【作者】 王乐三；

【导师】 孙振球；

【作者基本信息】 中南大学 ， 社会医学与卫生事业管理学， 2010， 博士

【摘要】 目的：评价PET-CT在NSCLC中的诊断准确度；寻找我省PET-CT在NSCLC诊断中的适宜诊断界值；校正灵敏度和特异度的证实偏倚；探索PET-CT诊断试验准确度的影响因素；分析NSCLC诊断的成本效果。方法：(1)对经“金标准”诊断的68例NSCLC和34例肺良性病变初诊患者采用18F-FDG PET-CT SUVmax法、CT目视法、PET-CT目视法、PET-CT目视法+SUVmax法4种诊断方法进行诊断学试验的准确度评价,比较各种诊断方法的灵敏度、特异度、正确率、阳性预测值、阴性预测值、约登指数、阳性似然比、阴性似然比和ROC曲线及曲线下的面积；(2)根据Youden指数最大原则、误诊率与漏诊率同等重要原则、正确率最大原则寻找我省18F-FDG PET-CT SUVmax鉴别非小细胞肺癌与肺良性病变的适宜诊断标准；(3)根据同期肺部良、恶性病变患者PET-CT检查阳性625例、PET-CT检查阴性332例的结果对灵敏度和特异度进行证实偏倚的校正；(4)采用二分类logistics回归和有序分类结果的累积logistics回归分析18F-FDG PET-CT诊断NSCLC与肺良性病变的准确度影响因素。(5)采用增量分析和敏感性分析比较18F-FDG PET-CT与CT的检出成本效果。结果：(1) 18F-FDG PET-CT SUVmax法、CT目视法、PET-CT目视法、PET-CT目视法联合SUVmax法4种诊断方法的灵敏度分别为91.2%、83.8%、89.7%、89.7%；特异度分别为55.9%、55.9%、55.9%、64.7%；Youden指数分别为0.471、0.397、0.456、0.544；正确率分别为79.4%、74.5%、78.4%、81.4%；阳性预测值分别为80.5%、79.2%、80.3%、83.6%；阴性预测值分别为76.0%、63.3%、73.1%、75.9%；阳性似然比分别为2.068、1.900、2.027、2.541；阴性似然比分别为0.157、0.290、0.184、0.159；经验ROC曲线下面积分别为0.7946、0.7814、0.8227、0.8722,双正态模型下光滑ROC曲线下面积分别为0.7973、0.8059、0.9055、0.9172,与机会线下面积0.5比较差异均有统计学意义,PET-CT联合SUVmax的诊断方法与SUVmax诊断方法和CT诊断方法之间的ROC曲线下面积有差异。(2) SUVmax在NSCLC与肺良性病变患者中分布呈偏态分布；NSCLC组与肺良性病变组的SUVmax均数±标准差分别为8.6±5.3和3.8±3.7,中位数(四分位数间距)分别为2.3(5.2)和7.2(6.1),两者差异有统计学意义(P<0.001)；鳞癌与腺癌SUVmax值的均数±标准差分别是9.5±4.5和7.2±6.,中位数(四分位数间距)分别为8.8(5.1)和5.5(6.3),两者差别无统计学意义(P=0.327)；18F-FDG PET-CT诊断NSCLC与肺良性病变的SUVmax与肿瘤大小相关系数rs=0.632(P＜0.001); 18F-FDG PET-CT诊断NSCLC与肺良性病变时,Youden指数最大原则下的适宜诊断界值为SUVmax=2.8;误诊率与漏诊率同等重要原则下的适宜诊断界值为SUVmax=5.45;正确率最大原则下的适宜诊断界值为SUVmax=2.8。(3)18F-FDG PET-CT极大似然估计法校正证实偏倚后的校正灵敏度、校正特异度分别为84.9%、15.1%。(4)经二分类logistic回归分析,肿瘤体积是SUVmax诊断方法中特异度的影响因素(P=0.037)；肿瘤体积是PET-CT诊断方法中灵敏度的影响因素(P=0.029)；肿瘤体积可能是PET-CT+SUVmax诊断方法中灵敏度的影响因素(P=0.057)。经有序结果的累积比数logistic回归分析,肿瘤大小是CT诊断法中ROC曲线的影响因素(通过影响灵敏度)；肿瘤大小是PET-CT诊断方法ROC曲线的影响因素(通过影响特异度)。(5) PET-CT和CT的检出成本／效果比分别为6242.62元／每例和322.11元／每例,成本／效果增量比为90610元／每例；PET-CT随灵敏度变化的成本／效果增量比为40271元／每例。结论：(1)18F-FDG PET-CT诊断NSCLC与肺良性病变的SUVmax与肿瘤大小呈正相关；NSCLC的SUVmax值高于肺良性病变的SUVmax值,鳞癌与腺癌SUVmax值的差异无统计学意义；SUVmax鉴别NSCLC与肺良性病变的适宜诊断标准为2.80；(2)18F-FDG PET-CT诊断NSCLC与肺良性病变的SUVmax法、CT目视分析法、PET-CT目视分析法、PET-CT目视分析联合SUVmax法的准确度比较,以半定量分析SUVmax法的灵敏度最高,以CT目视分析法的特异度最高,以目视分析结合半定量分析的PET-CT目视分析联合SUVmax法的准确度最高。(3)证实偏倚过高估计了18F-FDG PET-CT诊断非小细胞肺癌与肺良性病变的灵敏度和特异度。(4)肿瘤的大小和体积是18F-FDG PET-CT诊断非小细胞肺癌与肺良性病变准确度的影响因素,随着肿瘤的大小和体积的增加,灵敏度增大,误诊率增大。(5) PET-CT检出成本效果比和成本效果增量比高。更多还原

【Abstract】 Objective:To assess the diagnostic accuracy and find suitable critical value of maximum standardized uptake value(SUVmax)of 18F-FDG PET-CT for distinguishing non-small cell lung cancer (NSCLC) and pulmonary benign lesion. To correct the verification bias of sensitivity and specificity and explore the factors influenced diagnostic accuracy of it. To analyze the cost-effectiveness of PET-CT for the diagnosis ofNSCLC.Method:(1) Sixty-eight cases of NSCLC and thirty-four