 20 was replaced by "the last week")7 (Key supportive endpoint) 21 22  Health-related quality of life (EuroQoL-5D, version 2.0)8 (Key Reported at 0, 3 and 12 23 supportive endpoint) months 24 25 Further outcomes 26 27  Leg pain intensity (NRSs, range 0-10) last week 6 Reported at 0, 3 and 12 28 months 29  Hours with LBP during the last 4 weeks Reported at 0, 3 and 12 30 months 31 Reported at 3 and 12 months 32  Patients’ satisfaction ( 5-point Likert scale) 33 34  Global perceived effect (7-point Likert scale) Reported at 3 and 12 months 35 36  Work status including days with sick leave Reported at 0, 3 and 12 37 months by care providers and 38 every month during the trial 39 period by the patient 40  Co-interventions (other pharmacological (ATC-coded) and non- Reported at 0, 1,2, 3 and 12 41 pharmacological treatments) months by care providers and every month during the 42 follow up phase by the 43 patient 44 Other trial measurements 45 46 9 Reported at baseline 47  Emotional distress (Hopkins Symptom Checklist–25) 48 10 49  Fear-avoidance beliefs Questionnaire (FABQ) Reported at baseline 50 51  Subjective health complaints (SHC) 11 Reported at baseline 52 53  Background information (age, gender, BMI, ethnicity, marital Reported at baseline 54 status, educational level, work status, physical work load, leisure 55 time activity, smoking habits, past medical history including 56 previous surgery for disc herniation and duration of back pain) 57 58  Blinding questionnaire Reported at 3 and 12 months 59 60  Number of days the last week the patients took the study medication Reported every week during

4 https://mc.manuscriptcentral.com/bmj Page 25 of 101 BMJ

1 2 3 (0-7) treatment period by the 4 patient 5  Counts of returned capsules at end of treatment period Reported at 3 months by care 6 providers 7  Haematological parameters (White cell counts, thrombocytes, assessed and registered 8 haemoglobin (Hb) and hematocrit (Ht)), measures of kidney monthly during the 9 (creatinine), liver function (ASAT / ALAT), CRP and Glucose intervention period and at 10 one year follow up 11  Clinical evaluation (blood pressure, pulse, auscultation of hearth assessed and registered 12 and lunges)Confidential: For Review Onlymonthly during the 13 intervention period and at 14 one year follow up 15  Expectations of treatment effect Baseline 16  Q: “What is your personal expectation for antibiotic treatment 17 improving your back pain?” Alternatives: “My back pain will be 18 cured” “It results in a clear improvement” “It results in a small 19 improvement” “It results in no improvement” “Don’t know” 20 21 79 Data completeness were encouraged by SMS and assessed weekly for all participants. In case of missing data, 22 80 clinicians and/or patients were contacted to detect the cause. 23 24 81 25 82 26 27 83 28 84 29 30 85 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

5 https://mc.manuscriptcentral.com/bmj BMJ Page 26 of 101

1 2 3 4 Results 5 6 7 8 Confidential: For Review Only 9 Figure S2 - Appendix to flowchart 10 11 174 patients were listed in the “Other inclusion/exclusion criteria not fulfilled” in Figure 1. The reasons are as follows: 12 13 Do not wish to participate in study 53 14 Modic Changes not present 18 15 16 Allergy to penicillin 10 17 Elevated creatinine or ASAT/ALAT 3 18 Former lumbar disc surgery, but <12 months since operation 1 19 20 Former lumbar disc surgery, but Modic Change at different level 1 21 Reservations to eat product from swine (gelatin) 2 22 23 Regular use of glucocorticoids 1 24 Regular use of opioids (except codeine and tramadol) 5 25 Poor language skills (Norwegian) 5 26 27 Low compliance 15 28 Antibiotic treatment within 1 month of study participation 7 29 30 Other causes not listed as specific exclusion criteria (such as too old MRI scans, 58 31 spontaneous improvement in condition, massive comorbidity, seeking commercial 32 treatment in Denmark, low back pain not present, pregnancy in planning, planned 33 elective surgery) 34 35 36 The sum of these patients equals 179 (not 174). This is probably due to five double registrations and/or multiple 37 inclusion/exclusion criteria not fulfilled in one or more of the study centers. 38 39 40 41 42 6 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 27 of 101 BMJ

1 2 3 4 5 6 7 8 Table S2 - Responder analysesConfidential: For Review Only 9 Treatment groups Treatment comparison 10 11 Improvement in RMDQ score from baseline to 1 year Amoxicillin group Placebo group N* Numbers needed to treat P-value 12 (n= 89) (n= 91) (NNT) 13 14 15 Improved >30% –no./total no. (%) 40/84 (48%) 24/83 (29%) 167 5·3 ( 3·0 to 24) 0·01 16 17 Improved >50% –no./total no. (%) 23/84 (27%) 18/83 (22%) 167 18 (5·3 to -14) 0·39 18 19 20 Improved >75% –no./total no. (%) 15/84 (19%) 7/83 (8%) 167 10·6 (5·1 to -140) 0·07 21 22 Table shows number of patients who improved (>30%, >50% or >75%) from baseline to one year, out of total number who answered RMDQ 23 at baseline and one year. No imputations were performed in Responder analyses 24 * Number of cases included in the analyses 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 7 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 28 of 101

1 2 3 Figure S3 - Cumulative distribution function of responders 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 Cumulative proportion of the patients depending on change in the primary outcome (RMDQ) from baseline to one year. 33 34 35 36 37 38 39 40 41 42 8 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 29 of 101 BMJ

1 2 3 Table S3 - Response to blinding question at one year 4 Amoxicillin group (n= 89) Placebo group (n= 91) 5 6 Antibiotics 23 9 7 Placebo 36 50 8 Confidential: For Review Only 9 Unsure - antibiotic 11 7 10 Unsure - placebo 12 15 11 12 Unsure – missing subgroup 2 2 13 Missing 5 8 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 9 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 30 of 101

1 2 3 Figure S4 - Kernel density estimates 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Smoothed curves of the probability distribution for the number of patients of the RMDQ score at one year (kernel estimator with bandwidth 2·0-2·8). 32 33 The curves allow comparison of the estimated probability density functions of the RMDQ score at 1 year between the two treatment groups (amoxicillin 34 versus placebo) within each of three subgroups. Results were inconsistent. In the subgroup reporting they were unsure what treatment they received, the 35 green curves suggest a slightly higher probability of low RMDQ scores at 1 year in the amoxicillin treatment group (n=25, bold green curve) versus the 36 placebo treatment group (n=24, dotted green curve). Similar minor differences were found in the subgroup reporting they thought they received placebo 37 (amoxicillin treatment group n=36, bold blue curve; placebo treatment group n=50, dottet blue curve). Among those reporting they thought they received 38 amoxicillin, numbers were smaller and results were opposite: the red curves suggest a lower (not higher) probability of low RMDQ score at 1 year in the 39 40 41 42 10 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 31 of 101 BMJ

1 2 3 amoxicillin treatment group (n=23, bold red line) versus placebo treatment group (n=9, dotted red line). Patients perceiving the received placebo (blue) 4 were generally more likely than the others (green and red) to report higher (worse) RMDQ score at 1 year. 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 11 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 32 of 101

1 2 3 Table S4 – Response to blinding question at one year follow-up. 4 5 Odds Ratio 95% CI P-value 6 7 Amoxicillin group 2·3 0·9 to 5·5 0·07 8 Confidential: For Review Only 9 Responder at 3 months (>30% improvement from baseline) 3·1 1·3 to 7·3 0·01 10 11 12 Drug-related adverse events (any grade) 2·4 1·02 to 5·6 0·045 13 14 Increment (placebo group, non-responder and no adverse reaction) 0·029 0·006 to 0·136 <0·001 15 16 Odds ratios (CI, p-value) for treatment vs. (placebo or unsure) as response to the blinding question at one year follow-up in the multivariate logistic model. 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 12 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 33 of 101 BMJ

1 2 3 Table S5 - Sensitivity analyses 4 Treatment groups Treatment comparison (adjusted mean difference) 5 Analysis 6 Amoxicillin Placebo ANCOVA 7 N Mean±SD N Mean±SD N* Mean (95%CI) P-Value 8 Confidential: For Review Only 9 RMDQ 1-year 10 Per-protocol population 77 9·2 ±6·3 78 10·5 ±5·5 155 -1·4 (-3·0 to 0·2) 0·09 11 12 ITT population adjusted for study centre 85 9·0±6·2 84 10·7±5·6 180 -1·7 (-3·2 to -0·1) 0·03 13 14 ITT population adjusted for baseline differences 85 9·0±6·2 84 10·7±5·6 180 -1·5 (-3·0 to 0·05) 0·06 15 16 Expectation of treatment effect† 62 9·4±6·2 62 11·2±5·6 132 -2·0 (-3·8 to -0·1) 0·04 17 18 No expectation of treatment effect† 23 8·0±6·2 22 9·2±5·3 48 0·5 (-2·5 to 3·6) 0·73 19 * Number of cases included in the analyses after multiple imputations 20 † Post-hoc analyses, not described in the registry or the Statistical analysis plan. Expectation of treatment effect was defined as “My back pain will be cured”, or “It results in 21 a clear improvement”, no expectation of treatment effect was defined as “It results in a small improvement”, “It results in no improvement”, or “Don’t know” in the response 22 to the question on expectation of treatment effect (Table S3) 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 13 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 34 of 101

1 2 3 4 Table S6 - Primary and secondary outcomes for the separate MC types 5 6 Treatment groups Treatment comparison (adjusted mean difference) 7 Amoxicillin Placebo ANCOVA 8 Outcome N* Mean±SD N Mean±SD N¶ Mean (95%CI) p-value 9 Type I Modic changes 10 RMDQ (0-24)†† 11 Baseline 58 12.9±4.3 60 12.3±3.7 - - - 12 Confidential: For Review Only 12 months 55 8.2±6.0 56 10.3±5.4 118 -2.3 (-4.2 to -0.4) 0.02 13 14 ODI (0-100)† 15 Baseline 58 31.3±11.5 58 30.4±9.9 - - - 16 12 months 55 23.4±15.2 56 27.7±13.8 117‡ -5.1 (-9.3 to -0.8)§ 0.02 17 18 Back pain intensity (0-10) † 19 Baseline 58 6.5±1.1 59 6.3±1.3 - - - 20 12 months 55 5.2±2.2 55 4.5±2.5 117‡ -0.8 (-1.6 to 0.0)§ 0.06 21 EQ-5D (-0.59 - 1)† 22 Baseline 58 0.55±0.18 60 0.56±0.16 - - - 23 12 months 55 0.66±0.21 55 0.60±0.21 118 0.07 (0.01 to 0.14)§ 0.03 24 25 Type II Modic changes 26 RMDQ (0-24) 27 Baseline 30 12.3±5.5 30 13.7±3.5 - - - 28 12 months 30 10.5±6.5 28 11.4±5.9 62 -0.1 (-2.7 to 2.6) 0.95 29 30 ODI (0-100)† 31 Baseline 30 33.0±11.4 31 34.4±10.6 - - - 32 12 months 30 26.1±14.7 28 31.3±14.3 62 -4.5 (-10.6 to 1.6)§ 0.14 33 Back pain intensity (0-10) † 34 Baseline 30 6.3±1.3 31 6.2±1.9 - - - 35 36 12 months 30 5.0±1.9 28 5.2±2.5 62 -0.3 (-1.3 to 0.7)§ 0.52 37 EQ-5D (-0.59 - 1)† 38 Baseline 31 0.53±0.21 31 0.51±0.20 - - - 39 12 months 29 0.61±0.24 28 0.55±0.25 62 0.06 (-0.04 to 0.16)§ 0.22 40 41 RMDQ Roland-Morris Disability Questionnaire. Score from 0 to 24. Higher scores indicate more severe 42 pain and disability. 43 ODI Oswestry Disability Index. Score from 0 to 100. Higher scores indicate more severe pain and 44 disability. 45 EQ5D EuroQol’s health-related quality of life. Score from -0·59 to 1. Higher scores indicating better 46 quality of life. 47 * Number of answered questionnaires of outcome 48 † Post-hoc analyses, not described in the registry or the Statistical analysis plan 49 ‡ One patient excluded from analysis as the imputation model did not manage to impute all missing 50 variables 51 § Estimates smaller than the recommended thresholds for clinical important change within groups (ODI 52 13 to 20; LBP intensity NRS 2 to 3; Leg pain intensity NRS 2 to 3·5; EQ5D 0·105 to 0·3) 53 ¶ Number of cases included in the analyses after multiple imputations 54 55 56 57 58 59 60

14 https://mc.manuscriptcentral.com/bmj Page 35 of 101 BMJ

1 2 3 4 5 Figure S5 - Intensity grade 1 adverse events, frequency distribution 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 15 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 36 of 101

1 2 3 Figure S6 - Intensity grade 2 adverse events, frequency distribution 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 16 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 37 of 101 BMJ

1 2 3 Figure S7 - Serious adverse events or intensity grade 3-4 adverse events, frequency distribution 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 17 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 38 of 101

1 2 3 Table S7 - Most frequent adverse event types by attribution group and treatment group 4 Placebo group Amoxicillin group 5 Adverse event type Unrelated/ Drug related Unrelated/ Drug related Total Unlikely/NA Unlikely/NA 6 7 Diarrhoea 6 14 1 23 44 8 Back pain Confidential:20 1 13 For0 Review34 Only Influenza like illness 6 0 9 1 16 9 Headache 8 1 4 1 14 10 Nausea 0 3 0 9 12 11 Nasopharyngitis 6 0 4 0 10 12 Hepatic enzyme increased 2 1 4 3 10 Abdominal pain upper 0 2 1 6 9 13 Flatulence 1 3 0 4 8 14 Fatigue 0 2 2 3 7 15 Influenza 3 0 4 0 7 Cardiac murmur 4 0 2 0 6 16 Gastroenteritis 4 0 2 0 6 17 Rash maculo-papular 2 1 1 2 6 18 Arthralgia 3 0 2 0 5 Dry mouth 0 2 2 1 5 19 General physical health 3 0 2 0 5 20 deterioration 21 Pain in extremity 3 0 2 0 5 22 Rash 0 0 1 4 5 Tooth discoloration 0 0 3 2 5 23 Urinary tract infection 1 0 3 1 5 24 Abdominal discomfort 0 2 0 2 4 25 Cough 2 0 2 0 4 Dental plaque 1 0 1 2 4 26 Dizziness 0 2 2 0 4 27 Dyspnoea 3 0 0 1 4 28 Hypertension 2 0 2 0 4 Vulvovaginal candidiasis 0 0 0 4 4 29 Vulvovaginitis 0 0 0 4 4 30 Chest injury 2 0 1 0 3 31 Conjunctivitis 3 0 0 0 3 Constipation 1 2 0 0 3 32 Dry skin 0 0 2 1 3 33 Fall 1 0 2 0 3 34 Fungal infection 0 0 0 3 3 Gastrooesophageal reflux 35 0 0 0 3 3 disease 36 Leukocytosis 3 0 0 0 3 37 Musculoskeletal pain 0 0 3 0 3 38 Oesophagitis 0 0 2 1 3 Sinusitis 1 0 2 0 3 39 Urticaria 0 0 0 3 3 40 41 42 18 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 39 of 101 BMJ

1 2 3 Terminology compliant with CTCAE 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 19 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 40 of 101

1 2 3 References: 4 1. Hayden JA, Chou R, Hogg-Johnson S, Bombardier C. Systematic reviews of low back pain prognosis had variable methods and results—guidance for 5 future prognosis reviews. Journal of Clinical Epidemiology 2009;62:781-96.e1. 6 2. Bråten LC RM, Espeland A, Storheim K, Zwart JA, Hellum C, Aßmus J. Statistical analysis plan (SAP) for clinical outcomes in the AIM-study: A 7 randomized trial of antibiotic treatment in patients with chronic low back pain and Modic Changes (the AIM-study). ClinicalTrials.gov2018. 8 Confidential: For Review Only 9 3. Hernán MA, Robins JM. Per-Protocol Analyses of Pragmatic Trials. [S.l.]2017:1391-8. 10 4. Grotle M, Vøllestad M, Brox M. Cross-cultural adaptation of the Norwegian versions of the Roland-Morris Disability Questionnaire and the Oswestry 11 Disability Index. Journal of Rehabilitation Medicine 2003;35:241-7. 12 5. Chiarotto A, Boers M, Deyo RA, et al. A CORE OUTCOME MEASUREMENT SET FOR LOW BACK PAIN CLINICAL TRIALS. Orthopaedic Proceedings 13 2018;100-B:29-. 14 6. Fairbank JC, Pynsent PB. The Oswestry Disability Index. Spine 2000;25:2940. 15 7. Dworkin HR, Turk CD, Farrar TJ, et al. Core outcome measures for chronic pain clinical trials: IMMPACT recommendations. Pain 2005;113:9-19. 16 8. EuroQol - a new facility for the measurement of health-related quality of life. Health policy 1990;16:199-208. 17 18 9. Derogatis LR, Lipman RS, Rickels K, Uhlenhuth EH, Covi L. The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav Sci 19 1974;19:1-15. 20 10. Waddell G, Newton M, Henderson I, Somerville D, Main CJ. A Fear-Avoidance Beliefs Questionnaire (FABQ) and the role of fear-avoidance beliefs in 21 chronic low back pain and disability. Pain 1993;52:157-68. 22 11. Eriksen HR, Ihlebaek C, Ursin H. A scoring system for subjective health complaints (SHC). Scandinavian journal of public health 1999;27:63. 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 20 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 41 of 101 BMJ

