大规模人群FDG PET／CT意外发现甲状腺高代谢结节的回顾性分析

Retrospective evaluation of focal hypermetabolic thyroid nodules incidentally identified by ^18F-FDGPET/CT in a large population

目的对大规模人群中^18F-FDGPET／CT意外发现的甲状腺高代谢结节的发病率和良恶性进行回顾性分析。方法回顾2007年8月至2010年3月间无甲状腺癌病史或甲状腺手术史行PET／CT检查者共计8463例，其中145例意外发现甲状腺高代谢结节，经病理检查和随访明确结节良恶性68例（恶性37例，良性31例）。患者中男21例，女47例，年龄21-76（53．66±10．85）岁。根据结节的良恶性，进行单因素和多因素分析，单因素分析采用t检验、r检验或Fisher确切概率法，多因素分析采用二分法logistic回归。分别以SUVmax和logistic回归模型绘制ROC曲线，比较两者AUC。结果甲状腺高代谢结节的发病率为1．71％（145／8463），恶性百分率为54．41％（37／68）。良性和恶性结节的SUVmax分别为5．134±4．02和7．614±4．78，差异有统计学意义（t=2．235，P=0．029）。Logistic回归分析结果显示有统计学意义的自变量包括SUVmax、有无钙化、是否多发、结节大小和患者年龄。Logistic回归模型的AUC（AUCI．）为0．878±0．043（95％CI：0．793-0．962，P〈0．05），SUVmax的AUC（AUCs）为0．694±0．067（95％CI：0．562-0．825，P〈0．05），两者差异有统计学意义（P〈0．05）。结论甲状腺高代谢结节有较高的恶性百分率，SUVmax对此类结节良恶性鉴别的意义有限；综合结节的PET／CT影像特征，logstic回归模型可以提高诊断准确性。

Objective To investigate the prevalence of focal hypermetabolic thyroid nodules inci- dentally detected by ^18F-FDG PET/CT in a relatively large population and explore its value in differentiating malignancy from benign thyroid nodules. Methods During August 2007 to March 2010, 8463 patients with no history of thyroid cancer or thyroidectomy underwent is F-FDG PET/CT. Among them, 145 patients were found to have abnormal hypermetabolic thyroid nodules. Sixty-eight patients were conformed with histopatho- logical or clinical follow-up, including 37 with malignancy and 31 with benign nodules （male 21, female 47, average age （53.66 ± 10.85 ）y）. The SUV nodule size, single or multiple nodules, with or without calcification and patient＇s age were chosen as the parameters for predicting malignancy in hypermetabolicthyroid nodules. Univariate analysis was performed using t test,X^2 test and Fisher exact test. Binary logistic regression was performed for multi-variate analysis. The AUCs of SUVmax and logistic regression analysis were compared. Results The incidence of focal hypermetabolic thyroid nodules was 1.71% （145/8463）, with malignancy rate 54.41% （37/68）. The SUVmax of benign and malignant nodules were 5.13±4.02 and 7.61± d. 78, respectively （t = 2.235, P = 0. 029）. Logistic regression indicated that SUVmax, with or with- out calcification, single or multiple nodules, nodule size and patient＇s age were all the predictors for malig- nancy in hypermetabolie thyroid nodules. The AUC of logistic regressive model （AUCL ） and SUVmax （AUCs ） were 0.878 ＋0.043 （95% CI： 0.793-0.962, P 〈0.05） and 0.694 ±0.067 （95% CI： 0.562-0.825. P 〈0.05）, respectively （P 〈 0.05 ）. Conclusions Focal hypermetabolic thyroid nodules incidentally identi- fied by ^18F-FDG PET/CT come with high rate of thyroid malignancy. Differential diagnosis could be im- proved significantly using SUVmax and logistic regressive model aided by other parameters from 1SF-FDG PET/CT as well as patient＇s age.