cases of pulmonary benign lesion examined by PET-CT were proved by surgery, biopsy, pathology or follow-up. All cases were diagnosed by SUVmax analysis, CT visual analysis, PET-CT visual analysis and PET-CT visual analysis plus SUVmax. The sensitivity, specificity, accuracy, positive predictive value, negative predictive value, Youden’s index, positive likelihood ratio and negative likelihood ratio, the area under ROC curve were calculated and compared.(2) Three methods,which based on making Youden’s index maximum, making false positive rate and false negative rate equal, making accuracy maximum, were used to find suitable critical value of SUVmax of 18F-FDG PET-CT for distinguishing NSCLC and pulmonary benign lesion. (3) The verification bias of the sensitivity and specificity were corrected by using maximum likelihood estimate. The cases not to be proved by "gold standard" included 625 cases of NSCLC and 332 cases of pulmonary benign lesion diagnosed by PET-CT at the same period.(4) To explore the factors of the diagnostic accuracy by using binary logistic regression and ordinal logistic regression.(5) Comparison of cost-effectiveness Of 18F-FDG PET-CT and CT in the diagnosis of NSCLC by incremental analysis and sensitivity analysisResults:(1)The sensitivity of the SUVmax analysis, CT visual analysis, PET-CT visual analysis and PET-CT visual analysis plus SUVmax were 91.2%,83.8%,89.7%,89.7%, respectively. The specificity of it were 55.9%,55.9%,55.9%,64.7%, respectively. The Youden’s index of it were 0.471,0.397,0.456,0.544, respectively. The accuracy of it were 79.4%, 74.5%,78.4%,81.4%, respectively. The positive predictive values of it were 80.5%,79.2%,80.3%,83.6%, respectively. The negative predictive values of it were 76.0%,63.3%,73.1%,75.9%, respectively. The positive likelihood ratio of it were 2.068,1.900,2.027,2.541, respectively. The negative likelihood ratio of it were 0.157,0.290,0.184,0.159, respectively. empirical area under the ROC curve of it were 0.7946,0.7814,0.8227,0. 8722, respectively. The area under the fitted smooth ROC curve, based upon binormal model were 0.7973,0.8059,0.9055,0.9172, respectively. The area of empirical and the Fitted smooth ROC curve is in statistical significant(P<0.05),compared with the area under the reference line. The difference of the area under ROC curve of the method of PET-CT visual analysis plus SUVmax was in statistical significant(P<0.05) compared with the methods of SUVmax analysis and CT visual analysis(2) The SUVmax distribution of NSCLC and pulmonary benign lesion was skewness distribution. The means±standard deviation of SUVmax of NSCLC and pulmonary benign lesion were 8.6±5.3,3.8±3.7 respectively.The median(quartilerange) was 2.3(5.2),7.2(6.1),respectively. The difference of SUVmax between NSCLC and