Introduction: The specificity of current models for the preoperative prediction of lymph node invasion (LNI) is relatively low. Many times, the lymph nodes removed are negative and in case of nodal involvement, this is rarely bilateral. On these premises, we aimed to test the feasibility of unilateral lymph node dissection in a population of patients diagnosed and staged after the routine introduction of mpMRI-targeted biopsy in clinical practice.
Methods: We relied on a multi-institutional dataset encompassing data from 15 European centers. All patients were cN0 and underwent prostatectomy and bilateral extended pelvic lymph node dissection (ePLND). Prostatic lobe involvement was considered as side-specific. Similarly, LNI was considered as side specific. The outcome of the study was the presence, on final pathology, of uni- or bilateral LNI. Initially, we evaluate the rate of LNI contralaterally to the index lesion in case of absence of high risk (HR) features (non-organ confined at mpMRI, GG=4 and PSA>20). Subsequently, a model to predict LNI (ipsi- and contralaterally to the index lesion) including side-specific variables was generated. Ultimately, the rate of bilateral LNI was evaluated based on the predicted LNI probability.
Results: A total of 1,816 patients with complete data were identified, 1,072 (59%) had HR features. Median (IQR) age was 67 (62, 71) years, median (IQR) PSA was 8.3 (5.8, 13.5) ng/ml. Lymph node involvement was documented in 233 (13%) patients: specifically, 185 (10.2%) had unilateral LNI and 61 (3.2%) had bilateral LNI, in 54 (89%) of them, patients had HR features. In absence of HR features and index lesion’s GG1-2 (319 patients) and contralateral GG1 or negative biopsy, LNI was not documented contralaterally in any case. In case of index lesion’s GG 3 and contralateral GG1 or negative biopsy, LNI was documented in 1 (0.4%) patient. We created a multinomial model for ipsilateral and contralateral LNI including preoperative PSA, extracapsular extension on mpMRI, seminal vesicle invasion on MRI, GG in the index lesion, GG contralaterally to the index lesion, maximum lesion’s diameter and percentage of positive cores in the contralateral lobe to the index lesion. The model for LNI contralaterally to the index lesion had an AUC of 84% (95% CI: 79-89). With a cutoff of 1%, 764 contralateral lymph node dissections would be omitted with 1 (0.4%) contralateral LNI missed.
Conclusions: Our findings suggest that ePLND could be omitted contralaterally to the index lesion among individuals without HR clinical features within a threshold probability of LNI of 1%. Irrespective of the predictive model for LNI that each center uses, our model can be applied to decide whether unilateral ePLND could be considered whenever a ePLND is indicated.