Pre-test Ga-PSMA-ligand PET/CT positivity in early biochemical recurrent prostate cancer after radical prostatectomy-validation of a prediction model.

OBJECTIVES

The aim of this study was the validation of a recently established comprehensive and compact prediction model for Ga-PSMA-11-ligand positron-emission tomography (PET) positivity with an independent subsequent patient series.

METHODS

A total of 292 consecutive patients with early biochemical recurrence after radical prostatectomy and PSA values between 0.2 and 1 ng/ml who underwent Ga-PSMA-11-ligand PET/computed tomography (CT) between January 2016 and June 2017 were retrospectively included. The cohort was divided into a very low PSA value (0.2-0.5 ng/ml, n = 151) and a low PSA value (> 0.5-1 ng/ml, n = 141) subgroup. First, pre-test positivity probabilities for each patient were calculated according to the previously published comprehensive prediction model using all clinical variables (PSA value, ISUP grade group, T- and N-stage, patient under androgen deprivation therapy (ADT), previous radiation therapy) and the compact model using just the most predictive factors PSA value, ADT, and grade group. Then, all Ga-PSMA-11-ligand PET/CTs were analysed by one experienced nuclear medicine physician, and the results were correlated to the calculated pre-test probabilities.

RESULTS

In the very low PSA value subgroup, mean pre-test probability for positive findings in Ga-PSMA-11-ligand PET/CT was 57% (95% CI 55-60%) according to the compact model and 59% (95% CI 56-61%) according to the comprehensive model. In the low PSA value subgroup, mean pre-test probability was 72% (95% CI 70-74%) in the compact model and 74% (95% CI 72-76%) in the comprehensive model. After image analysis, 59% (89/151) of the patients in the very low PSA value subgroup revealed positive imaging findings. Seventy-nine percent (112/141) of the patients in the low PSA value subgroup presented with positive findings in the Ga-PSMA-11-ligand PET/CT. The accuracy (AUC) of the prediction models was 0.71 (95% CI 0.65-0.78) for the compact model and 0.74 (95% CI 0.68-0.80) for the comprehensive model.

CONCLUSION

External validation of the recently proposed prediction models showed a high concordance of the calculated pre-test probabilities and actual Ga-PSMA-11-ligand PET/CT findings in the validation cohort confirming the prediction models' ability to determine the presence of a positive lesion at Ga-PSMA-11-ligand PET. However, the predictive accuracy of the nomogram itself is suboptimal and should be used with caution. Furthermore, the model's generalizability may be hampered due to the study design (in-house validation). Nevertheless, given the limited health resources and the costs of hybrid imaging techniques, prediction models might be a benefit in patient selection.