Please login or create an account. If you do not have access to this content, you will be shown a 30 second preview and licensing options.
- Presentation
Validation of the Melanoma Institute of Australia Sentinel Node Prediction Tool
Description
The speaker describes a validation study of the Melanoma Institute of Australia Sentinel Node Prediction Tool, a nomogram designed to estimate an individual patient’s risk of sentinel lymph node metastasis using clinical and pathologic factors rather than broad staging buckets alone. Using examples of two T1A melanoma patients with very different risk profiles, the talk highlights how the tool can produce personalized estimates that may better inform decisions about sentinel lymph node biopsy. The study validated the tool in about 60,000 patients from the National Cancer Database and found that its discrimination was good, with an AUC of 0.733, similar to the original cohort’s 0.739. Applying a 10% biopsy threshold would have reduced the number of biopsies performed by about half while increasing the positivity rate from 14.6% to 22.8%. However, the tool had weaker calibration in the lowest-risk group, which includes many T1A and T1B tumors, and individual confidence intervals can meaningfully affect decision-making. Overall, the tool is presented as a free, practical aid for personalized risk assessment, but not a perfect replacement for clinical judgment, and further prospective validation is recommended.
View moreConclusions
- The Melanoma Institute Australia sentinel node metastasis prediction tool provides more individualized risk estimates than standard stage-based buckets by incorporating clinical and pathologic features.
- In a large NCDB external validation cohort, the tool showed fair discrimination with an AUC of 0.733, closely matching its original performance of 0.739.
- Using the tool’s 10% risk threshold would have reduced the number of patients offered sentinel lymph node biopsy while increasing the positivity rate among those selected for biopsy.
- The model appears most uncertain in the lowest-risk tumors, where calibration was poorer and many T1a/T1b cases fall.
- Because confidence intervals can be wide and net benefit results are mixed, the tool should be used as an aid to decision-making rather than as a definitive standard.
- The prediction tool is attractive because it is free and readily available, but further prospective validation is still needed before it can be considered fully established.
- Caleb Freeman, MD
- External validation of the Melanoma Institute Australia Sentinel Node Metastasis Risk Prediction Tool using the National Cancer Database#10.1016/j.jaad.2026.05.096
- 60,165 patients underwent sentinel lymph node biopsy and had the needed clinicopathologic factors
- AUC was 0.733 in the NCDB cohort compared with 0.739 in the original MIA cohort
- AUC: 0.733 (95% CI: 0.726 to 0.739)