National Comprehensive Cancer Network guidelines state that clinical stage III melanoma patients may undergo ultrasound surveillance of the nodal basin in lieu of complete lymph node dissection (CLND). This has led to an inability to accurately classify patients according to the American Joint Committee on Cancer (AJCC) eighth edition staging system because it uses the total number of positive lymph nodes from the CLND to assign a pathologic N stage. We propose a new model for clinical stage III melanoma patients that does not rely on the total number of positive lymph nodes. Instead, it uses Breslow depth, ulceration status, sentinel lymph node metastasis size, and extracapsular extension to stratify patients into groups 1 to 4. We compared our model's ability to predict melanoma-specific survival (MSS), distant metastasis-free survival (DMFS) and locoregional recurrence, and distant metastasis-free survival (DMFS-LRFS) to the current AJCC system with and without CLND-data using a Cox proportional hazards model and Akaike Information Criteria weights. Although not reaching our predetermined level of statistical significance of 95%, our model was 5 times more likely to better predict MSS compared with the AJCC model with CLND. In addition, our model was significantly better than the AJCC model without CLND in predicting MSS. Our model performed significantly better than the AJCC model in predicting DMFS and DMFS-LRFS regardless of whether data from CLND were included.

中文翻译：

---

III 期黑色素瘤：优于美国癌症联合委员会第八版分期系统的拟议分期模型。


国家综合癌症网络指南指出，临床 III 期黑色素瘤患者可以接受淋巴结盆腔超声监测，以代替完全淋巴结清扫术 (CLND)。这导致无法根据美国癌症联合委员会 (AJCC) 第八版分期系统对患者进行准确分类，因为它使用 CLND 的阳性淋巴结总数来分配病理 N 分期。我们为临床 III 期黑色素瘤患者提出了一种不依赖于阳性淋巴结总数的新模型。相反，它使用 Breslow 深度、溃疡状态、前哨淋巴结转移大小和囊外扩展将患者分为 1 至 4 组。我们比较了我们的模型预测黑色素瘤特异性生存 (MSS)、远处无转移生存 (DMFS) 的能力）和局部区域复发，以及使用 Cox 比例风险模型和 Akaike 信息标准权重的当前 AJCC 系统（有或没有 CLND 数据）和无远处转移生存 (DMFS-LRFS)。尽管没有达到我们预定的 95% 统计显着性水平，但与带有 CLND 的 AJCC 模型相比，我们的模型更好地预测 MSS 的可能性高出 5 倍。此外，我们的模型在预测 MSS 方面明显优于没有 CLND 的 AJCC 模型。无论是否包含来自 CLND 的数据，我们的模型在预测 DMFS 和 DMFS-LRFS 方面均明显优于 AJCC 模型。