摘要
目的探究含正电子发射计算机断层显像/X线计算机体层成像仪(PET-CT)及临床数据的联合模型预测甲状腺癌术后^(131)I治疗效果的临床价值。方法前瞻性选取2022年4月至2024年4月南阳市中心医院收治的298例甲状腺癌患者作为研究对象,所有患者均行甲状腺全切术及^(131)I治疗。术后6个月,失访8例。根据^(131)I治疗6个月后治疗效果分为成功组(n=235)和失败组(n=55)。比较两组患者的临床资料、游离碘甲状腺原氨酸(FT3)、促甲状腺激素(TSH)、刺激性甲状腺球蛋白(sTg)、游离甲状腺激素(FT4)、甲状腺球蛋白(Tg)减少指数和PET-CT参数[最大标准摄取值(SUVmax)、平均标准摄取值(SUVmean)、代谢性肿瘤体积(MTV)],采用Spearman相关系数分析各指标间的相关性,并将数据集分为训练集203例和验证集87例,采用Lasso交叉验证算法筛选指标,建立临床数据、PET-CT参数的联合模型,绘制受试者工作特征(ROC)曲线分析其对甲状腺癌术后^(131)I治疗效果的预测价值,并在验证集中进行验证。结果失败组患者的sTg为(5.53±1.54)ng/mL,明显高于成功组的(3.96±1.41)ng/mL,TSH、Tg减少指数及甲状腺静态显像阳性、^(131)I首次治疗剂量>100 mCi所占比例分别为(36.62±3.31)mIU/L、(40.68±11.75)%、30.91%、40.00%,明显低于成功组的(46.68±4.52)mIU/L、(53.22±12.59)%、80.00%、81.70%,差异均有统计学意义(P<0.05);Spearman相关系数分析结果显示,SUV_(max)、SUV_(mean)、MTV与sTg呈正相关,与Tg减少指数、TSH、甲状腺静态显像、^(131)I首次治疗剂量呈负相关(P<0.05);经过Lasso交叉验证算法筛选确定SUVmax、SUVmean、MTV、s Tg、Tg减少指数5个特征参数入选,回归系数分别为0.602、0.511、0.432、0.368、-0.211,将此模型命名为M5;M5模型在训练集和验证集中AUC为0.921、0.924,经Delong检验得出差异无统计学意义(P>0.05)。结论基于PET-CT及临床数据的联合模型对甲状腺癌术后^(131)I治疗效果具有预测价值,可为临床诊治及疾病转归提供科学依据。
Objective To explore the clinical value of a combined model incorporating positron emission tomography-computed tomography(PET-CT)and clinical data for predicting the therapeutic effect of ^(131)I after thyroid cancer surgery.Methods This prospective study enrolled 298 thyroid cancer patients who underwent total thyroidectomy followed by ^(131)I therapy at Nanyang Central Hospital between April 2022 to April 2024.At 6 months post-operation,eight patients were lost to follow-up.According to the therapeutic response to ^(131)I at 6 months,the remaining patients were divided into a success group(n=235)and a failure group(n=55).Clinical data,free triiodothyronine(FT3),thyroid-stimulating hormone(TSH),stimulated thyroglobulin(sTg),free thyroxine(FT4),thyroglobulin(Tg)reduction index,and PET-CT parameters(maximum standardized uptake value[SUVmax],mean standardized uptake value[SUVmean],metabolic tumor volume[MTV])were compared between the two groups.Spearman correlation coefficients were used to analyze inter-indicator correlations.The dataset was split into a training set(203 cases)and a validation set(87 cases).A least-absolute-shrinkage-and-selection-operator(Lasso)cross-validation algorithm was employed for feature selection to develop a combined model incorporating clinical data and PET-CT parameters.Receiver operating characteristic(ROC)curves were plotted to assess the predictive value for ^(131)I treatment outcomes,with subsequent validation in the validation set.Results The failure group exhibited significantly higher sTg levels compared to the success group:(5.53±1.54)ng/mL vs(3.96±1.41)ng/mL,P<0.05.Values for TSH,Tg reduction index,and the proportion of positive thyroid static imaging and rate of initial ^(131)I dose>100 mCi were(36.62±3.31)mIU/L,(40.68±11.75)%,30.91%,and 40.00%,respectively,significantly lower than the corresponding(46.68±4.52)mIU/L,(53.22±12.59)%,80.00%,and 81.70%in the success group(all P<0.05).Spearman correlation analysis revealed SUV_(max),SUV_(mean),and MTV were positively correlated with sTg,but negatively correlated with Tg reduction index,TSH,thyroid static imaging positivity,and initial ^(131)I dose(all P<0.05).Lasso cross-validation identified five predictive parameters:SUVmax,SUVmean,MTV,sTg,and Tg reduction index,with regression coefficients of 0.602,0.511,0.432,0.368,and-0.211,respectively,comprising the"Model M5".The Model M5 demonstrated AUC values of 0.921(training set)and 0.924(validation set),with no statistically significant difference in predictive performance by Delong test(P>0.05).Conclusion The combined model based on PET-CT and clinical data shows high predictive value for the therapeutic effect of ^(131)I therapy in postoperative thyroid cancer surgery,providing a scientific basis for clinical decision-making and prognostic evaluation.
作者
王素
朱丽
姚暖
尚毓
WANG Su;ZHU Li;YAO Nuan;SHANG Yu(Department of Nuclear Medicine,Nanyang Central Hospital,Nanyang 473000,Henan,CHINA.)
出处
《海南医学》
2025年第11期1556-1562,共7页
Hainan Medical Journal
基金
2023年度河南省医学科技攻关计划联合共建项目(编号:LHGJ20233973)。