1 2 3 4

5 6 7 8 9 10 Antibiotic treatment in patients with chronic low back pain 11 12 Confidential:and Modic For Changes: Review Only 13 14 a randomized controlled trial. 15 16 17 18 19 20 21 22 23 24 Protocol Identification Number: The AIM-Study (Antibiotics In Modic changes) 25 EudraCT Number: 2013-004505-14 26 27 28 29 SPONSOR: Oslo University Hospital 30 31 32 33 COORDINATING John-Anker Zwart, professor dr.med 34 INVESTIGATOR: 35 FORMI, OUS Ullevål, Bygg 37B, 36 Postboks 4956 Nydalen, 0424 Oslo 37 Tel : 22 11 86 21 38 39 E-mail: [email protected] 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 1 of 41 BMJ Page 42 of 101

1 2 3 CONTACT details 4 5 6 Project managing: FORMI, OUS Ullevål 7 8 Att: Kjersti Storheim, PT, PhD 9 Bygg 37B, Postboks 4956 Nydalen, 0424 Oslo 10 Tel : 22117740 11 12 Confidential:E-mail:[email protected] For Review Only 13 Coordinating John-Anker Zwart, professor dr.med 14 Investigator: 15 FORMI, Bygg 37B, Postboks 4956 Nydalen, 0424 Oslo 16 Tel : 22 11 86 21 17 E-mail: [email protected] 18 19 20 Principal Lars Grøvle, MD, PhD 21 Investigator: Revmatologisk avdeling, Sykehuset Østfold, avd. Moss, 1603 Fredrikstad 22 23 Tel: 08600 24 E-mail: [email protected] 25 26 27 Principal Jan-Sture Skouen, professor dr.med 28 Investigator: Nakke- og ryggpoliklinikken, Helse Bergen HF Haukeland 29 universitetssjukehus, Postboks 1400, 5021 Bergen 30 Tel: 93116344 31 32 E-mail: [email protected] 33 34 Principal Øystein P Nygaard, professor dr.med 35 Investigator: 36 Nasjonalt Senter for Spinale Lidelser, St. Olavs Hospital, Postboks 3250 37 Sluppen, 7006 Trondheim 38 Tel: 72822353 39 40 E-mail: [email protected] 41 Principal Audny Anke, MD, PhD 42 Investigator: Fysikalskmedisinsk avdeling, Universitetssykehuset i Nord-Norge, 43 Rehabiliteringsklinikken, Sykehusveien 38, 9038 Tromsø 44 45 Tel: 07766 46 E-mail: [email protected] / [email protected] 47 48 Principal Anne Froholdt, MD, PhD 49 Investigator: Fysikalskmedisinsk poliklinikk, Drammen Sykehus, Avdeling for Nevrologi, 50 revmatologi og reHabilitering, 3004 Drammen 51 Tlf: 32 80 30 00 52 E-mail: [email protected] 53 54 Jens Ivar Brox, professor dr.med 55 Principal 56 Investigator: Avdeling for fysikalskmedisin og rehabilitering, Oslo Universitetssykehus 57 Ullevål, Postboks 4956 Nydalen, 0424 Oslo 58 Tlf: 23076029 59 E-mail: [email protected] 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 2 of 41 Page 43 of 101 BMJ

1 2 Imaging: Ansgar Espeland, professor dr.med 3 4 Radiologisk avdeling, Helse Bergen HF Haukeland universitetssjukehus, 5 Postboks 1400, 5021 Bergen 6 Tel: 90595437 7 8 E-mail: [email protected] 9 Infection medicine: Olav Lutro, MD 10 11 Infeksjonsseksjonen, Medisinsk avdeling, Haukeland Universitetssykehus, 12 Confidential:Postboks 1400, 5021 For Bergen. Review Only 13 Tlf: 41019349 14 E-post: [email protected], evt [email protected] 15 16 Statistics: Jörg Aßmus, PhD 17 Kompetansesenter for klinisk forskning, Armauer Hansens hus, 3.etasje, 18 Haukeland universitetssykehus, N-5021 Bergen 19 Tel: 55975539 20 E-mail: [email protected] 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 3 of 41 BMJ Page 44 of 101

1 2 3 4 Signature page 5 6 Title Antibiotic treatment in patients with chronic low back pain and Modic Changes: a 7 randomized controlled trial. 8 9 10 Protocol ID The AIM-Study (Antibiotics In Modic changes) 11 no: 12 Confidential: For Review Only

13 14 EudraCT no: 2013-004505-14 15 16 17 I hereby declare that I will conduct the study in compliance with the Protocol, ICH GCP and the 18 19 applicable regulatory requirements: 20 Name Title Role Signature Date 21 John-Anker Professor Coordinating investigator 22 Zwart dr.med 23 Kjersti Storheim PT, PhD Project manager 24 25 26 Lars Grøvle MD, PhD Principal Investigator, 27 Sykehuset Østfold, avd. 28 Moss 29 Jan Sture Skouen Professor Principal Investigator, 30 dr.med Haukeland 31 universitetssjukehus 32 33 Bergen 34 Øystein P Professor Principal Investigator, St. 35 Nygaard dr.med Olavs Hospital, 36 Trondheim 37 Audny Anke MD, PhD Principal Investigator, 38 Universitetssykehuset i 39 40 Nord-Norge, Tromsø 41 Anne Froholdt MD, PhD Principal Investigator, 42 Drammen Sykehus 43 Jens Ivar Brox Professor Principal Investigator, 44 dr.med Oslo Universitetssykehus 45 Ullevål 46 47 Ansgar Espeland Professor National responsibility 48 dr.med for imaging, Haukeland 49 universitetssjukehus 50 Bergen 51 Olav Lutro MD National responsibility 52 for infection medicine, 53 54 Haukeland 55 universitetssjukehus 56 Bergen 57 Jörg Aßmus PhD Statistician, Haukeland 58 universitetssjukehus 59 Bergen 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 4 of 41 Page 45 of 101 BMJ

1 2 3 PROTOCOL SYNOPSIS 4 5 6 Antibiotic treatment in patients with chronic low back pain and Modic Changes: a randomized 7 controlled trial. 8 9 10 11 Sponsor Oslo University Hospital 12 Confidential: For Review Only 13 14 Phase and study type Phase III, therapeutic confirmatory 15 16 Investigational Medical Active substance: Amoxicillin (Amoksicillintrihydrat) (Amimox 17 Product (IMP) «Meda») tablets 750 mg 1x3 for (oral intake). 18 19 Placebo: Maize starch 20 21 22 Centers: Oslo University Hospital Ullevål, Sykehuset Østfold Moss, Haukeland 23 University Hospital Bergen, St.Olavs Hospital Trondheim, the 24 University Hospital of Northern-Norway Tromsø and Drammen 25 Hospital. 26 27 Study Period: Estimated date of first patient enrolled: March 15th 2015 28 29 Anticipated recruitment period: 2 years 30 31 Estimated date of last patient completed: March 15th 2018 32 33 34 Treatment Duration: 100 days 35 36 Follow-up: 1 year 37 38 39 Objectives & Endpoints The overall aim of the present project is to re-examine the finding of a 40 recent randomized controlled trial (RCT) that antibiotic treatment can 41 cure patients with chronic low back pain (LBP), a former disc 42 herniation and present Modic Changes (MCs) type I. Our trial will be 43 the first to re-examine this finding. We also want to add important new 44 knowledge to the research field beyond the only former RCT by 45 46 broadening the inclusion criteria to include both patients with type I 47 and type II MCs, improving the MRI assessment of MCs, further 48 clarifying the pathogenesis of MCs by studying genetic variability, 49 gene and protein expression of inflammatory biomarkers, and 50 conducting health economic analysis. 51 52 Main objective and endpoint: To evaluate the effect of Amoxicillin 53 versus placebo on disease-specific disability evaluated by the Roland 54 Morris Disability Questionnaire (RMDQ) at one year (12 months) 55 follow-up in patients with chronic LBP and MCs type I or II adjacent 56 to a previously herniated disc (hypothesis A). 57 58 59 Secondary objective (SO) and endpoint: 60  SO 1. To evaluate the effect of Amoxicillin versus placebo on

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 5 of 41 BMJ Page 46 of 101

1 2 RMDQ at one year (12 months) follow-up separately in patients 3 with type I and type II MCs, respectively (hypotheses B and C). 4 5 Exploratory and key supportive objectives and endpoints (KSOs) 6  KSO 2. To evaluate the effect of Amoxicillin versus placebo on 7 Oswestry Disability Index (ODI) at one year (12 months) follow- 8 up in the whole cohort of included patients (hypothesis D). 9 10  KSO 3. To evaluate the effect of Amoxicillin versus placebo on 11 LBP intensity at one year (12 months) follow-up in the whole 12 Confidential:cohort of included For patients Review (hypothesis OnlyE). 13 14  KSO 4. To evaluate whether the short tau inversion recovery 15 (STIR) signal (intensity and extent) of MCs on baseline MRI 16 predicts RMDQ score at one year (12 months) follow-up 17 (hypothesis F). 18 19  KSO 5. To assess whether change in STIR signal (intensity and 20 extent) of MCs from baseline to one year (12 months) follow-up is 21 related to RMDQ score at one year (12 months) follow-up 22 (hypothesis G). 23 24  KSO 6. To evaluate the effect of Amoxicillin versus placebo on 25 health-related quality of life (the EQ-5D) at one year (12 months) 26 follow-up in the whole cohort of included patients (hypothesis H). 27  To evaluate cost-effectiveness of Amoxicillin versus placebo at 28 one year (12 months) follow-up in the whole cohort of included 29 30 patients. 31  To evaluate whether positive pain provocation tests at baseline 32 predicts RMDQ score at one year (12 months) follow-up. 33 34  To evaluate the difference in incidence of AEs and SAEs between 35 the two intervention groups from inclusion to one year (12 months) 36 follow-up in the whole cohort of included patients. 37  To evaluate, separately in the two intervention groups, whether 38 lack of a clinically important improvement in RMDQ, ODI, and 39 LBP intensity, respectively, from baseline to post-treatment (100- 40 41 das after start of treatment) is associated with lack of a clinically 42 important improvement in these outcomes from baseline to one- 43 year (12 months) follow-up. 44  Further clinical objectives and endpoints: To evaluate the effect of 45 Amoxicillin versus placebo on: 46 o RMDQ at one year (12 months) follow-up separately in 47 48 patients previously undergoing back surgery for disc 49 herniation and patients NOT previously undergoing back 50 surgery for disc herniation. 51 o other outcome measures not mentioned above (leg pain 52 intensity, hours with LBP during the last 4 weeks, 53 54 bothersomeness, days with sick leave, co-interventions, 55 patients’ satisfaction, global perceived effect) at one year 56 (12 months) follow-up in the whole cohort of included 57 patients 58 59 o secondary outcome measures at one year (12 months) 60 follow-up separately in patients with type I and type II MCs at baseline, respectively

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 6 of 41 Page 47 of 101 BMJ

1 2 o primary and secondary outcomes measures at 100-days 3 follow-up in the whole cohort and separately in patients 4 with type I and type II MCs at baseline, respectively. 5 6  Further radiological objectives and endpoints: To assess whether 7 characteristics of MCs on baseline MRI predict ODI score or LBP 8 intensity at one year (12 months) follow-up. To compare change in 9 characteristics of MCs from inclusion to one year (12 months) 10 follow-up between treatment groups, and to assess whether this 11 12 Confidential:change in MCsFor is related Review to RMDQ score, Only ODI score, and pain 13 intensity at to one year (12 months) follow-up. To determine the 14 reliability of different MCs characteristics by different MRI 15 methods. To assess the relationships of these MC characteristics to 16 each other and to clinical variables. 17 18  Genetic objectives and endpoints: To investigate the effect of 19 Amoxicillin on epigenetic patterns, longitudinal gene- and protein 20 expression, genetic variation, from baseline to post-treatment (100 21 days after start of treatment) and from baseline to one year (12 22 months) follow-up in patients with MCs type I or II, and to 23 evaluate correlations with clinical data. To investigate the effect of 24 25 Amoxicillin on bowel flora, resistant bacteria and resistance genes. 26 27 28 29 Declaration of success: When the primary hypothesis is proven. 30 31 32 33 34 Study Design: A randomized placebo-controlled and double-blinded multicenter 35 study 36 37 Inclusion Criteria:  Age between 18 and 65 years 38  LBP of > 6 months duration in the area below the 12th rib and above the 39 gluteal folds with a Numerical Rating Scale (NRS) pain intensity score of 40  5 (mean of three NRS scales; current LBP, the worst LBP within the 41 last 2 weeks, and usual/mean LBP within the last 2 weeks). 42 43  MRI-confirmed lumbar disc herniation within the preceding 2 years 44  MC type I and/or type II in the vertebral body marrow at the same level 45 as the previously herniated disc. For patients with former surgery for disc 46 herniation, the MC has to be located at an operated level. 47 48  Written informed consent 49 Exclusion Criteria  Allergy to penicillin or cefalosporins 50  Allergy/hypersensitivity to any of the excipients of the study drug 51 52  Current pregnancy or lactation 53  Elevated kidney (creatinine) or hepatic (ALAT/ASAT) values outside 54 normal range 55 56  Phenylketonuria (Følling disease) 57  Mononucleosis or leukaemia 58  Any specific diagnosis that may explain patient’s low back symptoms 59 (e.g. tumor, fracture, spondyloarthritis, infection, spinal stenosis). 60  Former low back surgery (L1-S1) for other reasons than disc herniation

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 7 of 41 BMJ Page 48 of 101

1 2 (e.g. fusion, decompression, disc prosthesis) 3  Former surgery for disc herniation, but < 12 months have elapsed since 4 surgery. 5 6  Former surgery for disc herniation, but MC located at non-operated 7 level(s) only. 8  Reservation against intake of gelatine (the capsules contains gelatine, 9 which among other things is produced by ingredients from pigs) 10 11  Regular use of glucocorticoids 12 Confidential: Regular use Forof opioids Review with the exception ofOnly codeine and tramadol 13  Not understanding Norwegian language 14 15  Unlikely to adhere to treatment and/ or complete follow-up (e.g ongoing 16 serious psychiatric disease, drug abuse, plans to move) 17  Antibiotic treatment within the preceding one month before treatment 18 start 19  Contraindications to MRI (e.g. cardiac pacemaker electrodes, metal 20 21 implant in eye or brain, claustrophobia). 22  Unwilling to participate 23 24 25 Sample Size: 160 patients 26 27 Efficacy Assessments: Main outcome measure: the Norwegian version of the disease-specific 28 Roland and Morris Disability Questionnaire (RMDQ), a patient 29 reported questionnaire measuring pain related disability. 30 31 Secondary outcome measures: Clinical: ODI, LBP intensity and 32 health-related quality of life (the EQ-5D). Radiological: STIR signal 33 34 of MCs. 35 36 Other outcome measures (to be used in supportive analysis): Clinical: 37 leg pain intensity, hours with LBP during the last 4 weeks, 38 bothersomeness, days with sick leave, patients’ satisfaction, global 39 perceived effect, and co-interventions. Fecal sample. Health economy: 40 the RMDQ, and the EQ-5D. Radiological: characteristics of MCs other 41 42 than STIR signal. Genetic: epigenetic patterns, genetic variation, gene- 43 and protein expression. 44 45 Safety Assessments: Haematological parameters (leucocytes, thrombocytes, eosinophils, 46 haemoglobin (Hb) and hematocrit (Ht)) and measures of kidney 47 (creatinine) and liver function (ASAT / ALAT), will be assessed and 48 registered monthly during the intervention period, together with a short 49 clinical evaluation (blood pressure, pulse, auscultation of hearth and 50 51 lunges) to monitor side effects. Adverse events (AEs) will be 52 registered at all study visits. 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 8 of 41 Page 49 of 101 BMJ

