Purpose: To determine inter-rater reliability of sarcoidosis-related computed tomography (CT) findings that can be used for scoring of thoracic sarcoidosis.
Materials and methods: CT images of 51 patients with sarcoidosis were scored by five chest radiologists for various abnormal CT findings (22 in total) encountered in thoracic sarcoidosis. Using intra-class correlation coefficient (ICC) analysis, inter-rater reliability was analysed and reported according to the Guidelines for Reporting Reliability and Agreement Studies (GRRAS) criteria. A pre-specified sub-analysis was performed to investigate the effect of training. Scoring was trained in a distinct set of 15 scans in which all abnormal CT findings were represented.
Results: Median age of the 51 patients (36 men, 70%) was 43 years (range 26 - 64 years). All radiographic stages were present in this group. ICC ranged from 0.91 for honeycombing to 0.11 for nodular margin (sharp versus ill-defined). The ICC was above 0.60 in 13 of the 22 abnormal findings. Sub-analysis for the best-trained observers demonstrated an ICC improvement for all abnormal findings and values above 0.60 for 16 of the 22 abnormalities.
Conclusions: In our cohort, reliability between raters was acceptable for 16 thoracic sarcoidosis-related abnormal CT findings.
Key points: • Thoracic sarcoidosis is common; knowledge on reliability of CT scoring is limited. • Scoring CT abnormalities in pulmonary sarcoidosis can achieve good inter-rater agreement. • CT scoring validation in thoracic sarcoidosis is important for diagnostic and prognostic studies.