LIMITED INTER-RATER RELIABILITY IN ANTEROLATERAL LIGAMENT DETECTION IN PEDIATRIC KNEE MRIS

BACKGROUND: Despite the rapidly expanding anterolateral ligament (ALL) literature base, little has focused on the reliability of its detection on MRI. This is especially important in the pediatric population where the scarcity of pediatric cadaveric specimens hinders the accurate estimation of its prevalence. PURPOSE: The purpose of this study was to assess the visibility of ALL on common MRI pulse sequences and to calculate the inter-rater reliability among multiple expert readers in a cohort of normal pediatric knee MRIs. METHODS: A series of 3T pediatric knee MRIs obtained from 8/2018-4/2019 were reviewed to identify patients without significant knee pathology. Knee laterality, age, gender, and regional physeal status were also recorded. One orthopedic surgeon, 1 pediatric musculoskeletal (MSK) radiologist, and 1 adult MSK radiologist, independently reviewed each MRI pulse sequence (T1-weighted, PD-weighted, and PD-weighted fat-suppressed images) separately, to determine the presence or absence of the 3 parts of the ALL (femoral, tibial, and meniscal proper). Inter-rater reliability was calculated using Fleiss’ Kappa (K). RESULTS: Inclusion criteria were met for 75 patient knees with a mean age of 12.7±3.42y (3-19) and a greater proportion of females (57.3%). The distribution of regional physeal status was: Open-48.0%, Closing-29.3%, and Closed-22.7%. Inter-rater reliability was slight to fair (K: 0.141-0.340) for visualizing any specific ALL component on a unique MRI sequence (Table 1). For visualizing a specific attachment on any MRI sequence, there was slight to moderate agreement (K: 0.018-0.411). Visualizing any ALL attachment on any MRI sequence also yielded slight agreement (K: 0.190). Inter-rater reliability between specialities (i.e. Ortho vs. MSK Radiology) did not differ substantially from group findings. No single study sequence demonstrated enhanced reliability for all ALL components. CONCLUSION: The reliability of ALL detection on MRI across multiple clinicians was slight to moderate for all components in this cohort of normal 3T pediatric knee MRIs. No single sequence was consistently better for ALL detection agreement. This is in contrast with previous work showing substantial agreement among raters and describing PD-weighted imaging as preferable for viewing the ALL. Experience in musculoskeletal imaging did not confer improved agreement. Establishing diagnostic reliability should be a prerequisite for use of an imaging finding in the research or clinical setting. Existing literature drawing conclusions from ALL findings on MRI may need to be interpreted with caution if reliability testing is not described.