4514. Validation and Utility of Radiology-Pathology Concordance Categories for Ultrasound-Guided Targeted Liver Biopsies
Authors * Denotes Presenting Author
  1. Miriel Handler *; Beth Israel Deaconess Medical Center
  2. Francesca Rigiroli; Beth Israel Deaconess Medical Center
  3. Sara Babapour; Beth Israel Deaconess Medical Center
  4. Timothy Sotman; Beth Israel Deaconess Medical Center
  5. Alexander Brook; Beth Israel Deaconess Medical Center
  6. Olga R Brook; Beth Israel Deaconess Medical Center
This study aims to validate reproducibility and assess outcomes of concordant, discordant, and indeterminate radiology-pathology concordance categories for ultrasound (US)-guided targeted liver biopsies.

Materials and Methods:
This IRB-approved, HIPAA-compliant study included consecutive patients that underwent US-guided targeted liver biopsies between July 1, 2016 and July 1, 2021 in a tertiary academic institution. Three procedural radiologists with variable clinical experience independently and retrospectively determined concordance between radiology findings and pathology results while blinded to the outcomes. The result was considered concordant when intraprocedural imaging confirmed optimal sampling, and the pathology was consistent with imaging findings. All malignant results were considered to be concordant. The result was discordant when the pathology showed no lesional tissue or benign pathology with worrisome radiologic features. The result was indeterminate when intraprocedural imaging confirmed sampling of the target lesion, and the pathology could potentially explain imaging findings, however it was not classic. Pathology of repeated biopsy or the surgical sample was used as the reference standard. When not available, subsequent imaging or clinical follow up was used as the reference standard. Agreement between three proceduralists’ assessments was assessed with Fleiss’s kappa, and the rate of malignancy during the follow-up was determined for each category.

Of the US-guided biopsies, 778/894 (87%) had malignant diagnosis and thus deemed concordant. Five of 116 biopsies with nonmalignant results were excluded due to lack of confirmed final diagnosis. Of nonmalignant results, 52/111 (47%) biopsies were classified as concordant, 45/111 (41%) discordant, and 14/111 (13%) indeterminate with imaging findings. On follow up of non-malignant biopsies, 26/45 (58%) of discordant, 4/14 (29%) indeterminate, and 0/52 (0%) concordant results were malignant. The prevalence of malignancy in the three groups was significantly different (<em>p</em> <0.001). Interobserver agreement was substantial with Fleiss’s kappa of 0.77. Segment VIII lesions had highest rate of discordant cases, 11/17 (65%) vs 34/94 (36%), <em>p</em> = 0.03.

Validation of retrospectively assigned radiology-pathology concordance categories showed substantial interreader agreement, high malignancy rate in the discordant category and moderate malignancy rate in indeterminate category of US-guided targeted liver biopsies. Clinical relevance: assessment of radiology-pathology concordance for US-guided targeted liver biopsies can aid in determining risk of missed cancer diagnosis and suggesting the next steps in patient’s management.