1 2 1. INTRODUCTION AND SIGNIFICANCE 3 Low back pain (LBP) is the number one cause for living with disability globally and expenses for 4 the Norwegian society is estimated to reach at least NOK 24 billions annually. The origin and 5 pathogenesis of most LBP is unknown and guidelines endorse an approach where most patients are 6 7 viewed within a biopsychosocial model and are not given a patho-anatomical diagnosis. A recent 8 Danish study published during spring 2013 advances a remarkable cause and treatment for a 9 selected group of patients who suffer persistent disability and have focal vertebral bone marrow 10 changes (Modic Changes) on magnetic resonance imaging (MRI). The study was based upon a 11 hypothesis that chronic LBP may arise from a low-grade infection of the intervertebral disc by the 12 anaerobic ProprioneConfidential: acne bacteria, resulting Forin pain andReview bone edema (Modic Only Changes). In a 13 randomized placebo controlled trial it was concluded that 100 days of antibiotic treatment with 14 15 amoxicillin clavulanate tablets three times a day was significantly more effective than placebo in all 16 primary and secondary outcomes. The results were met by media frenzy, launched as a medical 17 breakthrough and evoked parallels to the Nobel Prize-winning discoveries that a chronic 18 Helicobacter pylori infection of the stomach lining causes 80 to 90 percent of ulcers. Presently, 19 criticism has been multifarious reaching from the risky antibiotic side effects in individuals and the 20 spread of antibiotic resistance, methodological issues and to potential conflicts of interest. 21 22 23 Nonetheless, referring to the large problem of chronic LBP, the finding that antibiotic treatment can 24 cure a sub-group of chronic LBP patients with Modic Changes should be re-assessed in well- 25 designed trials, of which our trial will be the first. The present MRI evaluation of Modic Changes 26 has important limitations. We will therefore also refine the characterization of Modic Changes by 27 using further MRI methods, and assess whether characteristics of Modic Changes affect clinical 28 course or treatment effect. In addition, we will study genetic variation, longitudinal gene and 29 30 protein expression of inflammatory biomarkers to further clarify the pathogenesis of Modic 31 Changes, and conduct health economic analysis. Our research will improve the clinical and 32 radiological management of patients with chronic LBP. To prevent inappropriate clinical use of 33 antibiotics in this large and challenging patient population based on a single trial, a new trial is 34 urgently needed. Our trial is rooted in a well-functioning national research network, and it will 35 strengthen Norway’s position internationally within the field of LBP research. This project will 36 promote scientific renewal, develop research-based knowledge about human health and disease, and 37 38 generate new knowledge on issues relevant to the society. 39 40 41 2. BACKGROUND 42 43 2.1 BACKGROUND DISEASE 44 45 About 70 – 85 % of people report LBP during adult life, and about 10% develop chronic LBP (i.e. 46 experience pain for longer than 3 months). Globally, LBP is ranked the number one cause for living 88 47 with disability , and patients with chronic LBP have a higher consummation of health- and welfare 48 services than most other groups of patients47;90. After two years out of work, less than 1% returns to 49 ordinary work44. The costs for the individuals and the society are huge and are estimated to reach at 50 least NOK 24 billions annually 59. 51 52 53 Over the last three decades there has been a major shift in the clinical and research approach to 54 LBP. Prior to this, clinical practice and research were mainly based upon a biomedical model with 55 patients receiving specific pathoanatomical diagnoses and treatments directed to these diagnoses. 56 An example of this type of thinking is Mixter and Barr’s 1934 classic paper describing surgical 57 treatment of disc prolapse64. This approach was challenged in 1987 by two landmark publications 58 by Gordon Waddell89 and The Quebec Task Force on spinal disorders81 that pointed out that most 59 LBP diagnoses were nominal and of doubtful validity. The “biopsychosocial model” of LBP, 60 “simple back pain” and “non-specific spinal disorders” have been standard approaches over the last

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 9 of 41 BMJ Page 50 of 101

1 2 decades, and clinical guidelines promote an approach where 85-90% of patients do not receive a 3 patho-anatomical diagnosis45;55. Since a patho-anatomic diagnosis is not pursued, clinicians apply 4 generic symptomatic treatments such as advice to stay active and avoid bed-rest, analgesic 5 medicines, exercise and manipulation. However, the many systematic reviews of treatments for 6 13;30;54 7 LBP reveal that existing treatments for non-specific LBP have, at best, only small effects . 8 9 In most other areas of medicine, a specific diagnosis is considered the cornerstone of effective 10 management. Hence, one explanation for ineffective treatments for non-specific LBP may be that 11 clinicians are unable to direct treatment to the specific pathology underlying back pain and instead 12 rely upon genericConfidential: treatments in heterogeneous For patients Review45. A focus of today’s Only clinical research 13 regarding LBP is therefore on identifying subgroups of patients with a (un)favourable prognosis or 14 20;86 15 likely to respond favourably to specific treatments . The classification of these subgroups, 36;45 16 however, is seldom based on patho-anatomical findings or diagnoses . The recent Danish study 17 treating a sub-group of patients with a supposed biomedical source of LBP (Modic Changes) with 18 antibiotics, reporting substantially greater effect than all currently established treatments12, might 19 therefore be a potentially important breakthrough in the understanding and management of LBP. 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 Figure 1 – MRI of the lumbar spine showing Modic Changes type II (white arrows) adjacent to a herniated disc (black arrow); type 39 II Modic changes (fatty type) are hyper-intense on T1 weighted (left) and T2 weighted (right) sequences (see text below). 40 41 42 Modic changes 43 The vertebral bone marrow changes known as Modic Changes (MCs, Figure 1) were first described 44 in 1987/8827;66 and have received considerable attention in recent years. MCs are classified into type 45 I, II and III66. On histological examination, type I MCs show disrupted and fissured endplates with 46 regions of degeneration, regeneration, reactive bone formation, endplate edema and vascular 47 65;66 48 granulation , type II MCs display endplate disruption and fatty degeneration, and type III MCs 66 49 contain sclerosis . The different types of MCs are defined based on T1- and T2 weighted magnetic 50 resonance imaging (MRI), and are visible in the endplates and vertebrae below and above the actual 51 disc85. Type I (edema type) is defined by a hypo-intense signal in T1 weighted MRI and hyper- 52 intense signal on T2 weighted MRI, type II (fatty type) by hyper-intense signal on T1- and T2 53 weighted MRI, and type III (sclerotic type) by hypo-intense signal on T1- and T2 weighted MRI. 54 55 Over months or years type I may change to normal or type II, and type II may change to normal or 56 type III. Other developments between types and combined types are also found. The different types 66;77 57 are thought to reflect a common underlying process . The prevalence of all types of MC has been 58 reported to be about 10% in populations without LBP and between 16% and 62% in LBP patients94. 59 Several studies have established an association between LBP and MCs with Odds Ratios (ORs) 60 from 2.0 to 19.952.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 10 of 41 Page 51 of 101 BMJ

1 2 The aetiology and pathogenesis of MCs is unclear. Some have described a possible relationship 3 between the degenerative disc, including the MCs, and the forces acting within the disc3;4;9;61. The 4 fact that MC type I has been found to convert to normal or to MC type II after fusion supports a 5 mechanical component in the development of MCs60;87. It has also been postulated that nucleus 6 25 7 tissue entering the vertebrae can cause an autoimmune reaction . Burke et al discovered a higher 8 level of pro-inflammatory mediators (IL-6, IL-8, PGE2) in patients with discogenic LBP compared 9 to those with sciatica, suggesting that production of proinflammatory mediators within the nucleus 10 pulposus may be a major factor in the genesis of a painful lumbar disc23. This theory may be 11 supported by a placebo-controlled trial reporting favourable results of intradiscal methylene blue 12 injections inConfidential: patients with chronic discogenic For LBP (methylene Review blue is an Only inhibitor of nitric oxide- 13 induced guanalyte cyclase, and nitrix oxide has been implicated in the inflammatory processes of 14 23;75 15 disc degeneration and discogenic pain) . Ohtori et al found PGP 9.5-immunoreactive nerve fibers 16 and TNF-immunoreactive cells in the endplates from patients with MCs; significantly more than in 17 patients without MCs. They concluded that endplate abnormalities are related to inflammation and 18 axon growth induced by TNF, and postulated that TNF expression and PGP 9.5-positive nerve in- 19 growth in abnormal endplates may be a cause of LBP71. 20 21 22 2.2 BACKGROUND THERAPEUTIC INFORMATION 23 A current hypothesis is that MCs may occur due to edema surrounding an infected disc; i.e. the 24 infection is in the disc and the MC is a secondary change in the bone due to cytokine and propionic 25 acid production12. Stirling et al took biopsies from nucleus material and found infection with low 26 virulent anaerobic organisms (propionibacterium acnes (P.acnes) and corynebacterium propinquum) 27 in 53% of the patients with lumbar disc herniation, compared to none of the patients operated on for 28 other spinal disorders83. This was later replicated by Corsia et al24. Fritzell et al detected bacterial 29 39 30 DNA in degenerative discs of patients operated for disc herniation or post-discectomy syndrome , 31 and Agarwal et al found positive bacterial cultures (mainly P.acnes) in disc material from patients 8 32 operated on for disc herniation . Synchronous with their publication of the randomized controlled 33 trial (RCT) of antibiotic treatment of patients with MCs12, the Danish research group published a 34 study of nucleus material harvested from 61 patients undergoing primary surgery for lumbar disc 35 herniation10. Microbiological cultures were positive in 46 % of the patients. Anaerobic cultures 36 were positive in 43 %, and of these 7 % had dual microbial infections, containing both one aerobic 37 38 and one anaerobic culture. Further, at discs with a nucleus with anaerobic bacteria, 80 % developed 39 new MCs in the vertebrae adjacent to the previous disc herniation (OR 5.6), but none of those with 40 aerobic bacteria and only 44 % of patients with negative cultures developed new MCs. It was 41 concluded that this findings supports the theory that the occurrence of MCs in the vertebrae 42 adjacent to a previously herniated disc may be due to edema surrounding an infected disc10. Also 43 the results from their uncontrolled pilot study of antibiotics in 32 patients who had MCs type I and 44 chronic LBP following a lumbar disc herniation may indicate a link between MCs and bacterial 45 11 46 infection . 47 48 It is not clear how these low virulent anaerobic organisms gain access to the disc. Albert et al argue 49 that these anaerobic mouth and skin commensal organisms most likely enter the disc and give rise 50 to a slowly developing infection during normal bacteraemia as a result of the neovascularisation 51 associated with disc degeneration or herniation10;12. However, all biological material analysed for 52 bacteria is harvested from patients having surgery and intraoperative contamination can not be 53 7 54 excluded . It is also not clear how nucleus pulposus and/or cytokine and propionic acid following a 55 disc infection leak into the endplates and become manifested as MCs in the bone. Yet, structural 56 defects in the nucleus and endplates occur as early as 15 years of age19, and disc degeneration 57 implies an aberrant, cell-mediated response to progressive structural failure5. One may therefore 58 hypothesize that structural disc and endplate defects allow spread of disc- and/or infection material. 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 11 of 41 BMJ Page 52 of 101

1 2 One concern with long-term treatment with Amoxicillin is the side effects, of which the main 3 concern is the effects of the natural bacterial flora and the emergence of resistant bacteria. It has 4 been shown that a course of Amoxicillin will lead to considerable change of the microflora of the 5 intestines and emergence of resistant enterobacterial species84. There is also evidence of an ability 6 of resilience, ie a host ability to return to original flora weeks or months after ended treatment 7 26 8 period . However, most of these studies have been done on short courses of amoxicillin, and there 9 is little knowledge of the effect of long-term treatment with amoxicillin on the human intestinal 10 microbiota. 11 12 2.3 PRE-CLINICALConfidential: & CLINICAL EXPERIENCES For Review WITH INVESTIGATIONAL Only 13 14 MEDICINAL PRODUCT 15 Two former clinical studies, both conducted by the same research group, have evaluated the effect 16 of antibiotic treatment for a selected group of patients with chronic pain and MCs in the vertebrae 17 adjacent to a previous disc herniation. The first study was an uncontrolled pilot study reporting a 18 clinically important and statistically significant (p 0.001) improvement in all outcome measures 19 (LBP intensity, number of days with pain, disease-specific and patient-specific function, and global 20 perceived effect) at the end of treatment, and at long-term follow-up11. The second study was a RCT 21 concluding that the antibiotic protocol was significantly more effective for this group of patients 22 12 23 than placebo in all the primary and secondary outcomes . 24 25 Based on the before outlined hypothesis that a low-grade infection with low virulent anaerobic 26 organisms in the herniated disc, mainly with P. acnes, causes edema in the adjacent vertebra, the 27 intervention in the Danish studies was amoxicillin-clavulanate. In the RCT they used 500/125 mg 28 12 29 1x3 or 2x3 for 100 days (for low- and high-dose study arm, respectively) . They reported a clear 30 trend towards better outcome for high-dose amoxicillin- clavulanate, but not significant. However, 31 the study was not designed for testing dose-response and results were merged for low- and high- 32 dose amoxicillin- clavulanate12. In Norway, amoxicillin is only available without added clavulanate, 33 and European culture studies of P.acnes do not suggest beta-lactamase69. Adding clavulanate also 34 increases risk of side effects. In order to re-examine the Danish RCT, reporting substantially greater 35 effect than all currently established treatments, we will give Amoxicillin (Amoksicillintrihydrat) 36 37 Amimox «Meda» tablets 750 mg 1x3 for 100 days (oral intake). The 24-hour dose is within the 38 antibiotic treatment recommendations described in “Produktresumé“ by the Swedish Medical 39 Products Agency (Lëkemedelsverket) and the long duration is in line with antibiotic treatment 40 commonly prescribed for post-operative discitis. Further, in the Danish RCT the effect took 6-8 41 weeks to manifest, which is consistent with the clinical course of resolving infection in poorly 42 vascularised infected tissue, like the intervertebral disc12. 43 44 45 2.4 RATIONALE FOR THE STUDY AND PURPOSE 46 The finding that antibiotic treatment is effective for selected patients with chronic LBP and MCs 47 should be re-examined in independent and well-designed RCTs, for several reasons: 48 49 First; chronic LBP is a large problem and many LBP patients have MCs on MRI. Antibiotic 50 treatment has side effects in individuals and use of antibiotics in the large sub-group of chronic LBP 51 52 patients with MCs may spread antibiotic resistance. Hence, before this sub-group receives antibiotic 53 treatment in clinical practice, the effect of the treatment should be confirmed in further RCTs. 54 Presently, we note a growing demand for antibiotic treatment of LBP patients and a new RCT 55 should be performed urgently to prevent inappropriate use of antibiotics based on a single trial. 56 57 Second; the MRI-based selection of patients for antibiotic treatment is not without problems. The 58 type of MCs (especially type I vs. II) has been used to guide on treatment12;17;68. Only patients with 59 12 60 type I MCs were included in the Danish RCT on antibiotic treatment for chronic LBP . No studies have examined whether such treatment is effective for patients with type II MCs. The

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 12 of 41 Page 53 of 101 BMJ

