Abstract
Abstract
Purpose
To develop and internally validate a low-risk selection model to guide omission of completion inguinal lymph node dissection (ILND) after a positive dynamic sentinel node biopsy (DSNB) and to identify predictors of an inguinal basin harboring additional inguinal lymph node metastases following a positive DSNB.
Methods
Data were drawn from the Danish National Penile Cancer Database (DaPeCa-data), comprising patients with penile cancer treated from 2000 to 2025. The analysis included 300 DSNB-positive inguinal basins, and each basin was analyzed independently. Candidate predictors were screened using a penalized logistic regression across multiple imputed datasets, with variables meeting a pre-specified selection threshold carried forward to a confirmatory multivariable model. An exhaustive search across combinations of clinical criteria was used to derive low-risk selection models, subject to a pre-specified safety constraint of a negative predictive value (NPV) ≥ 0.92.
Results
Among 300 DSNB-positive basins, 188 underwent ILND, of which 33 (18%) had additional nodal metastases. Metastasis size ≥ 10 mm, having more than one positive lymph node at DSNB, and bilateral positive DSNB were the strongest independent predictors of additional metastases after a positive DSNB. A two-criterion model (metastasis size ≤ 10 mm and no urethral invasion) spared 63 out of 248 evaluable basins (25.4%) from ILND, with 5 false-negative cases equal to an NPV of 0.921. A stricter three-criterion model spared 21 out of 248 basins (8.5%) with 1 false negative and an NPV of 0.952. Sensitivity analysis restricted to directly observed outcomes lowered the higher-yield model's NPV to 0.908.
Conclusions
In penile cancer patients with a positive DSNB, the strongest predictors of the presence of additional inguinal lymph node metastases were: sentinel node metastases of 10 mm or more; two or more positive sentinel lymph nodes and bilateral sentinel lymph node metastases. A simple, bedside-usable selection model may safely spare a meaningful subset of DSNB-positive patients from completion ILND, but the prediction performance was sensitive to outcome-imputation assumptions and requires external validation before clinical implementation.