Title Identifying biological pathways in the MRI findings of people with low back pain Authors and affiliations Rikke K Jensen1, Tue Secher Jensen1, Per Kjaer1,2, Peter Kent1 1Research Department, Spine Centre of Southern Denmark, Hospital Lillebaelt, Institute of Regional Health Services Research, University of Southern Denmark, Part of the Clinical Locomotion Network, Middelfart, Denmark. 2Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark, Part of the Clinical Locomotion Network, Odense, Denmark Contact email firstname.lastname@example.org Background Investigations into the association between lumbar MRI findings and low back pain (LBP) are complicated by multiple MRI findings being present at the same time. Findings such as lumbar intervertebral disc protrusions or endplate changes almost always co-exist with other degenerative disc findings such as the reduction of disc height and signal intensity. Despite this, the majority of previous research has focused solely on the associations between single imaging findings and pain or other clinical outcomes. Only recently have researchers started to engage with this complexity of MRI findings. An initial strategy to advance this area of investigation would be to recognise which MRI findings typically occur together and whether those clusters appear to reflect discrete biological pathways. Therefore, the objectives of this proof-of-concept study were to identify which vertebral MRI findings cluster together and describe plausible biological pathways that these clusters might represent. Methods All participants were people with chronic LBP seeking a comprehensive evaluation at an outpatient spine clinic in a Danish university hospital. Data for this study was extracted from the MRI findings of 631 people (3,155 lumbar spine motion segments) whose images were quantitatively coded using a detailed research protocol as part of the recruitment phase of two clinical trials. Reproducibility is high when using this MRI coding protocol (kappa 0.52-0.97). The MRI variables included in this study were information on intervertebral disc height and signal intensity, disc protrusions, high intensity zones, size and type of vertebral endplate signal changes, vertebral endplate irregularities and defects, osteophytes, and spondylolisthesis. Latent class analysis (probabilistic data mining) was used to distinguish the best fitting clusters of MRI findings. The distribution of lumbar disc levels in each cluster was also reported. Based on known histological changes inherent in the degeneration process of the motion segment, the clusters were grouped into hypothetical biological pathways. Results Latent class analysis identified twelve clusters of MRI findings. One cluster, characterised by no abnormal MRI findings, contained 52% of the motion segments and represented the normal, pre-degenerative motion segments. The following hypothetical pathways were derived from the content of the other clusters: (i) two clusters representing progressive stages of disc degeneration in the lower lumbar levels; (ii) four clusters representing progressive stages of disc protrusions and endplate changes in the lower lumbar levels; (iii) two cluster with endplate changes at either the upper or the lower endplates; (iv) two clusters containing progressive endplate changes and disc degeneration at the upper lumbar levels only; and lastly, (v) one cluster containing osteophytes at the upper lumbar motion segments. Conclusions MRI findings of lumbar vertebral motion segments were grouped into twelve clusters and those clusters fitted into a model of five different biological pathways of degeneration. Future research will test the association between these clusters and clinically important characteristics such as pain and activity limitation.