1 2 differentiation between type I and II may be less relevant, since the different types of MCs may 3 represent different stages of a common process66;77. Also, the MRI evaluation of MCs is 4 challenging. Different observers may report up to twofold different prevalence of MCs (any type) at 5 a specific endplate (e.g. 25% vs. 50%) despite of good interobserver agreement based on kappa 6 values15. Compared to 0.2 Tesla MRI (used in the Danish RCT), 1.5 Tesla MRI (commonly used in 7 14 8 clinical practice) may reveal more MCs especially type II but relatively fewer type I . The 9 conventional T1- and T2 weighted sequences used to define MCs are less likely to separate fat (type 10 II) from water (type I) than are fat-water separation MRI methods62;63 that are not routinely applied 11 in spine imaging. Diffusion weighted MRI shows diffusion of water molecules; it can be used to 12 differ osteoporoticConfidential: from neoplastic spine fractures For76 andReview might provide insightOnly into MCs as well34. 13 Separation of pure edema from a more “active” lesion would require further MRI methods to assess 14 15 contrast enhancement over time; such methods are used in rheumatologic studies and might be 16 16 useful also in spine imaging research . The size of the MCs must be consistently evaluated to 17 assess its relation to treatment effect. Again, the prevalence of different sizes may differ between 18 observers15. Unclear differentiation between different MC types in the same lesion makes it difficult 19 to determine the size of each type. Clearly, the MRI assessment of MCs should be improved, so that 20 patients can be classified more consistently according to the underlying lesions, especially when the 21 22 result is used in treatment decisions. In pilot studies, we have developed protocols for fat-water 23 separation MRI, diffusion weighted MRI, and dynamic contrast enhanced (DCE-) MRI of MCs. 24 25 Third; since biological material analysed for bacteria is harvested from patients having surgery; we 26 cannot rule out that the bacteria found in the nuclear material may be due to intraoperative 27 contamination rather than infection7. Also, the lack of biological material proving bacterial infection 28 in non-operated patients implies that other, strong indications of a bacterial infection are needed, 29 e.g. a new RCT confirming the effect of antibiotic treatment. Albert et al included analysis of serum 30 12 31 inflammatory markers but did not detect any differences between groups at end of treatment . 32 However, herniated or degenerated discs and the subsequent leakage of nucleus pulposus may 33 initiate profound, complex immunological- and inflammatory processes. Inflammatory factors 34 including cytokines, such as interleukin 1β (IL-1β), interleukin 6 (IL-6), tumor necrosis factor 35 (TNF), and transforming growth factors (TGF) could stimulate an on-going systemic inflammation 36 and affect pain sensitization. Moreover, high levels of IL-6 in serum have recently been suggested 37 80 38 to be a predictor for slow recovery in patients with lumbar radicular pain due to disc herniation . 39 Hence, there is accumulating evidence that cytokines are involved in the pathophysiology of 40 sciatica and low-back pain, but there is conflicting evidence with respect to the effect of TNF alfa 41 inhibitors in clinical trials67. It seems, however, in experimental studies that a combination of 42 cytokine inhibitors reduces the effects on nerve conduction velocity that might be of relevance for 43 future clinical trials for the treatment of sciatica and LBP72. It is accordingly of relevance to assess 44 45 whether persistent chronic LBP possibly caused by a low grade inflammation, is reflected by an 46 altered gene or protein expression of inflammatory biomarkers in these patients and if there is a 47 correlation between such expression and clinical improvement. 48 49 Previous studies have indicated that type II MCs are associated with genetic variation in the 50 interleukine-1A (IL-1A) and matrix metalloproteinases-3 (MMP-3) genes, which are involved in 51 disc degeneration40;53. Furthermore, it has been demonstrated that genetic variability in genes 52 encoding proteins relevant for pain processing (COMT, OPRM1) or disc degeneration (MMP) may 53 48;49;73 54 be important for the recovery after disc herniation . PARK2, a novel genetic variant, has 93 55 recently been demonstrated to be associated with lumbar disc degeneration . To further clarify the 56 pathogenesis of MCs we want to emphasize the role of different genetic variants and the association 57 with the clinical outcome after treatment. 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 13 of 41 BMJ Page 54 of 101

1 2 Fourth; new treatment methods are not only a clinical, but also an economic challenge for the 3 health care services56. We will carry out a cost-effectiveness evaluation for the main outcome and 4 for health-related quality of life by EQ5D after one year (12 months) follow-up. 5 6 7 8 3. STUDY OBJECTIVES AND RELATED ENDPOINTS 9 Our overall objective is to re-examine the finding that antibiotic treatment can cure patients with 10 chronic LBP and MCs adjacent to a previously herniated disc. Our trial will be the first to re- 11 examine this finding. We also want to add important new knowledge to the research field beyond 12 the Danish RCTConfidential: by broadening the inclusion For criteria toReview include both patients Only with type I and type II 13 MCs, improving the MRI assessment of MCs, further clarifying the pathogenesis of MCs by 14 15 studying genetic variability, gene and protein expression of inflammatory biomarkers, and 16 conducting health economic analysis. Our specific objectives are: 17 18 Main objective and endpoint: To evaluate the effect of Amoxicillin versus placebo on disease- 19 specific disability evaluated by the Roland Morris Disability Questionnaire (RMDQ) at one year 20 (12 months) follow-up in patients with chronic LBP and MCs type I or II adjacent to a previously 21 22 herniated disc (Hypothesis A). 23 24 Thus, our Main objective is to re-examine the clinical effect of antibiotic treatment reported in the 25 former Danish study from baseline to one year (12 months) follow-up. We will use the same 26 primary outcome measure (RMDQ), but the effect will be evaluated in patients with MCs type I or 27 MCs type II since, as outlined above, we argue that some MCs type I patients may be classified as 28 MCs type II patients and vice versa dependent on the magnet strength of MRI machines used, and 29 30 MCs type I and type II most likely represent a common process (that can be influenced by a 31 common treatment) (hypothesis A). As a secondary objective (SO 1) we evaluate the effect of 32 Amoxicillin versus placebo on RMDQ at one year (12 months) follow-up separately in patients with 33 type I and type II MCs, respectively (hypotheses B and C). 34 35 Exploratory and key supportive objectives (KSOs) and endpoints): 36 37  KSO 2. To evaluate the effect of Amoxicillin versus placebo on Oswestry Disability Index 38 (ODI) at one year (12 months) follow-up in the whole cohort of included patients (hypothesis 39 D). 40  KSO 3. To evaluate the effect of Amoxicillin versus placebo on LBP intensity at one year (12 41 42 months) follow-up in the whole cohort of included patients (hypothesis E). 43  KSO 4. To evaluate whether the short tau inversion recovery (STIR) signal (intensity and 44 extent) of MCs on baseline MRI predicts RMDQ score at one year (12 months) follow-up 45 (hypothesis F). 46 47  KSO 5. To assess whether change in STIR signal (intensity and extent) of MCs from baseline to 48 one year (12 months) follow-up is related to RMDQ score at one year (12 months) follow-up 49 (hypothesis G). 50 51  KSO 6. To evaluate the effect of Amoxicillin versus placebo on health-related quality of life 52 (the EQ-5D) at one year (12 months) follow-up in the whole cohort of included patients 53 (hypothesis H). 54 55  To evaluate cost-effectiveness of Amoxicillin versus placebo at one year (12 months) follow-up 56 in the whole cohort of included patients. 57  To evaluate whether positive pain provocation tests at baseline predicts RMDQ at one year (12 58 59 months) follow-up. 60  To evaluate the difference in incidence of AEs and SAEs between the two intervention groups from inclusion to one year (12 months) follow-up in the whole cohort of included patients.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 14 of 41 Page 55 of 101 BMJ

1 2  To evaluate, separately in the two intervention groups, whether lack of a clinically important 3 improvement in RMDQ, ODI, and LBP intensity, respectively, from baseline to post-treatment 4 (100-das after start of treatment) is associated with lack of a clinically important improvement 5 6 in these outcomes from baseline to one-year (12 months) follow-up. 7  Further clinical objectives and endpoints: To evaluate the effect of Amoxicillin versus placebo 8 on: 9 o RMDQ at one year (12 months) follow-up separately in patients previously undergoing 10 back surgery for disc herniation and patients NOT previously undergoing back surgery 11 for disc herniation. 12 Confidential: For Review Only 13 o other outcome measures not mentioned above (leg pain intensity, hours with LBP during 14 the last 4 weeks, bothersomeness, days with sick leave, co-interventions, patients’ 15 satisfaction, global perceived effect) at one year (12 months) follow-up in the whole 16 cohort of included patients 17 18 o secondary outcome measures at one year (12 months) follow-up separately in patients 19 with type I and type II MCs at baseline, respectively 20 o primary and secondary outcomes measures at posttreatment (100 days after start of 21 22 treatment) in the whole cohort and separately in patients with type I and type II MCs at 23 baseline, respectively. 24  Further radiological objectives and endpoints: To assess whether characteristics of MCs on 25 baseline MRI predict ODI score or LBP intensity at one year (12 months) follow-up. To 26 27 compare change in characteristics of MCs from inclusion to one year (12 months) follow-up 28 between treatment groups, and to assess whether this change in MCs is related to RMDQ score, 29 ODI score, and pain intensity at one year (12 months) follow-up. To determine the reliability of 30 different MCs characteristics by different MRI methods. To assess the relationships of these MC 31 characteristics to each other and to clinical variables. 32 33  Genetic objectives and endpoints: To investigate the effect of Amoxicillin on epigenetic 34 patterns, longitudinal gene- and protein expression, genetic variation, from baseline to post- 35 treatment (100 days after start of treatment) and from baseline to one year (12 months) follow- 36 up in patients with MCs type I or II, and to evaluate correlations with clinical data. To 37 investigate the effect of Amoxicillin on the faecal flora and the emergence of resistant bacteria 38 and resistance genes. 39 40 41 3.1 PRIMARY ENDPOINT 42 Pain and disability as measured by the Norwegian version of the disease-specific Roland and Morris 43 Disability Questionnaire (RMDQ)41;78, one year (12 months) after start of treatment in the ITT 44 45 population. 46 47 3.2 SECONDARY ENDPOINT 48 Pain and disability as measured by the Norwegian version of the disease-specific Roland and Morris 49 Disability Questionnaire (RMDQ)41;78, one year (12 months) after start of treatment in the MC type 50 I and II sub-groups. 51 52 53 3.3 EXPLORATORY ENDPOINTS; ODI, LBP intensity, STIR signal of MCs and health-related 54 quality of life (the EQ-5D), one year (12 months) after start of treatment. 55 56 3.4 OUTCOME MEASURES 57 Background variables. 58 Data will be collected at baseline only and will include sociodemographic data related to patient 59 age, gender, BMI, ethnicity, marital status, educational level, work status, physical work load, 60 leisure time activity, smoking habits, subjective health complaints (SHC)35, emotional distress

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 15 of 41 BMJ Page 56 of 101

1 2 (Hopkins Symptom Checklist–25 HSCL-25)29, fear-avoidance beliefs (FABQ work/physical 3 activity)91, LBP history/duration (including former treatment, e.g. surgery for disc herniation, 4 physiotherapy, chiropractic, etc), expectations about treatment effect and pain drawing (localized 5 versus widespread pain)58;70 In addition, routine clinical investigations (pain provocation tests 6 7 (springing test, active flexion / extension of the lumbar spine) and neurological tests (strength, toe- 8 /heel walking, sensibility, reflexes, straight leg raising test, reverse Lasegue test)) will be structured 9 and registered in a CRF at baseline for predictive purposes. 10 11 In order to check blinding, patients will be asked seven days after start of treatment, post-treatment 12 (100 days afterConfidential: start of treatment) and at one For year follow-up Review (12 months Onlyafter start of treatment) 13 about which study medicine (Amoxicillin / placebo / unsure) they think they received during the 14 15 intervention period of the study. 16 17 3.4.1 Primary and secondary outcome measure 18 The primary outcome measure of the study will be pain and disability measured by the Norwegian 19 version of the disease-specific Roland and Morris Disability Questionnaire (RMDQ)41;78. The 20 RMDQ ranges from 0 to 24, with lower score indicating less severe pain and disability. A clinically 21 22 important change will be defined as a 30 % reduction of the individual’s baseline score. Patients 23 will complete the RMDQ at baseline, post-treatment (100 days after start of treatment), and at one 24 year (12 months) after start of treatment. 25 26 3.4.2 Exploratory outcome measures 27 Exploratory clinical outcome measures will as a basic principle be measured at baseline, post- 28 treatment (100 days after start of treatment), and at one year (12 months) after start of treatment. 29 30 See trial flow chart of the end of the protocol for details and for deviations from the basic 31 principles. Secondary clinical outcome measures are low back pain (mean of three Numeric Rating 32 Scales (NRSs, range 0-10); current LBP, the worst LBP within the last 2 weeks, and usual/mean 33 LBP within the last 2 weeks (weekly reports in ViedocMe during the intervention period (for 34 weekly reports during the intervention period; the wording "last 2 weeks" will be replaced by "the 35 last week")) and leg pain (NRS, range 0-10), hours with LBP during the last 4 weeks (number of 36 days during the last 28 days (4 weeks) the participant had experienced LBP (0-28 days), and, on an 37 38 typical day, how many of the hours awake they experienced LBP (0-16 h). The number of days and 32 39 hours are multiplied (a 0-448 scale), bothersomeness (NRS, range 0-10), health-related quality of 1 40 life (EuroQoL-5D, version 2.0) , days with sick leave, functional capacity (Oswestry Disability 41 Index (ODI) 2.0)37, patient’s satisfaction with treatment (5-point Likert scale, not measured at 42 baseline, only 100 days after start of treatment) and at one year (12 months) after start of treatment) 43 and global perceived effect (7-point Likert scale not measured at baseline, only 100 days after start 44 45 of treatment) and at one year (12 months) after start of treatment). Compliance to the medicine 46 protocol (how many days last week did you take the study medication (0-7)), side effects / adverse 47 events (CTC-coded), and co-interventions (other pharmacological treatment (ATC-coded) and non- 48 pharmacological treatment will also be registered. Evidence indicates that there are some 49 methodological shortcomings related to the RMDQ43. ODI is another disease-specific disability 50 score frequently used in LBP research18, especially recommended for the evaluation of treatment 51 effects on patients recruited from secondary health care. Hence, a key secondary objective of our 52 53 study (KSO 2) is to evaluate the treatment effect by using the ODI in addition to RMDQ since this 54 will further improve the validity of the data forming the basis for conclusions drawn from of our 55 study. Low back pain will be evaluated weekly from baseline to post-treatment (100 days after start 56 of treatment) since, in the Danish RCT, the effect of antibiotic treatment was reported to take 6-8 57 weeks to manifest12 and we want to watch this phenomenon closely throughout the intervention 58 period since this information is considered to be an important premise for clinical implications 59 following our study (KSO 3). 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 16 of 41 Page 57 of 101 BMJ

1 2 Haematological parameters (for safety reasons), including measures of kidney and liver function, 3 will be assessed monthly during the intervention period, together with a short clinical evaluation 4 (for monitoring patients’ clinical status related to antibiotic intake, not outcome measures). 5 Haematological parameters to be assessed during the study period are specified below under 6 7 “assessment of safety” (section 6.9). 8 9 MRI: protocol, hypotheses, image evaluation, variables 10 We plan to obtain the following MRI images at baseline and at one year (12 months) follow-up: 11 sagittal T1- and T2 weighted images, axial T2 weighted images, sagittal short tau inversion 12 recovery (STIR)Confidential: images, sagittal fat-water separationFor Reviewimages, sagittal diffusion Only weighted images 13 (DWI), and sagittal T1 weighted DCE images (only if the patient can receive gadolinium injection). 14 15 The same MRI protocol and the same type of 1.5 Tesla MRI scanners will be used at all study sites. 16 A follow-up MRI is taken between 12 and 13 months after treatment start (i.e. 12 to 14 months after 17 baseline MRI).Our a priori hypotheses regarding characteristics of MCs on 1.5 Tesla MRI are as 18 follows: 19 I (KSO 4): Positive effect of antibiotic treatment at one year (12 months) follow up is more likely 20 when the baseline MCs: 21 22 a) contain more type I than type II, based on conventional T1- and T2 weighted MRI, STIR 23 sequences (provide fat suppression), and fat-water separation techniques, 24 b) are more intense on STIR and becomes more rapidly intense on DCE-MRI, or 25 c) are larger. 26 II (KSO 5). Positive treatment effect is related to a reversion of the characteristics in I. a-c from 27 baseline MRI to one year (12 months) follow-up MRI. 28 29 30 As a basis for examining these hypotheses, we will refine the evaluation of MCs by using the above 31 MRI methods, and determine A) the reliability of different MC characteristics by the different MRI 32 methods and B) the relationships of these MC characteristics to each other and to clinical variables. 33 Previous data on A-B exist for conventional T1- and T2 weighted MRI (MCs had high reliability 34 based on kappa values and were related to LBP52) but not for the other planned MRI methods. Thus, 35 our study will provide highly original data on the reliability and validity of new, refined MRI 36 characterizations of MCs. This will advance further LBP research, beyond the present RCT. 37 38 39 The MRIs will be de-identified and independently evaluated by three radiologists for a range of MC 40 characteristics (beyond those initially evaluated on conventional T1- and T2 images to assess RCT 41 eligibility). Variables include: On T1/T2 weighted images, disc herniation (yes/no), MC type (type I 42 vs. I/II-III vs. II-III/I vs. II vs. II/III), MC craniocaudal size (<10%, <25%, 25-50%, >50% of 43 vertebral height), MC volume (<10%, <25%, 25-50%, >50% of vertebra), number of endplates with 44 45 MCs. On STIR, MC signal intensity (0, 1 or 2), size, volume, number of endplates with high signal. 46 On fat-water separation images, MC type and size. On DWI, apparent diffusion coefficients/maps. 47 On DCE-MRI, intensity and size of enhancement, dynamic enhancement. 48 49 Analyses of MRI results will include observer agreement analyses (kappa, McNemars test, Bland 50 Altmann plots) and multiple regression analyses to assess relationships with clinical variables. 51 52 53 Health economy 54 The cost-effectiveness analysis will compare the potential effect of the treatment by using the 79 1 55 primary outcome RMQ and the EQ5D as the measure of effectiveness from baseline to one year 56 (12 months) follow-up. Costs of the study treatment (direct costs) will be estimated using a bottom- 57 up approach. Costs to the healthcare system incurred due to LBP (indirect costs) will be recorded by 58 a monthly cost diary, including registrations of number of visits to a general practitioner, physical 59 or manual therapist, medical specialist, social worker, and alternative therapist, number of days of 60 hospitalization and/or rehabilitation, use of medication (both on prescription and over the counter

