Interobserver and intraobserver agreement on prostate-specific membrane antigen PET/CT images according to the miTNM and PSMA-RADS criteria.

Abstract

In this study, we aimed to measure interobserver and intraobserver agreement in Ga-68-prostate-specific membrane antigen (PSMA) PET/computed tomography (CT) image interpretation. In addition, the limitations of these criteria and levels of personal confidence reported by the readers when reporting the findings were determined. The effects of interpersonal differences on clinical decisions were also investigated. PSMA PET images from 133 cases were reported independently by four different readers at different times according to the molecular imaging TNM (miTNM) and PSMA-reporting and data system (RADS) templates. There was substantial interobserver agreement for overall positivity, miT, miN and miM staging (Fleiss' κ = 0.65, 0.625, 0.731, and 0.779). Substantial agreement levels were observed in reporting of seminal vesicle invasion, the number of lymph node stations with metastasis, total number of intraprostatic areas containing tumors, and lymph node metastasis staging (Fleiss' κ = 0.622 and 0.779). The highest variation was seen in the reporting of intraprostatic distribution: In International Society of Urological Pathology (ISUP) grade group 1, moderate agreement was observed, and it was seen that the agreement level for the T staging increased with an increasing ISUP group in the staging group (Fleiss' κ = 0.531 vs. 0.655). There was near-perfect interobserver agreement in the reporting of five-point PSMA-RADS scoring [intraclass correlation coefficient (ICC) κ = 0.904; 95% CI, 0.865-0.934]. Disagreement according to miTNM staging had a major effect on clinical management in only 9% (n = 12) of the patients. PSMA PET has a lower interobserver variability and higher reproducibility than other imaging methods used for imaging of prostate cancer do, including CT, MRI, and bone scintigraphy. The miTNM template provides a reporting format that is highly reproducible and has a high level of agreement among readers, but the prostatic template needs development. In contrast, the PSMA-RADS system leads to slightly increased interobserver reporting differences and reduces personal confidence, but at the same time, it still exhibits almost-perfect agreement in terms of scoring.