Radiological phenotypes in pulmonary sarcoidosis: a reliability study of newly defined high-resolution computer...

An accurate morphological classification of distinct pulmonary phenotypes in sarcoidosis is lacking. Recently, a multinational Delphi study was conducted to reach a consensus on recognizable high-resolution computer tomography (HRCT) phenotypes in pulmonary sarcoidosis as a basis for a more distinctive classification. The reliability of these phenotypes has not yet been evaluated.

HRCT scans of adult sarcoidosis patients from the pulmonology department of a single sarcoidosis referral center were scored by three blinded independent readers. Seven phenotypes were distinguished as described in the Delphi study. They were divided into two subgroups: "non-fibrotic" and "likely-to-be fibrotic". Intra- and inter-reader reliability for scoring the predominant phenotype and the subgroup was assessed using weighted Kappa (K_w) statistics.

Forty-five patients (mean age, 47 years ± 12, 28 men) were included. For the scoring of the predominant phenotype, inter-reader reliability between all readers was substantial with an overall Fleiss' kappa of 0.69 (CI 0.622-0.765, P < .001). We observed a substantial inter-reader reliability between readers A and B (K _w of 0.76), between readers B and C (K_w of 0.66) and between readers A and C (K _w of 0.66). For the scoring of the subgroups "non-fibrotic" vs. "likely-to-be fibrotic," overall Fleiss' Kappa was substantial (K = 0.78, CI 0.607-0.944, P < .001). We observed a K _w score of 0.76 between readers A and B; 0.81 between readers A and C; 0.76 between readers B and C. Intra-reader reliability was substantial between the scores of A in scoring the predominant phenotypes (K _w of 0.71) and it was almost perfect in scoring the subgroups (K _w of 0.95). Intra-reader reliability was substantial between the scores of B in scoring the predominant phenotype (K _w of 0.66) and subgroups (K _w of 0.72).

The inter- and intra-reader reliability of the newly proposed HRCT phenotypes obtained from the Delphi study is very acceptable.

This study is the first to assess the reliability of these HRCT phenotypes and supports the use of them for classification purposes in future clinical and pathogenetic studies.