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 17 of 41 BMJ Page 58 of 101

1 2 medication), and number of days of sick leave. The costs of work absenteeism will be estimated by 3 the number of days absent from work multiplied by the average wage rate. 4 5 Genetic variation, longitudinal gene and protein expression 6 7 We will investigate longitudinal changes in gene (RNA) and protein expression from before to after 8 treatment (100 days after start of treatment), and at one year (12 months) follow-up. Total RNA will 9 be isolated from whole blood samples using Tempus Spin RNA Isolation Kits (Applied 10 Biosystems) and changes in gene expression will be assessed using reverse transcriptase 11 quantitative real-time PCR (RT-qPCR). Protein expression in serum will be examined by enzyme- 12 linked immunosorbentConfidential: assay (ELISA). We willFor also examineReview epigenetic Onlypatterns by investigating 13 methylation alterations from before to after antibiotic treatment. In addition, we will also investigate 14 15 genetic variations. Genomic DNA will be isolated from whole blood samples and genotyped by 16 TaqMan methodology. In addition, we will collect faeces samples at baseline, 100 days, and at one 17 year (12 months) follow up. The patients will be given equipment to collect faecal samples at home. 18 The material will be monitored for gut flora and resistance by culture and by sequentiation of 19 bacterial genome including search for resistance genes. The patients will be asked about diet, use of 20 antibiotics, any travels abroad, pets and smoking. Analyses will be performed by the NeuroGenetic 21 22 Lab at the Communication and Research Unit for Musculoskeletal Disorders (FORMI) at Oslo 23 University Hospital. The results of the analysis of stools will not be traced back to the individual 24 patient. Neither will the results be available to any of the investigators or anyone who could contact 25 the patients during the study time, to secure blinding of the main study. 26 27 28 4. OVERALL STUDY DESIGN AND STUDY FLOW 29 30 31 4.1 STUDY DESIGN 32 The study is a phase III-study. 33 34 A randomized placebo-controlled and double-blinded multicenter study will be conducted at four 35 university hospitals (Oslo University Hospital Ullevål, Haukeland University Hospital Bergen, 36 St.Olavs Hospital Trondheim, University Hospital Northern-Norway Tromsø) and two local 37 38 hospitals (Sykehuset Østfold and Drammen Sykehus) in Norway. A schematic diagram of trial 39 design is given at the end of the protocol. 40 41 Randomization lists are created using STATA 13 (StatCorp LP, College Station, TX, USA) and 42 stratified by MODIC type (1/2) and previous disc herniation surgery (yes/no) with a 1:1:1:1 43 allocation and random block sizes of 4 and 6. This will ensure similar numbers of patients receiving 44 45 antibiotics or placebo within each stratum (1-MCs type I, no previous back surgery for disc 46 herniation, 2-MCs type I, previous back surgery for disc herniation, 3-MCs type II, no previous 47 back surgery for disc herniation, and 4-MCs type II, previous back surgery for disc herniation). We 48 intend to recruit two equally large patient groups, one with type I MCs and one with type II MCs, in 49 order to evaluate treatment effect separately in each MC type group and not only in the total sample. 50 We will achieve this by stratified randomization as explained above, and we will stop the inclusion 51 of the two MCs type II strata (with and without previous surgery) when we have enough MCs type 52 53 II patients, and continue the inclusion of the two MCs type I strata (or vice versa). Patients are 54 allocated to the MCs type I group if their baseline MRI shows MCs of craniocaudal size >= 10% 55 and of primary or secondary type I at a level (superior or inferior endplate) with disc herniation last 56 2 years. Patients are allocated to the MCs type II group if their baseline MRI shows MCs of 57 craniocaudal size >= 10% and of primary or secondary type II – but not primary or secondary type I 58 – at a level with disc herniation last 2 years. Patients are stratified for previous back surgery for disc 59 herniation (Yes/No) since it is not clear how the low virulent anaerobic organisms gain access to the 60 disc (during normal bacteraemia or as a result of intraoperative contamination). Hence, stratifying

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 18 of 41 Page 59 of 101 BMJ

1 2 for previous back surgery for disc herniation will ensure balanced distribution of this potential 3 source of infection between groups. A statistician not involved in the trial will generate 4 randomization lists. Allocation will be concealed for all people involved in the trial. Subjects, 5 investigators, monitors and data analysts will all be unaware of the treatment assignment. 6 7 8 The allocation list will be stored at Kragerø Tablettproduksjon AS, Kragerø, Norway, which will be 9 responsible for manufacturing the study medicine and numbering the study medicine according to 10 the randomization / allocation list. This will minimize the possibility that those enrolling and 11 assigning patients will obtain access to the list. For safety reasons, a copy of the allocation list will 12 be stored byConfidential: Department of Clinical Research For Support, Review Oslo University Hospital,Only who will be 13 responsible for unblinding during business hours. . 14 15 Each subject will be identified in the study by a unique Subject Identification Code assigned when 16 subject signs the Informed Consent Form. The Subject Identification Code will be the patient’s 17 identity throughout the study and will be written on all CRF’s. 18 19 4.2 STUDY FLOW 20 The recruiting clinician screens eligible patients for inclusion- and exclusion criteria and requests a 21 22 baseline MRI (according to the MRI-protocol defined by the study) to confirm and characterize 23 MCs if: 24 1) the patient is eligible for the RCT based on ALL clinical inclusion- and exclusion criteria AND a 25 lumbar disc herniation on MRI within the preceding 2 years 26 2) AND the earlier (within the preceding 2 years) OR a more recent MRI shows MCs at the 27 herniated level. 28 29 30 Written informed consent will be signed by the patient and the recruiting clinician (copy to the 31 patient) once all clinical inclusion- and exclusion criteria are confirmed and an earlier MRI shows 32 MCs at a herniated level, but before a new confirmative study-MRI is requested. This is because we 33 want to use all MRIs taken in line with the study-MRI-protocol for methodological purposes (ref 34 section 3.3.2, MRI: protocol, hypotheses, image evaluation, variables). This is also specified in the 35 written informed consent. 36 37 38 The baseline MRI is taken at a dedicated center/hospital at the local study site, transferred to 39 Haukeland University Hospital, and independently evaluated for MCs by two study radiologists 40 (and a third radiologist in cases of disagreement). The radiologists inform the recruiting clinician 41 whether the baseline MRI confirms eligibility for the RCT. The recruiting clinician will then call 42 inn the patient for the collection of baseline measures, randomization and handing out study 43 medication. Treatment starts within 4 weeks after baseline MRI is taken. 44 45 46 Follow-up is specified under section 3.3 and in the study flow chart (table 1). 47 48 49 50 5. STUDY POPULATION 51 52 5.1 SELECTION OF STUDY POPULATION 53 Patients from all health regions in Norway referred to the participating hospitals will be screened 54 55 for eligibility. Both conservatively and surgically treated patients (i.e. operated on for disc 56 herniation > 12 months prior to inclusion) will be included. In addition, patients registered in the 57 Norwegian Registry for Spine Surgery operated on for disc herniation and reporting severe LBP 58 pain at one-year follow-up in the registry, will be invited. 59 60 5.2 NUMBER OF PATIENTS

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 19 of 41 BMJ Page 60 of 101

1 2 A total of 160 patients will be included and randomized. See power calculations for further details 3 (section 9). 4 5 5.3 INCLUSION CRITERIA 6 7  Age between 18 and 65 years 8  LBP of > 6 months duration in the area below the 12th rib and above the gluteal folds with a 9 Numerical Rating Scale (NRS) pain intensity score of  5 (mean of three NRS scales; current 10 LBP, the worst LBP within the last 2 weeks, and usual/mean LBP within the last 2 weeks). 11  MRI-confirmed lumbar disc herniation within the preceding 2 years. 12 Confidential: For Review Only 13  MC type I and/or type II in the vertebral body marrow at the same level as the previously 14 herniated disc. For patients with former surgery for disc herniation, the MC has to be located at 15 an operated level. 16  Written informed consent 17 18 19 5.4 EXCLUSION CRITERIA 20  Allergy to penicillin or cefalosporins 21  Allergy/hypersensitivity to any of the excipients of the study drug 22  Current pregnancy or lactation 23  Elevated kidney (creatinine) or hepatic (ALAT/ASAT) values outside normal range 24 25  Phenylketonuria (Følling disease) 26  Mononucleosis or leukaemia 27  Any specific diagnosis that may explain patient’s low back symptoms (e.g. tumor, fracture, 28 spondyloarthritis, infection, spinal stenosis). 29  Former low back surgery (L1 – S1) for other reasons than disc herniation (e.g fusion, 30 31 decompression, disc prosthesis). 32  Former surgery for disc herniation, but < 12 months have elapsed since surgery. 33  Former surgery for disc herniation, but MC located at non-operated level(s) only. 34  Reservation against intake of gelatine (the capsules contains gelatine, which among other things 35 is produced by ingredients from pigs) 36 37  Regular use of glucocorticoids 38  Regular use of opioids with the exception of codeine and tramadol 39  Not understanding Norwegian language 40  Unlikely to adhere to treatment and/ or complete follow-up (e.g ongoing serious psychiatric 41 disease, drug abuse, plans to move) 42 43  Antibiotic treatment within the preceding one month before treatment start 44  Contraindications to MRI (e.g. cardiac pacemaker electrodes, metal implant in eye or brain, 45 claustrophobia). 46  Unwilling to participate 47 48 49 Pregnancy is an exclusion criterion and a pregnancy test will be included at the screening to confirm 50 lack of pregnancy at enrolment. Furthermore, we will ensure that adequate contraception is used 51 during the study in women of childbearing potential since amoxicillin may interact with oral 52 hormonal contraceptives. All included women using oral hormonal contraceptives will therefore be 53 advised to use additional non-hormonal contraceptives like condom, femidom or pessary during the 54 100 day intervention period. Creatinine and ALAT/ASAT will also be included at the screening to 55 confirm normal kidney and hepatic values. At screening, blood for white cell counts, evaluation of 56 57 CRP and Glucose will also be taken. 58 59 60 6 TREATMENT

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 20 of 41 Page 61 of 101 BMJ

1 2 The study treatment will be dispensed to the subject by authorized site personnel only (i.e. medical 3 doctors). No dose modifications are allowed. 4 5 6 6.1 ANTIBIOTIC TREATMENT 7 The rationale for the trial treatment, the dosage and duration of therapy is outlined earlier in the 8 protocol. The intervention will be Amoxicillin (Amoksicillintrihydrat) (Amimox «Meda») tablets 9 750 mg 1x3 for 100 days (oral intake). The tablets will be encapsulated in Capsugel DB-caps AAel 10 Swedish orange. 11 12 Confidential: For Review Only 13 6.2 PLACEBO TREATMENT 14 Patients in the placebo control group will receive placebo capsules for 100 days of daily (1x3), oral 15 intake. The placebo will consist of maize starch and will also be encapsulated in Capsugel DB-caps 16 AAel Swedish orange. 17 18 The medication assigned to each randomized patient (both the Amoxicillin and the placebo) will be 19 encapsulated, packaged and labelled by Kragerø Tablettproduksjon AS, in Kragerø, Norway. The 20 21 containers will be Nolato box/lid 110ml/42mm containing 100 capsules. Labelling, prepared by 22 Kragerø Tablettproduksjon AS, is enclosed at the end of the document. The appearance of the 23 containers, labelling and the capsules will be identical for both treatment groups. 24 25 6.3 DRUG ADMINISTRATIONS AND ACCOUNTABILITY 26 The study medication will be stored and administered trough ”Sykehusapotekene” at the 27 participating study centers. The responsible pharmacist will confirm receipt of study drug and will 28 29 use the study drug only within the framework of this clinical study and in accordance with this 30 protocol. Three containers, each filled with 100 capsules, will be handed out at the start of 31 intervention. Confer the labelling; one capsule should be taken in the morning, one in the middle of 32 the day, and one in the evening. Capsules should be swallowed down in whole and can be taken 33 with or without food. Patients will be informed to keep the medicine in its original container, in a 34 safe place out of reach of children, and in a cool dry place below 25°C. 35 36 37 Return, capsule countand destruction of the returned study drug will also be administered trough 38 ”Sykehusapotekene” at the participating study centers. Monitor will register accountability in 39 electronic systems (Viedoc). 40 41 6.4 DURATION OF THERAPY AND DISCONTINUATION CRITERIA 42 43 As outlined earlier in the protocol, duration of therapy will be 100 days. Protocol therapy will be 44 followed, unless unacceptable side effects (e.g. abnormal/deviating haematological parameters 45 (leucocytes, thrombocytes, eosinophils, haemoglobin (Hb), hematocrit (Ht), kidney (creatinine) or 46 liver function (ASAT / ALAT)) are encountered. Patient subjects may also discontinue protocol 47 therapy in the following instances: 1) Voluntary discontinuation by the patient. 2) Incorrect 48 enrolment. 3) Pregnancy. 4) Disease progression (i.e. acute radiculopathy with corresponding disc 49 herniation seen on an MRI). 5) Deterioration in the patients condition which in the opinion of the 50 51 Principal Investigator warrants study medication discontinuation. The reason for discontinuation 52 will be recorded. Any significant adverse events will be followed up. 53 54 6.5 CONCOMITANT MEDICATION AND CO-INTERVENTIONS 55 Cessation of other treatment or medication will not be required, but patients treated by NSAIDs 56 57 (like Ibux or Naproxen) will be encouraged to limit the intake of such medication since NSAIDs 58 may be a significant co-intervention. The principal investigator will individually and by utmost care 59 evaluate patients treated with drugs containing allopurinol, digoxin, anticoagulants or methotrexate 60 to avoid interactions during the treatment phase of the study. Such patients may be treated by alternative medication during the intervention period, and / or monitored frequently with clinical

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 21 of 41 BMJ Page 62 of 101

1 2 evaluations and blood samples. All concomitant medication (incl. vitamins, herbal preparation and 3 other “over-the-counter” drugs) used by the patient and other treatments / co-interventions (incl. 4 physiotherapy, chiropractic, acupuncture, GP visits, examinations by hospital specialists, back 5 surgery etc) received during the treatment phase of the study and during the follow up period will 6 7 be recorded in the patient’s file and CRF. 8 9 6.6 SUBJECT COMPLIANCE 10 Treatment compliance will be assessed by capsule counts by a pharmacist at last visit (100 days 11 after start of treatment), in addition, patients will be asked every week during the intervention 12 period how Confidential:many days the last week they took For the study Review medication (0-7). Only No other measures (e.g. 13 of blood and urine) will be conducted to assess dose compliance. We will not pre-define a fixed 14 15 limit for compliance since this was not defined in the Danish study. Further, the Danish study 12 16 reported that 94-95% of the patients consumed 95-100% of the tablets . In an explorative analysis 17 (made in addition to the primary intention to treat analysis), we may however exclude non- 18 compliers and re-assess the effect of the treatment on RMDQ from baseline to one year (12 months) 19 follow-up. 20 21 22 6.7 EMERGENCY UNBLINDING 23 The code will be broken only if it’s needed to provide the best health care for patients (eg if drug 24 interactions occur). In case of emergency unblinding, medical doctors at the study sites can contact 25 Medical monitor John-Anker Zwart; [email protected] / cell phone 977 56 394 (24h/7d), 26 who will judge whether the SAE is a SUSAR and contact Department of Clinical Research Support, 27 Oslo University Hospital (Martha Colban) for breaking the code (24h/7d). Hence, the investigator 28 will have unrestricted and immediate access to break the treatment code during 24 hours. 29 30 31 6.8 WITHDRAWAL OF SUBJECTS 32 Patients who withdraw or are withdrawn from the study according to reasons listed in section 6.4 33 will stop further treatment. If the reason for withdrawal is voluntary discontinuation by the patient, 34 no further information or material will be collected about the patient. Data that have already been 35 collected will not be deleted. All included patients will be reported in the trial and no subjects will 36 be replaced. Principal investigators / study sites that have included patients withdrawing from the 37 38 trial will establish a new clinical status of the patient and direct treatment according to normal 39 clinical practice, alternatively transfer the patients back to referring hospital / physician. 40 41 Patients withdrawing or are withdrawn from the study during the intervention period will stop 42 further treatment but will be scheduled for an “end of study visit” with clinical evaluation, blood 43 samples, questionnaires and monitoring of adverse events and concomitant medications, according 44 45 to table 1 in the end of the protocol. Patients withdrawing or are withdrawn from the study during 46 the follow-up period (e.g. the period from post-treatment to one year (12 months) follow-up) will 47 also be scheduled for an “end of study visit”. All withdrawals will be scheduled for the one year (12 48 months) follow-up visit, unless the patient actively resists to be called up. 49 50 Patients not attending follow-up visits will receive a new call-up at next scheduled visit and will not 51 be registered as dropouts until after the last scheduled visit (1 year). The investigator will make a 52 53 reasonable effort to ascertain the reason(s) for withdrawal, while fully respecting the subject's rights 54 not to give his/her reason(s) for withdrawing prematurely from the trial. 55 56 6.9 ASSESSMENT OF SAFETY 57 Most adverse reactions of Amoxicillin are mild and reversible. However, the long-lasting 58 intervention period (100 days) requires extra attention to possible adverse reactions. 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 22 of 41 Page 63 of 101 BMJ

