The interobserver agreement for 68Ga-PSMA-11 PET/CT study interpretations in patients with prostate cancer is unknown. Methods:68Ga-PSMA-11 PET/CT was performed in 50 patients with prostate cancer for biochemical recurrence (n = 25), primary diagnosis (n = 10), biochemical persistence after primary therapy (n = 5), or staging of known metastatic disease (n = 10). Images were reviewed by 16 observers who used a standardized approach for interpretation of local (T), nodal (N), bone (Mb), or visceral (Mc) involvement. Observers were classified as having a low (<30 prior 68Ga-PSMA-11 PET/CT studies; n = 5), intermediate (30-300 studies; n = 5), or high level of experience (>300 studies; n = 6). Histopathology (n = 25, 50%), post-external-beam radiation therapy prostate-specific antigen response (n = 15, 30%), or follow-up PET/CT (n = 10, 20%) served as a standard of reference. Observer groups were compared by overall agreement (% patients matching the standard of reference) and Fleiss' κ with mean and corresponding 95% confidence interval (CI). Results: Agreement among all observers was substantial for T (κ = 0.62; 95% CI, 0.59-0.64) and N (κ = 0.74; 95% CI, 0.71-0.76) staging and almost perfect for Mb (κ = 0.88; 95% CI, 0.86-0.91) staging. Level of experience positively correlated with agreement for T (κ = 0.73/0.66/0.50 for high/intermediate/low experience, respectively), N (κ = 0.80/0.76/0.64, respectively), and Mc staging (κ = 0.61/0.46/0.36, respectively). Interobserver agreement for Mb was almost perfect irrespective of prior experience (κ = 0.87/0.91/0.88, respectively). Observers with low experience, when compared with intermediate and high experience, demonstrated significantly lower median overall agreement (54% vs. 66% and 76%, P = 0.041) and specificity for T staging (73% vs. 88% and 93%, P = 0.032). Conclusion: The interpretation of 68Ga-PSMA-11 PET/CT for prostate cancer staging is highly consistent among observers with high levels of experience, especially for nodal and bone assessments. Initial training on at least 30 patient cases is recommended to ensure acceptable performance.
