Kinematic assessment of lumbar segmental instability using digital fluoroscopic video

Lumbar segmental instability (LSI) has been a theoretical and controversial source of low back pain, largely because of the lack of consensus on what constitutes LSI. Digital fluoroscopic videos (DFV) have had limited success in measuring lumbar kinematics because of poor image quality and associated measurement errors. The purposes of this study were to develop a reliable DFV technique to measure lumbar kinematics and determine if the resulting variables distinguish between patients suspected to have LSI and healthy control subjects. A technique that combined digital image processing and distortion compensation was developed to measure lumbar vertebral kinematics using DFV. In a reliability study with a group of 20 subjects, the average intra-image reliability (ICC) was .986. The average inter-image reliability was .878. The 95% confidence interval for inter-image measurement error was 2° and 1.2 mm. This technique was applied to two symptom-based groups of subjects (20 with LSI and 20 healthy controls). The DFV were then analyzed by three spine surgeons to determine normality of movement. Subsequently, the groups were reorganized into two motion-based groups (11 with LSI and 14 healthy controls). Independent t-tests were used to compare the differences between those with LSI and healthy controls. Variables with a p <.20 and a positive likelihood ratio (+LR) >2.0, based on a cut-off score on a receiver operator characteristic curve, were considered as possible candidates for a model to distinguish group membership. A 10 variable model was developed when the reference criterion was the symptom-based groups. This model had the greatest accuracy (87.5%, sensitivity = .95, specificity = .80) when subjects had four or more of the variables present. An eight variable model (+LR >2.5) was developed to distinguish the motion-based groups. This model’s greatest accuracy was 96%. The higher +LR values and the greater accuracy of this model demonstrate the effectiveness of expert review process to obtain more homogenous groups. The technique developed was both reliable and successful in using a cluster of kinematic variables to discriminate between group memberships. These models provide a foundation for the development of a diagnostic prediction rule.