He Yushuang, Yang Yujia, Wen Wen, Qiu Li, Li Zhihui, Lei Jianyong
Department of Ultrasound, West China Hospital of Sichuan University, Chengdu, China.
Department of Thyroid Surgery, West China Hospital of Sichuan University, Chengdu, China.
Quant Imaging Med Surg. 2023 Oct 1;13(10):7002-7011. doi: 10.21037/qims-23-332. Epub 2023 Sep 11.
For thyroid cancer staging, evaluation of extent of local invasion, including recurrent laryngeal nerve (RLN), may assist surgical decision-making.
This prospective study evaluated patients who underwent thyroidectomy at a single tertiary-level academic institution. Patients with complete clinical information and ultrasound imaging of thyroid carcinoma and RLN were enrolled. Those who had thyroidectomy before or did not fit the above conditions were excluded. Patients were assigned to either a development or validation cohort. Development of models was constructed in a primary cohort based on preoperative ultrasound features and clinicodemographic data from August 2020 to December 2021. Validation of the models was then performed on an independent cohort between January and March of 2022. Multivariate logistic regression and nomograms were mainly used for statistical analysis.
Using data from 816 patients (80 RLN invasion), we built nomogram 1 based on age [95% confidence interval (CI): 1.315 to 145.933, P=0.029], body mass index (BMI; 95% CI: 1.228 to 10.874, P=0.020), tumor size (95% CI: 4.677 to 1,373.1, P=0.002), tumor adjacent to medial (95% CI: 1.816 to 26.713, P=0.005) and posterior thyroid capsules (95% CI: 5.567 to 756.583, P=0.001), and distance <1 mm between tumor and the RLN (95% CI: 10.389 to 826.746, P<0.001). Nomogram 2 was built based on tumor adjacent to the posterior thyroid capsule (95% CI: 2.922 to 53,074.51, P=0.017), distance <1 mm between tumor and the RLN (95% CI: 1.478 to 1,241.646, P=0.029), and loss of typical fascicular echotexture of the RLN along the long axis (95% CI: 35.11 to 53,272.81, P<0.001). In the validation cohort, nomogram 1 and 2 showed sensitivities of 94.74% and 57.89%, specificities of 74.12% and 95.29%, positive predictive values (PPV) of 45.00% and 73.26%, negative predictive values (NPV) of 98.43% and 91.03%, accuracies of 76.92% and 88.46%, and C-indices of 0.86 and 0.89.
Preoperative ultrasound is a feasible approach to evaluate RLN invasion in patients with thyroid carcinoma. Nomogram 1 may sensitively identify the risk of RLN invasion, and it may be checked using the more specific and accurate nomogram 2 to reduce false positives.
对于甲状腺癌分期，评估局部侵犯范围，包括喉返神经（RLN），可能有助于手术决策。
这项前瞻性研究评估了在一家三级学术机构接受甲状腺切除术的患者。纳入具有完整临床信息以及甲状腺癌和喉返神经超声成像的患者。排除之前接受过甲状腺切除术或不符合上述条件的患者。将患者分配至开发队列或验证队列。基于2020年8月至2021年12月的术前超声特征和临床人口统计学数据，在一个主要队列中构建模型。然后在2022年1月至3月期间对一个独立队列进行模型验证。主要使用多变量逻辑回归和列线图进行统计分析。
利用816例患者的数据（80例喉返神经侵犯），我们基于年龄[95%置信区间（CI）：1.315至145.933，P = 0.029]、体重指数（BMI；95%CI：1.228至10.874，P = 0.020）、肿瘤大小（95%CI：4.677至1373.1，P = 0.002）、肿瘤与甲状腺内侧相邻（95%CI：1.816至26.713，P = 0.005）以及肿瘤与甲状腺后包膜相邻（95%CI：5.567至756.583，P = 0.001），以及肿瘤与喉返神经之间距离<1mm（95%CI：10.389至826.746，P<0.001）构建列线图1。基于肿瘤与甲状腺后包膜相邻（95%CI：2.922至53074.51，P = 0.017）、肿瘤与喉返神经之间距离<1mm（95%CI：1.478至1241.
