Can We Reliably Grade Elbow Sclerosis on CT? New Study Challenges Assumptions

Inter- and Intraobserver Reliability of CT Assessment of Canine Elbow Sclerosis in the Absence of Elbow Pathology

Luke F. Ellis, Arthur K. House, Mika Frances, Rebekah Knight, Helen Dirrig, Alison Hillman

Background

Elbow dysplasia in dogs encompasses several developmental joint pathologies, including medial coronoid process disease (MCPD), and remains diagnostically challenging. While CT is widely used for assessment, interpretation of elbow sclerosis—particularly in the medial coronoid process (MCP) and medial aspect of the humeral condyle (MAHC)—is often subjective. Objective quantification using Hounsfield units (HU) has been proposed, yet methodologies vary and repeatability remains unclear. The authors aimed to establish a highly standardized CT examination protocol and evaluate inter- and intraobserver reliability for subjective and objective assessment of elbow sclerosis in dogs without overt elbow pathology. They hypothesized that standardization would improve repeatability.

Methods

This retrospective study included 49 elbow CT scans from 25 clinically sound dogs over one year of age undergoing pre-breeding screening and diagnosed as free of elbow dysplasia. CT acquisition and multiplanar reconstruction were standardized using defined anatomical landmarks. Five blinded observers (two board-certified surgeons, two board-certified radiologists, one surgical resident) independently evaluated randomized studies. Sclerosis at the MCP (traditional and modified transverse planes) and MAHC (dorsal plane) was graded subjectively (0 = none, 1 = mild, 2 = marked) using reference images and written criteria. Objective assessment was performed by measuring HU within standardized regions of interest. Two observers repeated evaluation of 30 randomly selected scans. Inter- and intraobserver reliability were assessed using Kendall’s coefficient of concordance (W), weighted kappa, and intraclass correlation coefficients (ICC). Correlation between ordinal scores and HU measurements was analyzed using linear regression.

Results

Complete agreement among observers for ordinal sclerosis grading was moderate: 38.8% (MCP traditional), 28.6% (MCP modified), and 26.5% (MAHC). Kendall’s W indicated moderate interobserver reliability across regions. Intraobserver reliability for ordinal grading ranged from weak to moderate.

For HU measurements, interobserver reliability was good for the MCP in the traditional plane (ICC 0.85) and moderate for the MCP modified plane (ICC 0.69) and MAHC (ICC 0.67). Intraobserver reliability for HU measurements was good to excellent. Correlation between subjective sclerosis grades and HU values was low to moderate (adjusted R² 0.47–0.53).

Binary classification (presence vs. absence of sclerosis) did not meaningfully improve agreement. Specialty (surgery vs. radiology) accounted for negligible variance in scoring.

Limitations

The study population was relatively small and data distribution across sclerosis categories was imbalanced, limiting precision of reliability estimates. Most dogs had minimal or no sclerosis in certain regions, which affected kappa interpretation. Despite standardization efforts, some aspects of HU measurement (e.g., exact ROI placement, magnification, number of polygon points) were not fully standardized and may have contributed to variability. The retrospective design and single CT scanner setting may limit generalizability.

Conclusions

Despite implementation of a highly standardized CT acquisition and assessment protocol, both subjective and objective evaluation of elbow sclerosis in dogs without overt pathology demonstrated only moderate to poor repeatability. Correlation between ordinal grading and HU measurements was limited. These findings raise concerns regarding the clinical interpretation and reporting of elbow sclerosis in screening and diagnostic contexts, particularly in the absence of clear pathology. Further refinement of assessment methodologies is required before sclerosis can be reliably used as a meaningful diagnostic descriptor.

Disclaimer: The summary generated in this email was created by an AI large language model. Therefore errors may occur. Reading the article is the best way to understand the scholarly work. The figure presented here remains the property of the publisher or author and subject to the applicable copyright agreement. It is reproduced here as an educational work. If you have any questions or concerns about the work presented here, reply to this email.