静止超音波画像における内側肘関節裂隙幅の術者内および術者間信頼性：横断的信頼性研究

Quantifying the reproducibility of ultrasound measurements of the medial elbow joint space is essential for interpreting changes over time and for comparing raters. However, measurement reliability under controlled, acquisition-independent conditions remains limited. Herein, we aimed to quantify the intra-rater repeatability and inter-rater agreement for static medial elbow ultrasound measurements exclusively focusing on the analytical (measurement) component in healthy elbows.

In this cross-sectional reliability study, 130 high-quality static ultrasound images from healthy elbows (13 elbows × 10 images each) were acquired by a single experienced examiner using standardized limb positioning and evaluated twice by seven orthopedic surgeons with varying ultrasound experience. The medial elbow joint space width was defined as the narrowest distance between the distal medial edge of the humeral trochlea and the proximal edge of the ulnar trochlear notch. Measurements were blinded and repeated following a ≥ 2-week interval. Reliability was assessed using the following intraclass correlation coefficients: ICC(1,1) for intra-rater repeatability and ICC(2,1) for inter-rater agreement, with error metrics expressed as the standard error of measurement (SEM) and the minimal detectable change at 95% confidence (MDC95). In addition, the mean, standard deviation, and range were calculated to provide reference values.

Intra-rater ICC(1,1) spanned 0.860-0.973, with SEM 0.14-0.39 mm (MDC95 0.39-1.09 mm). Using two-session averages, inter-rater ICC(2,1) was found to be 0.872, with SEM 0.30 mm (MDC95 0.82 mm). Systematic measurement biases were observed, while more experienced raters showed narrower error ranges. The mean medial elbow joint space width in healthy elbows was 2.8 ± 0.8 mm (range 1.5-4.8 mm).

In this study of static ultrasound images from healthy elbows, intra-rater variability was found to be small, while inter-rater agreement was good among seven orthopedic surgeons with varying experience. These findings reflect measurement reproducibility under standardized static image conditions, independent of image acquisition variability. For practical interpretation, an inter-rater threshold of approximately 0.8-0.9 mm was required to exceed the measurement noise in this dataset, supporting the use of two-session averaging when feasible. Because the present study only evaluated the measurement component, these results likely represent the upper limit of achievable reliability in clinical practice.