1 2 Expected / known adverse reactions by intake of Amoxicillin (listed in the SmPC downloaded from 3 the homepage of The Norwegian Medicines Agency: http://slv.no/_layouts/Preparatomtaler/Spc/03- 4 2054.pdf) are: 5 6 7 Adverse Reactions (AR); “all untoward and unintended responses to an investigational medicinal 8 product related to any dose administered”. 9 Common (in ≥1/100 to <1/10 cases): feeling sick (nausea), vomiting, diarrhoea, indigestion, 10 inappetence, flatulence, heartburn, enanthema, soreness of the mouth or tongue, xerostomia, 11 changed sense of taste, skin reactions like exanthema, pruritus or urticaria. 12 Less commonConfidential: (in ≥ 1/1000 to1/100 cases): oral-For or vaginal Review thrush, headache, Only tiredness, moderate 13 and temporary rise in liver enzymes. 14 15 Rare (in ≥ 1/10 000 to <1/1000 cases): discoloration of the teeth (especially with the suspension), 16 dizziness. 17 18 Serious Adverse Event (SAE) / Serious Adverse Reaction (SAR); “Any untoward medical 19 occurrence that at any dose results in; Death, Is immediately life-threatening, Requires in-patient 20 hospitalization or prolongation of existing hospitalization, Results in persistent or significant 21 22 disability or incapacity, Is a congenital abnormality or birth defect, or Is an important medical event 23 that may jeopardize the subject or may require medical intervention to prevent one of the outcomes 24 listed above. A SAR is an SAE possibly related to the study medication. 25 Rare (in ≥ 1/10 000 to <1/1000 cases): eosinophilia and hemolytic anemia, laryngeal oedema, 26 serum sickness, allergic vasculitis and anaphylactic shock, hyperkinesis, cramps, acute interstitial 27 nephritis, crystalluria, fever, hepatitis and jaundice, angioneurotic oedema, Stevens-Johnson 28 syndrome, acute generalized exanthematous pustulosis, toxic epidermal necrolysis, bullous 29 30 exfoliative dermatitis. 31 Very rare (in < 1/10 000 cases): leukopenia, granulocytopenia, thrombocytopenia, pancytopenia, 32 myelosuppression, agranulocytosis, prolonged bleeding time and prolonged prothrombin time. 33 Following cessation of medication, these side effects are reversible. 34 35 Safety will be monitored by the assessments described below (haematological parameters etc), as 36 well as the collection of ARs and AEs (“Any untoward medical occurrence in a patient 37 38 administered a pharmaceutical product and which does not necessarily have a causal relationship 39 with this treatment”) at every visit. Patients will be instructed to contact the investigator 40 immediately should they manifest any signs or symptoms they perceive as serious. 41 42 The reporting procedures for SAEs and SUSARs (Suspected Unexpected Serious Adverse 43 Reaction) will follow the national (FOR-2009-10-30-1321) and European regulations (CT-3), which 44 45 means that the investigator will immediately report to the sponsor all serious adverse events (SAEs) 46 within 24 hours of its knowledge. Reports will be sent by Viedoc directly to the medical monitors e- 47 mail ([email protected]). The sponsor will further ensure that all suspected unexpected 48 serious adverse reactions (SUSARs) which are fatal or life-threatening are reported to the regulatory 49 authorities within 7 days after its knowledge. SUSARs not fatal or life threatening shall be reported 50 to the regulatory authorities within a maximum of 15 days of first knowledge. The coordinating 51 investigator will ensure that all the investigators and study staff will be familiar with these 52 53 procedures and timelines. All AE / SAE / SUSARs will be registered in the patients record / journal. 54 55 Haematological parameters (leucocytes, thrombocytes, eosinophils, haemoglobin (Hb) and 56 hematocrit (Ht)) and measures of kidney (creatinine) and liver function (ASAT / ALAT), will be 57 assessed and registered monthly during the intervention period, together with a short clinical 58 evaluation (blood pressure, pulse, auscultation of hearth and lunges) to monitor side effects. White 59 cell counts, evaluation of CRP and Glucose will also be undertaken monthly during the intervention 60 period and after end of treatment (day 100) and at 1-year follow up for further safety monitoring

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 23 of 41 BMJ Page 64 of 101

1 2 and evaluation of inflammatory mechanisms. A qualified physician, who is an investigator or a sub- 3 investigator for the trial, will be responsible for all trial-related medical decisions and will ensure 4 that adequate medical care is provided to a subject for any adverse events, including clinically 5 significant laboratory values, related to the trial. All required treatment will be registered in CRFs. 6 7 All noxious and unintended responses to the study will be considered adverse drug reactions. The 8 physician will further inform a subject when medical care is needed for intercurrent illness(es) of 9 which the investigator becomes aware. The investigator will inform the subject's primary physician 10 about the subject's participation in the trial, which will be specified in the written informed consent. 11 Subject’s participation will also be registered in the patients record / journal. 12 Confidential: For Review Only 13 The risk of adverse events should always be weighted against the expected clinical outcome of the 14 15 trial. Preclinical data do not indicate any danger for humans by intake of Amoxicillin and only 16 patients with chronic problems and considerable pain report (ref inclusion criteria) will be included. 17 Further, the patient group is a huge burden to the society. Summed up, we assume that our 18 intervention and safety procedures are reasonable. 19 20 7.0 ADDITIONAL ETHICAL ASPECTS 21 22 The protocol is approved by the Regional Committees for Medical Research Ethics in south east 23 Norway (REK Sør-Øst), and to the Norwegian Medicines Agency (SLV). The project was 24 registered at ClinicalTrials.gov by December 2014 under the identifier: NCT02323412, and will be 25 monitored by Department of Clinical Research Support, Oslo University Hospital. Written informed 26 consent will be obtained from all patients, with a copy to the included patients. The project will be 27 in accordance with the Helsinki Declaration and the project will adhere to the CONSORT 28 guidelines for transparent reporting of trials. The trial will be conducted in compliance with the 29 30 protocol and the ICH-GCP (Good Clinical Practice) guidelines. Only data described in the protocol 31 will be recorded and considered to be source of data. Data will be stored for 15 years after 32 conclusion of the project period for audit/inspections from e.g. the Norwegian Medicines Agency. 33 Insurance contracts for all study sites are obtained from “Legemiddelansvarsforeningen”. A Data 34 Monitoring Committee (DMC) including one medical doctor (ML), one ethician (LT) and one 35 statistician (AHP), al individuals not involved in the trial, is established. DMC will have closed 36 meetings twice every year, whereof one meeting will be scheduled to about 30 days before the 37 38 annually report from the study to the Norwegian Medicines Agency (SLV). The DMC will evaluate 39 safety data and blinded analysis of treatment effect and hence ensure patients safety. On the other 40 hand, the DMC will also ensure that the trial continues for an adequate period of time and is no 41 stopped too early to answer its scientific questions. 42 The whole trial may be discontinued at the discretion of the PI, the sponsor or the DMC in the event 43 of any of the following: 44 45  Occurrence of AEs unknown to date in respect of their nature, severity and duration 46  Medical or ethical reasons affecting the continued performance of the trial (i.e. inordinate 47 prevalence of AEs, SAEs, and SUSARs and the Norwegian Medicines Agency and/or the 48 Regional Committees for Medical Research Ethics in south east Norway request the study to be 49 terminated) 50 51  Blinded analysis of main outcome measure (RMDQ) performed by the statistician in DMC 52 shows that the improvement in one treatment arm is undeniably superior to in the other 53 treatment arm. Undeniable superior effect in one treatment arm over the other is defined as 54 mean difference of > 7.0 points on the RMDQ between treatment groups, adjusted for baseline 55 values, at 1-year follow-up in the first 80 included patients. 56  Difficulties in the recruitment of patients 57 58 59 In this study we will collect blood samples to investigate associations between genetic variations 60 and clinical outcome after treatment. The information obtained in this study will only to be used for research purposes and not for diagnosis for the participants.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 24 of 41 Page 65 of 101 BMJ

1 2 3 4 8.0 HYPOTHESIS AND ANALYSIS FOR MAIN AND KEY SECONDARY OBJECTIVES 5 The main hypothesis (A) corresponding to the main objective of the trial is: 6 7 8 A. Patients with MCs type I or II at baseline in the antibiotic treatment group reports significantly 9 lower RMDQ-score at one year (12 months) follow-up than patients in the placebo group. 10 We will use ANCOVA analysis, adjusted for baseline RMDQ-score, and perform an intention to 11 treat (ITT) analysis to examine this hypothesis in the whole cohort of patients with MCs type I or II. 12 The significanceConfidential: level will be 0.05. We will Forreport the Review exact p value with Only 95 % CI. 13 14 15 The two most important additional hypotheses B and C, which correspond to secondary objective 16 one (SO 1), are: 17 18 B. Patients with MCs type I at baseline in the antibiotic treatment group reports significantly lower 19 RMDQ-score at one year (12 months) follow-up than patients in the placebo group. 20 21 22 C. Patients with MCs type II at baseline in the antibiotic treatment group reports significantly 23 lower RMDQ-score at one year (12 months) follow-up than patients in the placebo group. 24 25 We will use ANCOVA analysis, adjusted for baseline RMDQ-score, and perform ITT analyses to 26 examine hypothesis B in the sub-sample of patients with type I MCs and hypothesis C in the 27 subsample of patients with type II MCs. The significance level will still be 0.05. We will report the 28 exact p value with 95 % CI for each of the two sub-samples. We consider a Bonferroni adjustment 29 30 of the significance level far too conservative for the assessment of hypotheses B and C, since the 31 effect of antibiotics in one of the two sub-samples is likely to be very dependent of the effect in the 32 other sub-sample (and in the total sample). Among the reasons for this (stated above) are that a) 33 some MCs type I patients may be classified as MCs type II patients and vice versa if we use a MRI 34 machine with a different magnet strength, and b) type I and type II MCs most likely represent a 35 common process (that can be influenced by a common treatment). Nevertheless, the prior Danish 36 study included only MCs type I patients and not MCs type II patients. We therefore need separate 37 38 additional results for treatment effect in each of these two patient groups, otherwise it would be 39 difficult to compare results for MCs type I patients between the present and the Danish trial or to 40 evaluate our rationale for extending the treatment to MCs type II patients. To have power to 41 examine both of the hypotheses B and C, we will include equally many Modic I and Modic II 42 patients in the trial. 43 44 45 Hypothesis D that corresponds to KSO 2 is as follows: 46 D. Patients with MCs type I or II at baseline in the antibiotic treatment group reports significantly 47 lower ODI-score at one year (12 months) follow-up than patients in the placebo group. 48 We will use ANCOVA analysis, adjusted for baseline RMDQ-score, and perform an intention to 49 treat (ITT) analysis to examine this hypothesis in the whole cohort of patients with MCs type I or II. 50 The significance level will be 0.05. We will report the p value with 95 % CI. 51 52 53 Hypothesis E that corresponds to KSO 3 is as follows: 54 E. Patients with MCs type I or II at baseline in the antibiotic treatment group reports significantly 55 lower LBP intensity NRS-score at one year (12 months) follow-up than patients in the placebo 56 group. 57 We will use ANCOVA analysis, adjusted for baseline values, and perform an intention to treat 58 (ITT) analysis to examine this hypothesis in the whole cohort of patients with MCs type I or II. The 59 significance level will be 0.05. We will report the p value with 95 % CI. 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 25 of 41 BMJ Page 66 of 101

1 2 Hypothesis F that corresponds to KSO4 is as follows: 3 F. In the antibiotic treatment group, high signal from MCs on STIR at baseline MRI predicts lower 4 RMDQ-score at one year (12 months) follow-up. 5 6 7 Hypothesis G that corresponds to KSO5 is as follows: 8 G. Reduced signal from MCs on STIR from baseline to one year (12 months) follow-up MRI is 9 associated with lower RMDQ-score at one year (12 months) follow-up. 10 11 Hypothesis H that corresponds to KSO6 is as follows: 12 H. PatientsConfidential: with MCs type I or II at baseline For in the antibiotic Review treatment groupOnly reports significantly 13 better quality of life (EQ-5D) at one year (12 months) follow-up than patients in the placebo 14 15 group. 16 17 We intend to use multiple regression analyses for hypotheses F and G. The significance level for 18 hypotheses D-H will be 0.05, and nominal p values will be reported. 19 20 As secondary analysis, we will perform responder analyses of RMDQ to supplement the 21 22 interpretability of the main analysis. We will analyze if a higher proportion of patients with MCs 23 type I or II at baseline reports a clinically relevant improvement of their RMDQ-score from baseline 24 to one year (12 months) follow-up in the antibiotic treatment group than in the placebo group. 25 Analysis (chi-square test) will be performed for three different cut-offs (>75%, >50% and >30% 26 improvement of RMDQ score from baseline to one year follow-up, respectively). Intention-to-treat 27 (ITT) principles will be used and the significance level will be 0.05. We will report the exact p 28 value and the number needed to treat (NNT) with 95 % CI. 29 30 31 Handling of missing data: Patients with missing data on on RMDQ at one year (12 months) follow- 32 up will be excluded from the analyses of hypothesis A, B, C, F and G. In these analyses, we could 33 have replaced missing data on the main outcome with data at 100-days follow-up, and assumed no 34 further improvement after 100 days (last value carried forward). However, a short-time effect of the 35 intensive 3 months treatment, measured immediately after ended treatment, is less relevant to the 36 use of such treatment in clinical practice. Even a striking effect after 100 days would not justify 37 38 such extensive treatment if the effect rapidly vanished. Thus, we will try to obtain as complete data 39 as possible on RMDQ at 1 year, and not use the 100 days data in the primary one year (12 months) 40 analysis of this main outcome. 41 42 For the three key secondary outcomes (ODI, LBP intensity and EQ-5D), similar considerations will 43 apply as for RMDQ. This implies that patients with missing data on ODI, LBP intensity or EQ-5D 44 45 respectively at one year (12 months) follow-up, will be excluded when analyzing hypotheses D, E 46 and H. The statistical handling of other missing data may depend on the types of data and on 47 clinical considerations, and cannot be planned in advance for all scenarios in the 48 explorative/supportive analyses. In all cases of missing data, we will decide the approach to the 49 missing data a priori in co-operation with the study statistician, prior to analyzing the data, to 50 ensure that the approach is based on reasoning in advance and that it is not selected post hoc based 51 on the results it provided. 52 53 54 55 9.0 POWER CALCULATIONS 56 57 We wished to design the study with enough power to evaluate the treatment effect separately in the 58 MC type I group and in the MC type II group. 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 26 of 41 Page 67 of 101 BMJ

