Melanoma: Staging/Prognosticating

PREDICTING SUBCLINICAL SPREAD

Clinical Risk Factors of Subclinical Spread in Melanoma in Situ and Invasive Melanoma. Dermatol Surg. 2025. PMID: 39412145.

• Objective: identify clinical risk factors associated with subclinical spread (defined as Tumors requiring more than 1 Mohs stage to clear margins) in patients with melanoma in situ (MIS) and invasive melanoma (IM) treated with Mohs micrographic surgery (MMS).

• Study Design: Retrospective cohort study, 327 patients with MIS (80.7%) or IM treated with MMS by one surgeon from 2020–2024

• Key Findings: Subclinical spread occurred in 44.6% of cases (146/327)

  • Significant risk factors for subclinical spread included: History of melanoma (p = .013), Larger prebiopsy tumor size (p = .034), Larger preoperative size (p = .011), Recurrent tumors (p = .019), Cheek location among facial tumors (p = .039)

  • No significant associations were found with: Age, Tumor type (MIS vs IM), special anatomic site (e.g., “H zones” of face)

• My take: I don't find this particularly clinically helpful​

SENTINEL LYMPH NODE BIOPSY

Predicting SLN Positivity

• The Melanoma Institute of Australia created a nomogram based on clinical factors to predict patients who are likely to have a positive sentinel node (here), and validated it in a cohort of ~60K US patients.  It showed an ROC C-statistic of 0.733, indicating a good, but not strong predictive model. 37454700​

Risk Prediction Models for Sentinel Node Positivity in Melanoma: A Systematic Review and Meta-Analysis. JAMA Dermatol. 2025 May. PMID: 40072444.​​

• Most Validated Models: MSKCC (Memorial Sloan Kettering) and MIA (Melanoma Institute Australia)

• Variables Commonly Used: Breslow depth, ulceration, patient age, Clark level, mitotic rate, and tumor location.

• Gene Expression Profiles (GEP): No significant difference in discrimination between models that included GEP and those that did not (P = .11).

• Clinical Utility: Nine models provided sufficient information for individualized risk estimates and preprocedural use. Some offer online calculators or integer-based risk scores.

• MSKCC and MIA models are currently the most robust and applicable in practice but differ in cohort inclusion criteria (e.g., Clark level thresholds).

• Genetic testing models are not clearly superior and may not be cost-effective or widely available.

• When using nomograms, consider alignment between your patient population and the original development cohort.

  • Models and there best use case/notes:

  • MSKCC - Intermediate/thick melanomas (≥1 mm); U.S.-based; wide validation.  Most validated; excludes thin unless Clark IV/V; slightly older cohort.

  • MIA - Broad Breslow range (includes <1 mm); Australian; excludes rare subtypes. Highly validated; flexible input; online calculator available

  • Tripathi - Thin melanomas; large, diverse SEER/NCDB data; C-stat ~0.85; designed for U.S. population with thin melanomas

  • Maurichi - Thin melanomas; high C-statistic but limited external validation; Outstanding reported performance (C-stat 0.96), but methodological concerns raised

  • Friedman - Thin melanomas (0.5–1.0 mm); U.S. NCDB data; Targeted to those right on the SLNB threshold

Global Applicability of a Risk Prediction Tool for Sentinel Node Positivity in Patients With Primary Cutaneous Melanoma. JAMA Dermatol. 2025. PMID: 40202725.

• Concise study summary: This retrospective, multicentre cohort study pooled data from 15,731 adults who underwent sentinel-node biopsy (SNB) for primary cutaneous melanoma at 11 centres across Europe, North America, South America and Australasia to externally validate the Melanoma Institute Australia (MIA) six-parameter SN-metastasis risk calculator. The full model (all six inputs) achieved an AUC of 73 % (95 % CI 70.6-75.3 %) with excellent calibration; performance dipped slightly (≈70-71 %) when 1-3 optional variables were missing, and decision-curve analysis showed net clinical benefit over “biopsy all” once the predicted risk exceeded 8% (if your biopsy threshold is <8%, this model isn't very helpful). Key limitations missing optional pathology data in >80 % of records, inclusion only of patients already selected for SNB (selection bias), inter-centre heterogeneity, and wide CIs for rare subtypes such as pure desmoplastic melanoma..

• Implications for today’s clinical practice: great if you threshold for biopsy is >8%, in which case using this tool will help refine who should be referred for biopsy. ​

• Plain-language takeaway for patients: a model like this will tell you "out of 100 patients, how many will have a positive sentinel node".   It does this pretty well, so that if the calculator says 8%, pretty close to 8 out of 100 with those same parameters will have a positive node.  However, when asked which 8, it will be right only 73% of the time -- 27 of the 100 patients will be given the wrong prediction. 

Risks and benefits of SNLBx

Sentinel lymph node biopsy status improves adjuvant therapy decision-making in patients with clinical stage IIB/C melanoma: A population-based analysis. J Am Acad Dermatol. 2023. PMID: 36442639.

MY SUMMARY: This is a challenging study to comprehend statistically, but conceptually it sought to weigh the risks and benefits of SLNBx for patients with stage IIB/C melanoma for whom adjuvant immunotherapy is FDA approved.  The challenge with answering the question is the variability of recommendations by oncologists for adjuvant immunotherapy, which often hinges on the risk of melanoma specific death.  The conclusion was that if sentinel lymph node biopsy is completed and used as a prognostic variable in contrast to using clinicopathologic factors alone to estimate melanoma specific death, treatment recommendations will be more accurate.  Numerically: 6 of 100 patients will avoid immunotherapy without missing the treatment of those destined to recur.  When considering patients <65yo, this was 11 of 100 patients.​​

​

31-GENE EXPRESSION PROFILE

The 31-Gene Expression Profile Test Outperforms AJCC in Stratifying Risk of Recurrence in Patients with Stage I Cutaneous Melanoma. Cancers (Basel). 2024. PMID: 38254778.

MY SUMMARY: Cohort of AJCC Stage IA and IB tumors.  Found 31-GEP better predicts recurrence and melanoma specific survival than AJCC IA or IB classification.  All classes of GEP (1A, 1B/2A, and 2B) all predicted recurrence.  Only 2B was an independent predictor of survival.