Combining clinical exams can better predict lumbar spine radiographic instability

Several clinical tests have been proposed to diagnose lumbar instability, but their accuracy is still in question. The primary purpose of this study was to evaluate the diagnostic accuracy of the clinical lumbar instability tests. The secondary goal was to design a model to detect lumbar instability.

A prospective diagnostic cross-sectional study.

A sample of 202 patients with chronic low back pain were participated in the study. Five lumbar instability tests including Aberrant movement, Passive lumbar extension, Prone segmental instability, H and I and pheasant tests were compared to flexion/extension radiography as the gold standard for diagnosing lumbar instability using two by two tables. Multiple Logistic Regression analysis was applied to develop a model using demographic information as well as the patients' pain intensity, disability level, lumbar lordosis and the clinical tests.

Among the five examined tests, Prone segmental instability, H and I and pheasant tests showed very small likelihood ratios and diagnostic odd's ratio. The largest values were for H and I test with the positive likelihood ratio of 1.28 (95% CI: 0.72 to 2.29) and diagnostic odd's ratio of 1.37 (95% CI: 0.66 to 2.83); the diagnostic accuracy measures were smaller for the other studied clinical tests. The model was developed using weight (t = 1.15, p = 0.03) and lumbar lordosis (t = 3.04, p = 0.00) (which showed a significant relationship with lumbar instability) and prone segmental instability test. The final model has the positive likelihood ratio of 2.07 (95% CI: 1.41 to 3.05) and diagnostic odd's ratio of 3.77 (95% CI: 2.03 to 7.01).

Each individual test had very small to no power in discriminating patients with lumbar instability. The developed model just slightly improved the accuracy of radiological instability detection.