1 2 In power calculations for each MC type group, we used a two-sided alfa of 0.05 and a power of 3 0.90, we wished to be able to identify a difference in mean RMDQ of 4 between the two treatment 4 groups at 12 months follow-up, and we assumed a standard deviation (SD) of RMDQ of 5. The 5 minimal important difference in mean RMDQ between groups is not clear28;46;74, but a change in 6 RMDQ of 2-3 in individual patients over time is very unlikely to be important57 and may represent 7 22 8 measurement error ; the repeatability of the RMDQ was 4 in a Norwegian study of chronic LBP 41 9 patients . We therefore used a difference of 4 in mean RMDQ between groups in the power 10 calculations. The SD of 5 we used is within the upper range of commonly reported SDs for RMDQ 11 in patients with persistent LBP (SD 3.3-4.731;41;50;51;92, SD 6.1 in one study42. We used the sample 12 size calculatorConfidential: for comparing two means that For is available Review here: Only 13 http://www.openepi.com/SampleSize/SSMean.htm 14 15 16 When we put to-sided alfa 0.05 (i.e. 95 % Confidence Interval), power 90, ratio of sample size 1 17 (antibiotic group / placebo group), difference in RMDQ 4, and SD of RMDQ 5 into the calculator, 18 the resulting sample size is 33 in each treatment group, or 66 in both treatment groups; i.e. 132 in 19 the total sample (both MC type groups). The Danish RCT12 had 11 % dropouts. Adding 20 % for 20 dropouts (26 patients) we calculated that we needed to include 158 patients in the study, rounded up 21 22 to 160 patients: 80 patients with type I MCs and 80 patients with type II MCs. 23 24 In the total sample, 66 patients in each treatment group provide 98 % power at a two-sided alfa of 25 0.01 to detect a difference of 4 in mean RMDQ between the groups (and 93 % power at a two-sided 26 alfa of 0.05 to detect a difference of 3) when assuming a SD of RMDQ of 5. When assuming a SD 27 of RMDQ of 6, 66 patients in each group provide 89.5 % power at a two-sided alfa of 0.01 and 97 28 % power at a two-sided alfa of 0.05 to detect a difference of 4 in mean RMDQ between the groups. 29 30 31 32 All data will be analyzed according to ITT principles. No formal interim analysis is planned and all 33 data will be stored in a blinded fashion in a clinical trial database and not be accessed before the 34 official locking of the database (one year after the last patient has been enrolled). No access to the 35 data is allowed before the time point of database lock. However, the statistician in the DMC will 36 perform blinded analysis for groups A and B (NOT Amoxicillin and Placebo) of main outcome 37 38 measure (RMDQ) when the first 80 included patients (i.e. 50% of the planned sample size) have 39 passed 1-year follow-up, to reveal if patients in one treatment arm responds undeniable over the 40 other. Undeniable effect in one treatment arm over the other is defined as mean difference of > 7.0 41 points on the RMDQ between group A and B, adjusted for baseline values, at 1-year follow-up in 42 the first 80 included patients. No details from these analyses will be communicated to the sponsor, 43 but based on the results (i.e. > 7.0 points superior improvement on RMDQ in one group compared 44 45 to the other), DMC may recommend termination of the study (ref: the trial discontinuation criteria 46 listed in section 7). At the time point of database lock we will have a finalized version of the 47 statistical analysis plan (SAP) ready. 48 49 Regarding the stool (faeces) investigations we will perform samples from patients recruited at all 50 study centres. The number of patients with a stool sample taken will be limited to 50-70 patients. As 51 it is difficult to compare bowel flora between individuals due to large interpersonal differences we 52 53 will perform analysis mainly by comparing samples of individual participants over time. Hence, a 54 limited study sample is adequate. Previous studies regarding the effect of amoxicillin on the natural 2;6;21;33;38;82 55 flora has all been limited to 10-40 participants 56 57 10.0 DATA QUALITY 58 Only data collected by CRFs and data generated from outcome measures listed in section 3.3 will 59 be viewed as valid data. Data will be stored at the research database at Oslo University Hospital. 60 The trial will be conducted in compliance with the protocol and the ICH-GCP (Good Clinical

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 27 of 41 BMJ Page 68 of 101

1 2 Practice) guidelines and under guidance of the department of Clinical Research Support, Oslo 3 University Hospital. 4 5 The investigator will permit trial-related monitoring, audits, IRB/IEC review, and regulatory 6 7 inspections and provide direct access to source data/documents. Data will be stored for 25 years. 8 9 10 11.0 PROGRESSION PLAN 11 The study will start when Regional Ethics Committee and the Norwegian Medicines Agency have 12 approved theConfidential: protocol. Detailed progression Forplan with Review milestones is specified Only in the application 13 form. The data collection is planned to be finalized in March 2018. 14 15 16 17 11.0 DISSEMINATION AND COMMUNICATION OF RESULTS. 18 Since LBP is of interest for all types of health personnel worldwide, some of the publications will 19 be published in general journals and some in more specific spine, imaging and health economy 20 journals. The results will be published in peer-reviewed international journals. The results will also 21 22 be presented in the national medical journal and as posters or lectures at conferences. Summaries of 23 the results will be presented in the format of popular science and through media. Emphasize will be 24 put on exposing patient users, important stakeholders and key opinion leaders to our results. 25 26 27 13.0 RESEARCH GROUP, ORGANISATION AND NETWORK. 28 The applicant is the project manager (and primary supervisor of the PhD-candidate), and the project 29 30 owner is Oslo University Hospital, Ullevål (FORMI). FORMI is led by prof. John-Anker Zwart and 31 prof. Lars Nordsletten (FOU-leader for Division of Surgery and Clinical Neuroscience, and research 32 leader at Orthopaedic department). Our departments have very active research fellowship 33 embracing (multiple) research groups, more than 10 professors, about 60 PhD-candidates / post- 34 doctoral fellowships and we have very good research infrastructure and multidisciplinary 35 competence from different health professions (orthopaedic surgery, neurology, radiology, physical 36 therapy, physical medicine and rehabilitation, nursing and chiropractic) and from basic science, 37 38 epidemiology, to clinical research. This will contribute to optimal conditions for the project and for 39 the feasibility of the PhD-project. We have extensive research collaboration with both national and 40 international institutions. Collaborators from other Norwegian institutions include Professor dr.med 41 Ansgar Espeland from Haukeland University Hospital (Bergen), who is responsible for the imaging 42 part of the study, professor Jan Sture Skouen from Haukeland University Hospital, professor 43 Øystein P Nygaard from National Center for Spinal Disorders (Trondheim), associated professor 44 45 Audny Anke from University Hospital Nothern-Norway, and clinicians and researchers from 46 Sykehuset Østfold and Drammen Hospital. Our multidisciplinary team brings together investigators 47 with skills and experience in clinical research in back pain. Each of the members of the team 48 currently collaborates with at least one other member and most collaborates with at least two other 49 members. These working partnerships provide evidence that the research team we propose is viable 50 and that the individual members have a demonstrated ability to work in teams on collaborative 51 projects. The present project will further promote national network building within the field of 52 53 Musculoskeletal Research, a field that recently has been promoted as a priority for research in 54 Norway by the National Co-operation Group on Health Research. Our experience in conducting 55 clinical trials supports the probability that the study can proceed in a timely fashion. 56 57 58 59 Reference List 60 14.0 REFERENCE LIST

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 28 of 41 Page 69 of 101 BMJ

1 2 1. EuroQol--a new facility for the measurement of health-related quality of life. The EuroQol Group. Health 3 Policy 1990;16:199-208. 4 5 2. Swedish Study Group. A randomized multicenter trial to compare the influence of cefaclor and amoxycillin on 6 the colonization resistance of the digestive tract in patients with lower respiratory tract infection. Infection 7 1991;19:208-15. 8 9 3. Adams MA. Biomechanics of back pain. Acupunct.Med. 2004;22:178-88. 10 11 4. Adams MA, Freeman BJ, Morrison HP et al. Mechanical initiation of intervertebral disc degeneration. Spine 12 (PhilaConfidential: Pa 1976.) 2000;25:1625-36. For Review Only 13 5. Adams MA, Roughley PJ. What is intervertebral disc degeneration, and what causes it? Spine (Phila Pa 1976.) 14 2006;31:2151-61. 15 16 6. Adamsson I, Edlund C, Seensalu R et al. The normal gastric microflora and Helicobacter pylorr; before, 17 during and after treatment with omeparazole and amoxycillin. Clinical Microbiology and Infection 1998;4:308- 18 15. 19 20 7. Aebi M. Is low back pain after disc herniation with Modic Type 1 changes a low-grade infection? Eur.Spine J. 21 2013;22:689. 22 23 8. Agarwal V, Golish SR, Alamin TF. Bacteriologic culture of excised intervertebral disc from immunocompetent 24 patients undergoing single level primary lumbar microdiscectomy. J.Spinal Disord.Tech. 2011;24:397-400. 25 26 9. Albert HB, Kjaer P, Jensen TS et al. Modic changes, possible causes and relation to low back pain. 27 Med.Hypotheses 2008;70:361-8. 28 29 10. Albert HB, Lambert P, Rollason J et al. Does nuclear tissue infected with bacteria following disc herniations 30 lead to Modic changes in the adjacent vertebrae? Eur.Spine J. 2013;22:690-6. 31 32 11. Albert HB, Manniche C, Sorensen JS et al. Antibiotic treatment in patients with low-back pain associated with 33 Modic changes Type 1 (bone oedema): a pilot study. Br.J.Sports Med. 2008;42:969-73. 34 35 12. Albert HB, Sorensen JS, Christensen BS et al. Antibiotic treatment in patients with chronic low back pain and 36 vertebral bone edema (Modic type 1 changes): a double-blind randomized clinical controlled trial of efficacy. 37 Eur.Spine J. 2013;22:697-707. 38 39 13. Artus M, van der Windt DA, Jordan KP et al. Low back pain symptoms show a similar pattern of improvement 40 following a wide range of primary care treatments: a systematic review of randomized clinical trials. 41 Rheumatology.(Oxford) 2010;49:2346-56. 42 43 14. Bendix T, Sorensen JS, Henriksson GA et al. Lumbar modic changes-a comparison between findings at low- 44 and high-field magnetic resonance imaging. Spine (Phila Pa 1976.) 2012;37:1756-62. 45 46 15. Berg L, Neckelmann G, Gjertsen O et al. Reliability of MRI findings in candidates for lumbar disc prosthesis. Neuroradiology. 2012;54:699-707. 47 48 16. Bierry G, Dietemann JL. Imaging evaluation of inflammation in the musculoskeletal system: current concepts 49 and perspectives. Skeletal Radiol. 2013;42:1347-59. 50 51 17. Blondel B, Tropiano P, Gaudart J et al. Clinical results of lumbar total disc arthroplasty in accordance with 52 Modic signs, with a 2-year-minimum follow-up. Spine (Phila Pa 1976.) 2011;36:2309-15. 53 54 18. Bombardier C. Outcome assessments in the evaluation of treatment of spinal disorders: summary and general 55 recommendations. Spine 2000;25:3100-3. 56 57 19. Boos N, Weissbach S, Rohrbach H et al. Classification of age-related changes in lumbar intervertebral discs: 58 2002 Volvo Award in basic science. Spine (Phila Pa 1976.) 2002;27:2631-44. 59 60 20. Brennan GP, Fritz JM, Hunter SJ et al. Identifying subgroups of patients with acute/subacute "nonspecific" low back pain: results of a randomized clinical trial. Spine 2006;31:623-31.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 29 of 41 BMJ Page 70 of 101

1 2 21. Brismar B, Edlund C, Nord CE. Impact of cefpodoxime proxetil and amoxicillin on the normal oral and 3 intestinal microflora. Eur.J.Clin.Microbiol.Infect.Dis. 1993;12:714-9. 4 5 22. Brouwer S, Kuijer W, Dijkstra PU et al. Reliability and stability of the Roland Morris Disability Questionnaire: 6 intra class correlation and limits of agreement. Disabil.Rehabil. 2004;26:162-5. 7 8 23. Burke JG, Watson RW, McCormack D et al. Intervertebral discs which cause low back pain secrete high levels 9 of proinflammatory mediators. J.Bone Joint Surg.Br. 2002;84:196-201. 10 11 24. Corsia MF, Wack M, Denys G. Low vitulence Bacterial infections of intervertebral discs and the resultant 12 spinalConfidential: disease processes. Abstract from Scoliosis For Research Review Society (SRS) annual Only meeting . 2003. Ref Type: Abstract 13 14 25. Crock HV, Goldwasser M. Anatomic studies of the circulation in the region of the vertebral end-plate in adult 15 Greyhound dogs. Spine (Phila Pa 1976.) 1984;9:702-6. 16 17 26. De La Cochetiere MF, Durand T, Lepage P et al. Resilience of the dominant human fecal microbiota upon 18 short-course antibiotic challenge. J.Clin.Microbiol. 2005;43:5588-92. 19 20 27. de RA, Kressel H, Spritzer C et al. MR imaging of marrow changes adjacent to end plates in degenerative 21 lumbar disk disease. AJR Am.J.Roentgenol. 1987;149:531-4. 22 23 28. Demoulin C, Ostelo R, Knottnerus JA et al. What factors influence the measurement properties of the Roland- 24 Morris disability questionnaire? Eur.J.Pain 2010;14:200-6. 25 26 29. Derogatis LR, Lipman RS, Rickels K et al. The Hopkins Symptom Checklist (HSCL): a self-report symptom 27 inventory. Behav Sci 1974;19:1-15. 28 29 30. Deyo RA. Treatments for back pain: can we get past trivial effects? Ann.Intern.Med 2004;141:957-8. 30 31 31. Dufour N, Thamsborg G, Oefeldt A et al. Treatment of chronic low back pain: a randomized, clinical trial 32 comparing group-based multidisciplinary biopsychosocial rehabilitation and intensive individual therapist- 33 assisted back muscle strengthening exercises. Spine (Phila Pa 1976.) 2010;35:469-76. 34 35 32. Dunn KM, Croft PR. Classification of low back pain in primary care: using "bothersomeness" to identify the 36 most severe cases. Spine (Phila Pa 1976.) 2005;30:1887-92. 37 38 33. Edlund C, Stark C, Nord CE. The relationship between an increase in beta-lactamase activity after oral 39 administration of three new cephalosporins and protection against intestinal ecological disturbances. 40 J.Antimicrob.Chemother. 1994;34:127-38. 41 42 34. Eguchi Y, Ohtori S, Yamashita M et al. Diffusion magnetic resonance imaging to differentiate degenerative 43 from infectious endplate abnormalities in the lumbar spine. Spine (Phila Pa 1976.) 2011;36:E198-E202. 44 45 35. Eriksen HR, Ihlebaek C, Ursin H. A scoring system for subjective health complaints (SHC). Scand.J.Public 46 Health 1999;27:63-72. 47 36. Fairbank J, Gwilym SE, France JC et al. The role of classification of chronic low back pain. Spine (Phila Pa 48 1976.) 2011;36:S19-S42. 49 50 37. Fairbank JC, Pynsent PB. The oswestry disability index. Spine 2000.Nov 15;25(22.):2940.-53. 2000;25:2940- 51 53. 52 53 38. Floor M, van AF, Rozenberg-Arska M et al. Effect of loracarbef and amoxicillin on the oropharyngeal and 54 intestinal microflora of patients with bronchitis. Scand.J.Infect.Dis. 1994;26:191-7. 55 56 39. Fritzell P, Bergstrom T, Welinder-Olsson C. Detection of bacterial DNA in painful degenerated spinal discs in 57 patients without signs of clinical infection. Eur.Spine J. 2004;13:702-6. 58 59 40. Goupille P, Jayson MI, Valat JP et al. Matrix metalloproteinases: the clue to intervertebral disc degeneration? 60 Spine (Phila Pa 1976.) 1998;23:1612-26.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 30 of 41 Page 71 of 101 BMJ

