INTRODUCTION: Dynamic assessment of lumbar spine intervertebral kinematics may help identify subgroups of patients with chronic low back pain (cLBP) who may preferentially benefit from specific interventions¹. One kinematics biomarker that may help classify individuals with cLBP is motion sharing among intervertebral motion segments², i.e. how the overall movement of the lumbar spine is distributed among the individual motion segments. A recent systematic review concluded that little or no data is available describing dynamic motion sharing in individuals with cLBP during flexion/extension or lateral bending³. This interim analysis from an ongoing study evaluates the potential for classifying individuals with cLBP based upon dynamic motion sharing among lumbar motion segments.

METHODS: Individuals having cLBP, defined as low back pain for more than 3 months with pain persistence greater than 50% of the time in the last six months, BMI <35 kg/m², not pregnant, and able to perform the required lumbar motions were enrolled. After obtaining written informed consent, participants performed two to three trials of maximal lateral bending to both sides and two to three trials of flexion/extension (Figure 1). Synchronized biplane radiographs of the lumbar spine were captured at 20 images per second over the 3 second duration of each movement trial. Bone tissues of L1 through S1 were segmented from CT scans and used to create subject-specific 3D bone models. Vertebral motion was tracked in the biplane radiographs using a validated tracking process that matched digitally reconstructed radiographs created from the CT-based bones to the biplane radiographs with an accuracy of better than 1mm in translation and 1° in rotation⁴. The percent contribution of each motion segment to the change in L1-S1 angle was computed at each 1° increment of L1-S1 rotation and averaged over the two to three corresponding movement trials for each participant. Segmental contributions to L1-S1 rotation were clustered using MATLAB’s built-in k-means clustering algorithm. Silhouette coefficients were calculated to determine the appropriate number of clusters, with silhouette scores closer to 1.0 indicating more accurate clustering.

RESULTS: Data processing has been completed for 20 out of the 85 participants who have completed motion testing (9 M, 11 F; average age: 46±17yrs.; BMI: 25.5±4.2kg/m²). Segmental contributions to L1-S1 rotation were used to group participants into 3 to 4 clusters at each motion segment for flexion (Figure 2), 2 to 5 clusters at each motion segment for extension, and 2 to 3 clusters at each motion segment for lateral bending, with silhouette scores ranging from 0.49 to 0.55 for flexion, 0.48 to 0.54 for extension, and from 0.59 to 0.71 for lateral bending (Table 1).

DISCUSSION:  This interim analysis suggests that lumbar spine intervertebral motion sharing may be a promising kinematics biomarker for classifying individuals with cLBP, with lateral bending providing more accurate clusters than flexion or extension. These findings need to be confirmed after processing the entire dataset. Future work will identify which movements and motion segments produce the most accurate clustering of individuals with cLBP and explore how kinematics clusters can improve diagnostics and/or treatment.