探讨甲状腺乳头状癌（PTC）术前细针穿刺洗脱液甲状腺球蛋白（FNA-Tg）诊断颈部淋巴结转移的阈值及效能。

回顾性分析2016年7月至2018年1月于浙江大学附属邵逸夫医院头颈外科行甲状腺结节手术的410例患者共541个淋巴结的FNA-Tg结果。以术后淋巴结病理结果为“金标准”，绘制ROC曲线得出FNA-Tg对颈部淋巴结转移的最佳诊断阈值及效能分析；根据血清甲状腺球蛋白（sTg）将淋巴结分为低sTg组（sTg≤10 ng/ml）和高sTg组（sTg＞10 ng/ml），绘制ROC曲线分析2组的FNA-Tg最佳诊断阈值；并对Tg比值法（FNA-Tg/sTg）和Tg差值法（FNA-Tg-sTg）的诊断阈值和效能进行分析。

绘制FNA-Tg诊断淋巴结转移与非转移的ROC曲线，得出FNA-Tg的最佳诊断阈值为0.835 ng/ml，取最佳截断值时，其敏感度、特异度分别为88.5%、97.5%；ROC曲线下面积为0.944（95%CI：0.914~0.974）。FNA-Tg水平与sTg水平相关（r=0.350，P＜0.001），根据sTg水平将淋巴结分为低sTg组296个和高sTg组245个，2组FNA-Tg的最佳诊断阈值分别为0.825 ng/ml（ROC曲线下面积为0.909；95%CI：0.838~0.981）、0.875 ng/ml（ROC曲线下面积为0.951；95%CI：0.918~0.984）。Tg比值法0.145为最佳诊断阈值（ROC曲线下面积为0.922；95%CI：0.889~0.955），其敏感度、特异度、阳性预测值、阴性预测值分别为82.7%、95.5%、85.3%、94.3%；Tg差值法＞0（ROC曲线下面积为0.841，95%CI：0.802~0.879）为诊断阈值时的敏感度、特异度、阳性预测值、阴性预测值分别为74.8%、93.3%、79.4%、91.5%。

FNA-Tg诊断术前PTC患者颈部淋巴结转移可采用不同的诊断阈值，当FNA-Tg为0.835 ng/ml，诊断性能较好；为避免sTg水平可能带来的影响，可综合考虑Tg比值法及Tg差值法。

To explore the threshold and efficacy of thyroglobulin assessment in the needle washout after fine needle aspiration (FNA-Tg) in the preoperative diagnosis of cervical lymph metastasis of thyroid papillary carcinoma (PTC).

A retrospective analysis was performed on the FNA-Tg results of a total of 541 lymph nodes in 410 patients who underwent thyroid node surgery at the Department of Head Neck Surgery, Sir Run Run Shaw Hospital of Zhejiang University College of Medicine from July 2016 to January 2018. Taking the postoperative lymph node pathological results as the "gold standard", the ROC curve was plotted to obtain the optimal diagnostic threshold of FNA-Tg for cervical lymph node metastasis and evaluate its diagnostic efficacy. According to the level of serum thyroglobulin (sTg), the lymph nodes were divided into either a low sTg group (sTg ≤10 ng/ml) or a high sTg group (sTg >10 ng/ml), the ROC curve was plotted to analyze the diagnostic threshold of FNA-Tg in the two groups. The diagnostic threshold and efficacy of FNA-Tg/sTg ratio and FNA-Tg-sTg difference were then analyzed.

The ROC curve of FNA-Tg for diagnosis of lymph node metastasis was plotted, and the optimal diagnostic threshold of FNA-Tg was determined to be 0.835 ng/ml. Using this threshold as the cut-off point, the sensitivity and specificity were 88.5% and 97.5%, respectively, and the area under the ROC curve (AUC) was 0.944 (95%CI: 0.914~0.974). The level of FNA-Tg was correlated with the level of sTg (r=0.350, P＜0.001). Then, the lymph nodes were divided into either a low sTg group (n = 296) or a high sTg group (n = 245) according to the level of sTg, and the optimal diagnostic threshold for the two groups was determined to be 0.825 ng/ml (AUC = 0.909, 95%CI: 0.838~0.981) and 0.875 ng/ml (AUC = 0.951, 95%CI: 0.918~0.984), respectively. The optimal diagnostic cut-off of FNA-Tg/sTg ratio was determined to be 0.145; using this cut-off, the AUC of FNA-Tg/sTg ratio was 0.922 (95%CI: 0.889~0.955), and the sensitivity, specificity, positive predictive value, and negative predictive value were 82.7%, 95.5%, 85.3%, and 94.3%, respectively. The optimal diagnostic cut-off of FNA-Tg-sTg difference was determined to be >0; using this cut-off, the AUC of FNA-Tg-sTg difference was 0.841 (95%CI: 0.802~0.879), and the sensitivity, specificity, positive predictive value, and negative predictive value were 74.8%, 93.3%, 79.4%, and 91.5%, respectively.

FNA-Tg can be used for the preoperative diagnosis of cervical lymph node metastasis in patients with papillary thyroid carcinoma. When 0.835 ng/ml is used as the cut-off value of FNA-Tg, the diagnostic performance is better. To avoid the possible impact of sTg level, FNA-Tg/sTg ratio and FNA-Tg-sTg difference could be used in combination.