Comparison of the Diagnostic Effectiveness of a Complete Logistic Model and Sentinel Lymph Node Biopsy in Patients with Skin Melanoma

Objective of the study. To compare the diagnostic effectiveness of the full logistic model and sentinel lymph node biopsy (SLNB) in predicting metastatic involvement of regional lymph nodes in patients with cutaneous melanoma in relation to the morphological status of lymph nodes. Material and methods. A single-center prospective study included 121 patients with cutaneous melanoma. Data from histological/immunohistochemical (IHC) examination of regional lymph nodes were used as the reference standard. The full logistic model was constructed on the basis of clinical, ultrasound, elastographic, scintigraphic, and morphological features; the result was expressed as the predicted probability of metastases (PIGX). The results of SLNB were analyzed as a binary test (metastases: yes/no). For both models, sensitivity, specificity, overall accuracy, and the area under the ROC curve (AUC) were calculated. Results. In comparison with IHC examination, SLNB demonstrated a sensitivity of 77.8%, specificity of 100%, and overall accuracy of 95.0%, with an ROC AUC of approximately 0.89. The full logistic model showed an ROC AUC of approximately 0.95; at a threshold of p = 0.272, its sensitivity was 85.2%, specificity 95.7%, and accuracy 93.4%, and at p = 0.349 – 81.5%, 97.9%, and 94.2%, respectively. In the range of thresholds around p ≈ 0.35, the full logistic model demonstrates a combination of sensitivity, specificity, and accuracy close to the characteristics of SLNB. Conclusion. The full logistic model has high discriminative ability and, with optimal threshold selection, provides diagnostic accuracy comparable to SLNB, with slightly higher sensitivity. The model may be considered as an additional tool for risk stratification and decision support regarding the need for invasive staging of regional lymph nodes.