To present interobserver variability in thecal sac (TS) delineation based on contours generated by 8 radiation oncologists experienced in spine stereotactic body radiation therapy and to propose contouring recommendations to standardize practice.

In the setting of a larger contouring study that reported target volume delineation guidelines specific to sacral metastases, 8 academically based radiation oncologists with dedicated spine stereotactic body radiation therapy programs independently contoured the TS as a surrogate for the cauda equina and intracanal spinal nerve roots. Uniform treatment planning simulation computed tomography datasets fused with T1, T2, and T1 post gadolinium magnetic resonance imaging for each case were distributed to each radiation oncologist. All contours were analyzed and agreement was calculated using both Dice similarity coefficient and simultaneous truth and performance level estimation with kappa statistics.

A fair level of simultaneous truth and performance level estimation agreement was observed between practitioners, with a mean kappa agreement of 0.38 (range, 0.210.55) and the mean Dice similarity coefficient (± standard deviation, with range) was 0.43 (0.36 ± 0.1 to -0.53 ±0.1). Recommendations for a reference TS contour, accounting for the variations in practice observed in this study, include contouring the TS to encompass all the intrathecal spinal nerve roots, and caudal to the termination of the TS, the bony canal can be contoured as a surrogate for the extra thecal nerves roots that run within it.

This study shows that even among high-volume practitioners, there is a lack of uniformity when contouring the TS. Further modifications may be required once dosimetric data on nerve tolerance to ablative doses, and pattern of failure analyses of clinical data sets using these recommendations, become available. The contouring recommendations were designed as a guide to enable consistent and safe contouring across general practice.