Background and Aim
Accurate estimation of sentinel node metastasis risk is critical for counselling and selection of patients with cutaneous melanoma for sentinel lymph node biopsy (SLNB). The Melanoma Institute Australia (MIA) Sentinel Node Metastasis Risk Prediction Tool is widely used but cautions that performance may vary outside the population in which it was developed. This study aimed to validate the calibration of the MIA tool in a regional Australian surgical centre.
Methods
A retrospective audit was conducted of patients undergoing wide local excision with SLNB for primary cutaneous melanoma at Toowoomba Hospital between January 2021 and December 2025. For each patient, the predicted probability of sentinel node positivity generated by the MIA tool was recorded. Model calibration was assessed by comparing observed versus predicted nodal positivity overall and across 10% predicted-risk bands. Calibration was further evaluated using a locally estimated scatterplot smoothing (LOESS) curve, plotting predicted risk against observed nodal outcomes, and compared with the line of identity.
Results
A total of 107 patients underwent SLNB during the study period, of whom 21 (19.6%) had a positive sentinel node. Observed nodal positivity increased with higher predicted-risk bands. In low-risk categories (≤19%), observed positivity was lower than predicted, while higher-risk categories demonstrated closer agreement. LOESS calibration analysis demonstrated over-prediction of nodal metastasis in lower-risk patients, with improved calibration at higher predicted probabilities.
Conclusions
In this regional cohort, the MIA Sentinel Node Metastasis Risk Prediction Tool demonstrated acceptable calibration overall but tended to over-predict sentinel node positivity in lower-risk patients. Local validation using real-world data provides important benchmarking information and may support more tailored counselling for patients considering SLNB in rural and regional practice.