A neural network-based model was found to be an effective risk stratifying approach for patients with T1b melanoma. These findings were published in JAMA Dermatology.

The National Comprehensive Care Network guidelines recommend that patients with cutaneous melanoma who have a 5% or greater risk of positivity be considered for sentinel lymph node (SLN) biopsy.

To evaluate improved risk stratification approaches, the neural network-based approach clinicopathologic factors (thickness, mitoses, ulceration, age) plus molecular analysis (31-gene expression profiling) model (i31-GEP-SLNB) was used in this study. Patients with T1a-HR, T1b, and T2 disease were evaluated with the i31-GEP-SLNB model, and the net benefit of the method was compared with the standard approach.

Continue Reading

The SLN positivity rate was 3% for T1a-HR, 5% for T1b, 12% for T2a, and 13% for T2b.

Using the i31-GEP-SLNB model would reduce SLN biopsy rate by 69%, it had a sensitivity of 43%, specificity of 69%, positive predictive value (PPV) of 4%, and negative predictive value (NPV) of 98% for T1a-HR. For T1b, the SLN biopsy rate would be reduced by 41%, with a sensitivity of 83%, specificity of 42%, PPV of 8%, and NPV of 98%. In T2a and T2b, the SLN biopsy rate would decrease by 13% and 4%, the sensitivities were 96% and 100%, specificities were 14% and 5%, PPVs were 13% and 13%, and NPVs were 96% and 100%, respectively.

The greatest net benefit (+0.012) and relative utility (+22%) of the i31-GEP-SLNB was observed to T1b melanoma.

Per 100 patients, the i31-GEP-SLNB stratification approach would decrease ALN biopsy-related charges by $81,052 to $350,129.

This study may have been limited by setting the risk threshold at 5%.

These data indicated the i31-GEP-SLNB model was effective at stratifying patients with T1b melanoma into risk categories and would likely reduce unnecessary SLN biopsies and associated costs.

Disclosure: One author declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original article for a full list of disclosures.


Marchetti MA, Dusza SW, Bartlett EK. Utility of a model for predicting the risk of sentinel lymph node metastasis in patients with cutaneous melanoma.JAMA Dermatol. 2022;158(6):680-683. doi:10.1001/jamadermatol.2022.0970