Interobserver reliability in the interpretation of diagnostic lumbar MRI and nuclear imaging

Abstract

Background context Assessment of patients with complaints of low back or leg pain varies with the subspecialty of the treating physician. The evaluation of the spine patient may include magnetic resonance imaging (MRI), bone scan, or single-photon emission computed tomography (SPECT) imaging. The interpretation of these tests and the examiner's biases will impact the outcome of patient treatment and the cost to the health-care system. Purpose To evaluate interobserver reliability of MRI and nuclear imaging studies and determine the predictability of nuclear medicine results based upon MRI findings. Study design/setting Retrospective radiographic review for patients with low back pain. MRI, planar bone scan imaging, and SPECT imaging techniques were evaluated. Patient sample Seventeen patients (80 lumbar levels) who presented to an orthopedic spine specialist with a complaint of mechanical low back or leg pain were randomly selected. Inclusion criteria were age greater than 21 years and a workup that included MRI, bone scan, and SPECT scan images. Outcome measures Interobserver reliability was determined using kappa values (0.6 to <0.8=substantial; 0.8–1=perfect). Interobserver reliability was evaluated for all studies and compared with a “group consensus.” Methods A team of orthopedists and radiologists of varying experience levels were assembled to interpret the imaging studies. All readings were performed independently followed by group interpretation and discussion. The reviewers were asked to assess each lumbar level (L1–L2 through L5–S1). Phase one: Is the level degenerative? Is there a spondylolisthesis present? Does the level have Modic changes on MRI? Phase two: Based on the MRI, will the planar bone scan be positive? After this prediction, was the planar scan positive? Identify the lesion location (anterior column vs. posterior column). Phase three: Based on their assessment and predicted results on planar bone scan, was the SPECT scan able to improve this assessment? Results High kappa values were demonstrated in the identification of a degenerative disc, spondylolisthesis, and Modic change (0.773, 0.728, and 0.669, respectively). Bone scan and SPECT scan yielded poorer kappa results (0.539 and 0.460, respectively). Reviewer-predicted bone scan results demonstrated a positive predictive value of 68% and a negative predictive value of 84%. Predicated SPECT results were similar (positive predictive value 66% and negative predictive value 84%). SPECT identified 24% more lesions in the lumbar spine when compared with bone scan. Conclusions MRI interpretation of the lumbar spine is comparable between specialties. Nuclear imaging studies (bone scan/SPECT) demonstrated a poorer correlation between examiners. The presence of MRI changes enables an accurate prediction of bone scan or SPECT scan findings. SPECT scan demonstrates an increased sensitivity in the detection of spinal abnormalities and the ability to localize a lesion when compared with planar bone scan.