Introduction: Studies of endplate structural defects (EPSD) may further the understanding of pathoanatomical mechanisms underlying back pain. However, with CT, as with MRI, the common measurement methods used to document EPSD have not been validated, leaving uncertainty about what the observations represent or how accurately they capture the presence or absence of EPSDs. This study aims to compare and determine the reliability and validity of two common EPSD assessment methods using clinical-CT, with micro-CT as the reference standard.

Methods: Using an evaluation manual, 418 endplates on sagittal slices of clinical-CT obtained from 19 embalmed cadavers (9 men and 10 women, aged 62-91 years) were independently assessed by three raters (two experienced radiologists and a novice) for EPSD using two methods (Brayda-Bruno: wavy/irregular, notched, Schmorl’s nodes and facture; and Feng: focal, corner, and erosive defects), with two-week intervals between each assessment, blinded to all prior assessments. A Nominal Group Technique was followed to resolve discordance and achieve consensus. The corresponding micro-CT from the harvested T7-S1 spines were assessed by another independent rater with excellent intra-rater agreement (Kappa=0.955, 95%CI: 0.912–0.998). All data were analyzed descriptively and inferentially for reliability and validity, including sensitivity and specificity, on SPSS version 22. The study was prospectively registered at ClinicalTrial.gov (NCT04808960). 

Results:  Inter-rater agreement between each pair of raters was good (Kappa=0.60-0.69) for assessing the presence of EPSD using both Brayda-Bruno and Feng assessments. However, poor agreement was observed for the assessment of specific EPSD phenotypes according to Brayda-Bruno (Kappa=0.43) and Feng (Kappa=0.58). All normal endplates according to Brayda-Bruno (n=151) were classified as normal using Feng. However, 26% (n=58) of normal endplates by Feng corresponded to wavy/irregular endplates in the Brada-Bruno system for which a category did not exist in Feng. Similarly, erosion, for which Brayda-Bruno lacked a category, was mainly (82.8%) classified as wavy/irregular. While the majority of notched defects (n=15, 46.9%) and Schmorl’s nodes (n=45, 79%) using Brada-Bruno’s classification were recorded as focal defects using Feng’s. When compared to micro-CT, fractures (n=53) and corner defects (n=28) were routinely missed on clinical-CT. Endplates classified as wavy/irregular on clinical CT corresponded to erosion (n=29, 21.2%), jagged defects (n=21, 15.3%) and calcification (n=19, 13.9%) on micro-CT. Schmorl’s nodes assessed on clinical CT often represented erosion (n=16, 32%) on micro-CT and focal defects represented endplate fractures (n=21, 27.6%). Overall, there was a sensitivity of 70.9% and specificity of 79.1% for Feng’s method, and 79.5% and 57.5%, respectively, using Brayda-Bruno’s. Schmorl’s nodes (60%) and Focal defects (52%) had the highest sensitivity. All EPSD clinical-CT and micro-CT dimensions significantly correlated (p<0.001), except defect depth. 

Discussion: There is good reliability and support for the validity of assessing the presence of EPSD with the two classification methods using clinical CT. However, neither method contained sufficient EPSD phenotypes to provide the needed specificity. The Brayda-Bruno system lacked a category for erosion, and Feng lacked a category for wavy/irregular defects. A standardized, combined method, with a publicly available measurement atlas, is needed to improve reliability and improve specificity of EPSD presence and interpretation.