Evaluation of the TARGET score for MRI-based detection of prostate cancer recurrence in patients treated with HIFU hemiablation.

[PURPOSE] To evaluate diagnostic performance and inter-reader agreement of the Transatlantic Recommendations for Prostate Gland Evaluation with MRI after Focal Therapy (TARGET).

[METHODS] This retrospective, multi-center study included patients who underwent high-intensity focused ultrasound (HIFU) hemiablation for localized prostate cancer (PCa), with follow-up multiparametric MRI (mpMRI) at 12 months and subsequent biopsy (reference standard). Three radiologists independently assessed the mpMRIs using the TARGET score (treated prostate side) and the PI-RADS score (v2.1, untreated side). Diagnostic performance metrics and area under the receiver operating characteristic curves (AUC) were computed for any-/clinically significant-PCa (cs-PCa) and two scoring thresholds (3 vs. 4). Inter-reader-agreement was analyzed using Fleiss к.

[RESULTS] Of all 40 patients included, 23 had biopsy-proven recurrent PCa in 26 lesions, with seven lesions harboring cs-PCa (5/7 on the treated side). TARGET showed moderate to high sensitivity (Reader 1-3: 60-100 %), high NPV (91-100 %) but lower specificity (54-69 %) and PPV (16-27 %) in the detection of cs-PCa, given a scoring threshold of ≥ 3. The diagnostic performance for cs-PCa (AUC: 0.57-0.79) exceeded that of any-PCa (0.54-0.64). PI-RADS showed an overall moderate efficacy in any-/cs-PCa-detection (untreated prostate side, AUC: 0.52-0.60). Overall inter-reader agreement was moderate (TARGET к = 0.60-0.64, PI-RADS к = 0.54-0.55).

[CONCLUSIONS] The strength of the TARGET score was its high NPV, implying its potential to rule out cs-PCa. However, the moderate sensitivity and specificity, low PPV and moderate inter-reader agreement indicate a limited reliability of TARGET to consistently identify cs-PCa recurrence. The generalizability of these findings is restricted by the overall small sample size.