Abstract
BACKGROUND: There is no established consensus on how to quantify and combine MRI findings to most accurately assess their cumulative relationship with different low back pain (LBP) phenotypes. OBJECTIVES: To explore associations between single MRI findings, sum scores and combinations of findings and LBP with and without leg pain. METHODS: Cross-sectional analyses from a population-based cohort of 344 people aged 45 years with lumbar MRI and self-reported data. LBP phenotypes were defined as ‘no-LBP’ (n = 198), ‘non-radicular LBP’ (n = 90), and ‘radicular LBP’ (n = 56). Each MRI finding was analysed as present at one or more segmental levels and as sum scores across levels. Two combinations of MRI findings were defined based on their hypothesised relevance to non-radicular or LBP with leg pain. Multinomial logistic regression was used to explore patterns in estimates of association. RESULTS: Disc degeneration, Modic type II, facet joint degeneration, and abnormal disc contour showed increasing estimates of association with non-radicular LBP when analysed as sum scores and present at more segmental levels. For the non-radicular combination of MRI-findings, participants with ≥ 4 findings had higher estimates of association with non-radicular LBP compared to those with no or fewer findings. For the radicular combination of MRI-findings, participants with ≥ 4 findings had increased estimates of association compared to those with no or fewer findings for both LBP phenotypes. CONCLUSION: Although no statistically significant associations were observed, four of seven MRI findings showed increasing estimates of associations with non-radicular LBP when assessed as sum scores. Participants with ≥ 4 MRI findings had increased estimates of association with LBP phenotypes compared to those with no or fewer findings. These observations require confirmation in larger studies. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12891-026-09614-2.