1 2 41. Grotle M, Brox JI, Vollestad NK. Cross-cultural adaptation of the Norwegian versions of the Roland-Morris 3 Disability Questionnaire and the Oswestry Disability Index. J Rehabil Med. 2003;35:241-7. 4 5 42. Grotle M, Foster NE, Dunn KM et al. Are prognostic indicators for poor outcome different for acute and 6 chronic low back pain consulters in primary care? Pain 2010;151:790-7. 7 8 43. Grotle M, Wilkens P, Garratt AM et al. Which Roland-Morris Disability Questionnaire? Rasch analysis of four 9 different versions tested in a Norwegian population. J.Rehabil.Med. 2013;45:670-7. 10 11 44. Hagen KB, Thune O. Work incapacity from low back pain in the general population. Spine. 1998;23:2091-5. 12 Confidential: For Review Only 13 45. Hancock MJ, Maher CG, Laslett M et al. Discussion paper: what happened to the 'bio' in the bio-psycho-social model of low back pain? Eur.Spine J. 2011;20:2105-10. 14 15 46. Henschke N, van EA, Froud R et al. Responder analyses in randomised controlled trials for chronic low back 16 pain: an overview of currently used methods. Eur.Spine J. 2014;23:772-8. 17 18 47. Ihlebaek C, Laerum E. Plager flest - koster mest. Muskel-skjelettlidelser i Norge. Nasjonalt ryggnettverk - 19 Formidlingsenheten. 1, 1-64. 2004. Oslo. 20 Ref Type: Report 21 22 48. Jacobsen LM, Schistad EI, Storesund A et al. The MMP1 rs1799750 2G allele is associated with increased low 23 back pain, sciatica, and disability after lumbar disk herniation. Clin.J.Pain 2013;29:967-71. 24 25 49. Jacobsen LM, Schistad EI, Storesund A et al. The COMT rs4680 Met allele contributes to long-lasting low 26 back pain, sciatica and disability after lumbar disc herniation. Eur.J.Pain 2012;16:1064-9. 27 28 50. Jensen OK, Nielsen CV, Sorensen JS et al. Type 1 Modic changes was a significant risk factor for 1 year 29 outcome in sick-listed low back pain patients: a nested cohort study using magnetic resonance imaging of the 30 lumbar spine. Spine J. 2014. 31 32 51. Jensen RK, Leboeuf-Yde C, Wedderkopp N et al. Rest versus exercise as treatment for patients with low back 33 pain and Modic changes. A randomized controlled clinical trial. BMC.Med. 2012;10:22. 34 35 52. Jensen TS, Karppinen J, Sorensen JS et al. Vertebral endplate signal changes (Modic change): a systematic 36 literature review of prevalence and association with non-specific low back pain. Eur.Spine J. 2008;17:1407-22. 37 38 53. Karppinen J, Daavittila I, Solovieva S et al. Genetic factors are associated with modic changes in endplates of 39 lumbar vertebral bodies. Spine (Phila Pa 1976.) 2008;33:1236-41. 40 41 54. Keller A, Hayden J, Bombardier C et al. Effect sizes of non-surgical treatments of non-specific low-back pain. 42 Eur.Spine J 2007;16:1776-88. 43 44 55. Koes BW, van TM, Lin CW et al. An updated overview of clinical guidelines for the management of non- 45 specific low back pain in primary care. Eur.Spine J 2010;19:2075-94. 46 47 56. Korthals-de B, I, van TM, van DH et al. Economic evaluations and randomized trials in spinal disorders: principles and methods. Spine (Phila Pa 2004;29:442-8. 48 49 57. Kovacs FM, Abraira V, Royuela A et al. Minimal clinically important change for pain intensity and disability in 50 patients with nonspecific low back pain. Spine (Phila Pa 1976.) 2007;32:2915-20. 51 52 58. Kvale A, Ellertsen B, Skouen JS. Relationships between physical findings (GPE-78) and psychological profiles 53 (MMPI-2) in patients with long-lasting musculoskeletal pain. Nord.J Psychiatry 2001;55:177-84. 54 55 59. Laerum E, Brage S, Ihlebaek C et al. Et muskel- og skjelettregnskap. Forekomst og kostnader knyttet til 56 skader, sykdommer og plager i muskel- og skjelettsystemet. Muskel og Skjelett Tiåret (MST). 1/2013, 1-85. 57 2013. 58 Ref Type: Report 59 60 60. Lang P, Chafetz N, Genant HK et al. Lumbar spinal fusion. Assessment of functional stability with magnetic resonance imaging. Spine (Phila Pa 1976.) 1990;15:581-8.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 31 of 41 BMJ Page 72 of 101

1 2 61. Leboeuf-Yde C, Kjaer P, Bendix T et al. Self-reported hard physical work combined with heavy smoking or 3 overweight may result in so-called Modic changes. BMC.Musculoskelet.Disord. 2008;9:5. 4 5 62. Low RN, Austin MJ, Ma J. Fast spin-echo triple echo dixon: Initial clinical experience with a novel pulse 6 sequence for simultaneous fat-suppressed and nonfat-suppressed T2-weighted spine magnetic resonance 7 imaging. J.Magn Reson.Imaging 2011;33:390-400. 8 9 63. Ma J. Dixon techniques for water and fat imaging. J.Magn Reson.Imaging 2008;28:543-58. 10 11 64. Mixter WJ, Barr JS. Rupture of the intervertebral disc with involvement of the spinal canal. N.Engl.J Med 12 1934;210Confidential:-4. For Review Only 13 65. Modic MT. Modic type 1 and type 2 changes. J.Neurosurg.Spine 2007;6:150-1. 14 15 66. Modic MT, Steinberg PM, Ross JS et al. Degenerative disk disease: assessment of changes in vertebral body 16 marrow with MR imaging. Radiology. 1988;166:t-9. 17 18 67. Mulleman D, Mammou S, Griffoul I et al. Pathophysiology of disk-related low back pain and sciatica. II. 19 Evidence supporting treatment with TNF-alpha antagonists. Joint Bone Spine 2006;73:270-7. 20 21 68. Nguyen C, Benichou M, Revel M et al. Association of accelerated switch from vertebral end-plate Modic I to 22 Modic 0 signal changes with clinical benefit of intradiscal corticosteroid injection for chronic low back pain. 23 Arthritis Rheum. 2011;63:2828-31. 24 25 69. NORM. Norsk overvåkingssystem for antibiotikaresistens hos mikrober. 2013. 20-8-2013. 26 Ref Type: Personal Communication 27 28 70. Ohlund C, Eek C, Palmbald S et al. Quantified pain drawing in subacute low back pain. Validation in a 29 nonselected outpatient industrial sample. Spine 1996;21:1021-30. 30 31 71. Ohtori S, Inoue G, Ito T et al. Tumor necrosis factor-immunoreactive cells and PGP 9.5-immunoreactive nerve 32 fibers in vertebral endplates of patients with discogenic low back Pain and Modic Type 1 or Type 2 changes on 33 MRI. Spine (Phila Pa 1976.) 2006;31:1026-31. 34 35 72. Olmarker K. Combination of two cytokine inhibitors reduces nucleus pulposus-induced nerve injury more than 36 using each inhibitor separately. Open.Orthop.J. 2011;5:151-3. 37 38 73. Olsen MB, Jacobsen LM, Schistad EI et al. Pain intensity the first year after lumbar disc herniation is 39 associated with the A118G polymorphism in the opioid receptor mu 1 gene: evidence of a sex and genotype 40 interaction. J.Neurosci. 2012;32:9831-4. 41 42 74. Ostelo RW, Deyo RA, Stratford P et al. Interpreting change scores for pain and functional status in low back 43 pain: towards international consensus regarding minimal important change. Spine (Phila Pa 1976.) 44 2008;33:90-4. 45 46 75. Peng B, Pang X, Wu Y et al. A randomized placebo-controlled trial of intradiscal methylene blue injection for the treatment of chronic discogenic low back pain. Pain 2010;149:124-9. 47 48 76. Pozzi G, Garcia PC, Stradiotti P et al. Diffusion-weighted MR imaging in differentiation between osteoporotic 49 and neoplastic vertebral fractures. Eur.Spine J. 2012;21 Suppl 1:S123-S127. 50 51 77. Rahme R, Moussa R. The modic vertebral endplate and marrow changes: pathologic significance and relation 52 to low back pain and segmental instability of the lumbar spine. AJNR Am.J.Neuroradiol. 2008;29:838-42. 53 54 78. Roland M, Fairbank J. The Roland-Morris Disability Questionnaire and the Oswestry Disability Questionnaire. 55 Spine 2000;25:3115-24. 56 57 79. Roland M, Morris R. A study of the natural history of back pain. Part I: development of a reliable and sensitive 58 measure of disability in low-back pain. Spine 1983;8:141-4. 59 60 80. Schistad EI, Espeland A, Pedersen LM et al. Association between baseline IL-6 and 1-year recovery in lumbar radicular pain. Eur.J.Pain 2014.

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 32 of 41 Page 73 of 101 BMJ

1 2 81. Spitzer WO, LeBlanc FE, Dupuis M. Scientific approach to the assessment and management of activity- 3 related spinal disorders. A monograph for clinicians. Report of the Quebec Task Force on Spinal Disorders. 4 Spine. 1987;12:S1-59. 5 6 82. Stark CA, Adamsson I, Edlund C et al. Effects of omeprazole and amoxycillin on the human oral and 7 gastrointestinal microflora in patients with Helicobacter pylori infection. J.Antimicrob.Chemother. 8 1996;38:927-39. 9 10 83. Stirling A, Worthington T, Rafiq M et al. Association between sciatica and Propionibacterium acnes. Lancet 11 2001;357:2024-5. 12 Confidential: For Review Only 84. Sullivan A, Edlund C, Nord CE. Effect of antimicrobial agents on the ecological balance of human microflora. 13 Lancet Infect.Dis. 2001;1:101-14. 14 15 85. Toyone T, Takahashi K, Kitahara H et al. Vertebral bone-marrow changes in degenerative lumbar disc disease. 16 An MRI study of 74 patients with low back pain. J.Bone Joint Surg.Br. 1994;76:757-64. 17 18 86. Vibe FK, O'sullivan P, Skouen JS et al. Efficacy of classification-based cognitive functional therapy in patients 19 with non-specific chronic low back pain: a randomized controlled trial. Eur.J Pain 2013;17:916-28. 20 21 87. Vital JM, Gille O, Pointillart V et al. Course of Modic 1 six months after lumbar posterior osteosynthesis. 22 Spine (Phila Pa 1976.) 2003;28:715-20. 23 24 88. Vos T, Flaxman AD, Naghavi M et al. Years lived with disability (YLDs) for 1160 sequelae of 289 diseases and 25 injuries 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 26 2012;380:2163-96. 27 28 89. Waddell G. 1987 Volvo award in clinical sciences. A new clinical model for the treatment of low-back pain. 29 Spine. 1987;12:632-44. 30 31 90. Waddell G. The Back Pain Revolution. First ed. London: Churchill Livingstone, 1998. 32 33 91. Waddell G, Newton M, Henderson I et al. A Fear-Avoidance Beliefs Questionnaire (FABQ) and the role of 34 fear- avoidance beliefs in chronic low back pain and disability. Pain 1993;52:157-68. 35 36 92. Wilkens P, Scheel IB, Grundnes O et al. Effect of glucosamine on pain-related disability in patients with 37 chronic low back pain and degenerative lumbar osteoarthritis: a randomized controlled trial. JAMA. 38 2010;304:45-52. 39 40 93. Williams FM, Bansal AT, van Meurs JB et al. Novel genetic variants associated with lumbar disc degeneration 41 in northern Europeans: a meta-analysis of 4600 subjects. Ann.Rheum.Dis. 2013;72:1141-8. 42 43 94. Zhang YH, Zhao CQ, Jiang LS et al. Modic changes: a systematic review of the literature. Eur.Spine J. 44 2008;17:1289-99. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 33 of 41 BMJ Page 74 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 34 of 41 Page 75 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj PROTOCOL VERSION NO. 7.2 – 18.02.2016 Sub-study resistant bacteria/genes faecal flora Page 35 of 41 BMJ Page 76 of 101

1 2 3 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 77 of 101 BMJ

1 2 Table 1. Trial flow chart 3

4 5 Screening Baseline measures End of treatment 6 consultation End of study (when confirmed Treatment Period visit (within 3 days Follow-up Period 7 visit MRI = maximum 4 after last dose) 8 Confidential:weeks after For MRI) Review Only 9 Time Within 6 weeks Day Day 66 Within 7 days 1 year after 10 before treatment Day 0 33 (+/- (+/- 7 Day 100 (+/- 7 days) after treatment start 11 starts 7 days) days) withdrawal (+/- 2 weeks) 12 13 Medical History X 14 15 Informed consent X 16 Inclusion/exclusion 17 X 18 Evaluation 19 1) 20 MRI X X X 21 Blood samples (safety and X X X X X X 22 inflammatory mechanisms)2) 23 3) 24 Clinical evaluation (safety) X X X X X X X 25 26 Faeces samples X X X X 27 28 Blood samples (genetic aims) X X X X X X 29 30 Clinical evaluation (pain/neuro) 4) X 31

32 Adverse events X X X X X 33 34 Concomitant medication5) X X X X X X 35 36 Co-interventions (non-pharm) 5) X X X X X X 37 5) 38 Sick listing X X X X X X 39 40 Questionnaires X X X X 41 42 37 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 78 of 101

1 2 Screening Baseline measures End of treatment 3 consultation End of study (when confirmed Treatment Period visit (within 3 days Follow-up Period 4 visit MRI = maximum 4 after last dose) 5 weeks after MRI) 6 Time Within 6 weeks Day Day 66 Within 7 days 1 year after 7 before treatment Day 0 33 (+/- (+/- 7 Day 100 (+/- 7 days) after treatment start 8 Confidential:starts For7 days)Review days) Onlywithdrawal (+/- 2 weeks) 9 6) 6) 6) 6) 10 Pain monitoring X X X X X X 11 Randomization 12 X 13 Treatment X 14 administration/dispensation 15 Compliance (capsule count)7) 16 X7) 17 18 * If treatment starts at the day of baseline measures, baseline measures and Day 0 is concurrent. 19 1) Baseline MRI according to the study protocol can be maximum 4 weeks old when treatment starts. A follow-up MRI is taken between 12 and 13 months after treatment start (i.e. 12 to 14 months after 20 baseline MRI). 2) Haematological parameters (leucocytes, thrombocytes, eosinophils, haemoglobin (Hb) and hematocrit (Ht)) and measures of kidney (creatinine) and liver function (ASAT / ALAT), every month or 21 more frequently if clinically indicated (SLV-imposed). Glucose, white cell counts and CRP is for further safety monitoring and evaluation of inflammatory mechanisms (self-imposed) 22 3) Blood pressure, pulse, auscultation of hearth and lunges 23 4) Pain provocation tests, neurological tests 24 5) Concomitant medication, co-interventions and sick-listing will be monitored monthly also during the follow-up period (100 days to 1 year) for health-economical calculations. 25 6) Pain-monitoring (LBP intensity) weekly during treatment period 26 7) Containers and capsules delivered to local “Sykehusapotek” at each study senter for return capsule count, registering of accountability in electronic systems and destruction of the returned study drug 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 38 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 79 of 101 BMJ

1 2 3 4 5 6 7 8 Confidential: For Review Only 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 39 43 https://mc.manuscriptcentral.com/bmj 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Page 80 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 81 of 101 BMJ

1 2 Labelling prepared by Kragerø Tablettproduksjon AS. 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

41 https://mc.manuscriptcentral.com/bmj BMJ Page 82 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 83 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 84 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 85 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 86 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 87 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 88 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 89 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 90 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 91 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 92 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 93 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 94 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 95 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 96 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 97 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 98 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 99 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 100 of 101

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj Page 101 of 101 BMJ

1 2 3 4 5 6 7 8 9 10 11 12 Confidential: For Review Only 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj BMJ Page 102 of 101

1 2 3 4 Abstract 5 Study question 6 This study aimed to assess the efficacy of three months of antibiotic treatment compared to placebo in patients 7 with chronic low back pain (LBP), prior disc herniation, and vertebral endplate changes visible on magnetic 8 resonance images (Modic changes (MCs)). 9 10 Methods 11 In this double-blind parallel-group multicentre trial in a hospital outpatient setting, replicating a former study, 12 180 patientsConfidential: with chronic LBP, previous disc herniation, For and Review type I (n=118) or type Only II (n=62) MCs were randomized to three months of oral treatment with either 750 mg amoxicillin or placebo three times daily. The 13 primary outcome was the Roland-Morris Disability Questionnaire (RMDQ) score (range 0-24) at 1-year follow- 14 up in the intention-to-treat population. We predefined the minimal clinically important between-group difference 15 in mean RMDQ score as 4. 16 17 Study answer and limitations 18 The difference in the mean RMDQ score between the amoxicillin and the placebo group was for the total cohort 19 (primary analysis) -1·6 (95% CI, -3·1 to 0·0; P=0·04), for the type I MC group (secondary analysis) -2·3 (95% 20 CI, -4·2 to -0·4; P=0·02), and for the type II MC group (secondary analysis) -0·1 (95% CI, -2·7 to 2·6; P=0·95). 21 A limitation was treatment without prior microbiological evidence of infection. 22 23 What this study adds 24 We did not find any clinically relevant effect of three months oral antibiotic treatment in patients with chronic 25 low back pain, MCs, and a former herniated disc. The difference between the treatment groups (2·3 points on the 26 RMDQ) for patients with type I MCs was substantially smaller than reported in the trial we were reassessing 27 (8·3 points on the RMDQ). 28 29 Funding, competing interests, data sharing 30 The study was funded by governmental organizations (Helse Sør-Øst and Helse Vest). The authors report no 31 competing interests. Requests for data should be addressed to [email protected]. 32 33 Study registration 34 ClinicalTrials.gov NCT02323412 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

https://mc.manuscriptcentral.com/bmj