Abstract

Recent randomized clinical trials of stage III melanoma patients have not shown a disease-specific or metastasis-free survival benefit for complete lymph node dissection versus observation following a positive sentinel lymph node. This leads to a dilemma for prognostication and inclusion in clinical trials as the current AJCC melanoma staging system uses lymph node status (garnered from complete lymph node dissections) to subclassify patients into stage IIIa – IIId. Furthermore, there are conflicting studies whether patients with sentinel node micrometastases have the same survival rate as those with negative sentinel nodes. We present a model that uses the sentinel node metastasis size, extracapsular extension status, ulceration status, and Breslow depth, to stratify stage III patients into risk groups. We use AIC weights to calculate the probabilities of the models. Our model performs as well as the AJCC model in predicting disease-specific survival if data from complete lymph node dissections is available, and better than the AJCC model in predicting disease-specific survival if data from the complete lymph node dissection is not available (probability of AJCC model without lymph node dissection data = 0.016). Our model performs better than the AJCC model in predicting metastasis-free and recurrence-free survival regardless of the presence or absence of complete lymph node dissection data (probability of our model = .986 and .990, respectively). Controlling for Breslow depth and ulceration, we show that patients with sentinel node micrometastasis have a significantly higher locoregional recurrence or distant metastasis rate compared to sentinel node negative patients (p = 0.039).