The Spinal Instability Neoplastic Score (SINS): An Analysis if Reliability and Validity from the Spine Oncology Study Group

Abstract

Spinal instability due to metastases is an indication for surgery, rather than radiotherapy. The Spine Oncology Study Group (SOSG) developed the Spinal Instability Neoplastic Score (SINS), to help radiation oncologists determine instability via a scoring system by quantifying several clinical and radiologic factors. SINS consists of 6 component variables: spine location, pain, lesion bone quality, radiographic alignment, vertebral body collapse and postero-lateral involvement of the spine by tumor. Relevant clinical and radiographic data from 30 de-identified cases of spinal tumor were assessed by SOSG members. Members were asked to categorize each case as “stable”, “potentially unstable”, and “unstable”, and the consensus opinion was used as the gold standard for predictive validity testing. On two occasions at least 6 weeks apart, each rater scored each case using SINS. Interobserver and intraobserver reliability for each SINS component was assessed using the Fleiss and Cohen κ statistics, respectively. Each total score was converted to a 3-category data field, with 0-6 as “stable”, 7-12 as “potentially unstable”, and 13-18 as “unstable”. Predictive validity was assessed using Cohen's κ for agreement between SINS categorization and the consensus opinion. The κ statistics for interobserver reliability were 0.790, 0.841, 0.244, 0.456, 0.462 and 0.492 for the fields of location, pain, bone quality, alignment, vertebral body collapse and postero-lateral involvement, respectively. The κ statistics for intraobserver reliability were 0.806, 0.859, 0.528, 0.614, 0.590 and 0.662 for the same respective fields. Intraclass correlation coefficients for inter- and intraobserver reliability of total SINS score were 0.846 (0.773 - 0.911, 95% CI) and 0.886 (0.868 - 0.902, 95% CI), respectively. The κ statistic for predictive validity was 0.712 (0.676 - 0.766, 95% CI). The sensitivity and specificity of SINS to detect potentially unstable or unstable lesions was 95.7% and 79.5%, respectively (false negative rate 4.3%). SINS demonstrated near-perfect interobserver and intraobserver reliability in determining 3 clinically relevant categories of stability. Among the components of SINS, spine location and pain had the highest levels of agreement among observers, while bone quality was the weakest. Final SINS category showed substantial agreement with the a priori determined “consensus opinion” (gold standard) and had low type II error (4.3%). SINS is an important tool for oncologists to determine appropriate surgical referrals. Real-world application of this classification system will need to be evaluated in a prospective fashion.