Interobserver and intraobserver agreement in PET/CT with [18F]DCFPyL according to TNM molecular and PSMA-RADS 2.0 criteria

Abstract

Purpose

The aim of this study was to determine the agreement between three observers with different levels of experience using the PSMA-RADS 2.0 criteria and the miTNM system for the interpretation of PET-PSMA with [¹⁸F]DCFPyL in males with prostate cancer.

Materials and Methods

PET-PSMA images from 114 prostate cancer patients were blindly reported twice by three different observers at intervals of 8 weeks. The evaluations were performed according to the molecular imaging TNM (miTNM) and PSMA-RADS 2.0 criteria. We used Fleiss’ Kappa to analyse inter and intraobserver agreements.

Results

Moderate overall agreement was obtained in the assessment of the PET-PSMA results (Fleiss'k = 0.53; 95% CI 0.45−0.62; p < 0.001), with significant agreement in the miT, miN and miM reports. There was a substantial level of agreement in the reporting of prostatic disease and lymphatic involvement (Fleiss'k = 0.66 and 0.65), being lower than that observed in the reporting of metastatic disease (Fleiss'k = 0.86), especially in the M0 group (Fleiss'k = 0.99). Upon re-evaluation of the images, observer 1 had moderate overall agreement for miT (Fleiss'k = 0.51) and substantial agreement for miN and miM (Fleiss'k 0.75 and 0.63, respectively).

Conclusions

The use of a structured scoring system such as PSMA-RADS 2.0, as well as the miTNM classification system in the interpretation of PET-PSMA images in prostate cancer patients, provides a highly reproducible report format. High levels of interobserver and intraobserver agreement are found, especially when ruling out disease, which supports its use in routine clinical practice.