pulmonary benign lesion was in statistical significant (P<0.001). The means±standard deviation of SUVmax of lung squamous cell carcinoma and lung adenocarcinoma were 9.5±4.5,7.2±6.2 respectively. The median(quartile range) were 8.8(5.1),5.5(6.3)respectively. The difference of SUVmax between lung squamous cell carcinoma and lung adenocarcinoma was not in statistical significant (P=0.327). The Spearman correlation coefficient between SUVmax and tumor size was 0.632(P<0.001). The optimal critical value of SUVmax was 2.8,5.5,2.8 respectively according to the rule of making Youden’s index maximum, making the false positive rate and false negative rate equal,making the accuracy maximum.(3) After correcting verification bias, the corrected sensitivity of the PET-CT visual analysis was 84.9%, the corrected specificity of it was 15.1%.(4)Tumor volumn was a factor affected the specificity of SUVmax analysis and also a factor affected the sensitivity of PET-CT visual analysis and PET-CT visual analysis plus SUVmax, using binary logistic regression analysis. The tumor size was a factor affected ROC curve of CT visual analysis (5-category scale:normal, benign, probably benign, suspicious, malignant) and PET-CT visual analysis (5-category scale: normal, benign, probably benign, suspicious, malignant).(5) PET-CT and CT diagnostic cost-effectiveness ratio were 6242.62RMB per cases and 322.11 RMB percases respectively. Incremental ratio of diagnostic cost-effectiveness was 90610 RMB per cases. Incremental ratio of diagnostic cost-effectiveness was 40271 RMB per cases with change of the sensitivity for the PET-CT.Conclusions:(1) There was a positive correlation between SUVmax and Tumor size. The SUVmax of NSCLC was higher than pulmonary benign lesion. The difference of SUVmax between lung squamous cell carcinoma and lung adenocarcinoma was not in statistical significant (P=0.327). The optimal critical value of SUVmax was 2.8.(2) The comparisons of the diagnostic accuracy among SUVmax analysis, CT visual analysis, PET-CT visual analysis and PET-CT visual analysis plus SUVmax showed that the sensitivity of SUVmax analysis (semi quantity) was the maximum, the specificity of CT visual analysis (visual analysis) was the maximum, the area under the ROC curve of PET-CT visual analysis plus SUVmax (visual analysis plus semi quantity)was the maximum.(3) The effect of verification bias overestimated sensitivity and specificity for diagnostic accuracy of 18F-FDG PET-CT in distinguishing NSCLC and pulmonary benign lesion.(4) The tumor size and volume were the factors influencing the diagnostic accuracy of 18F-FDG PET-CT in distinguishing NSCLC and pulmonary benign lesion.(5) The diagnostic cost-effectiveness ratio and incremental ratio of diagnostic cost-effectiveness were high for the PET-CT.更多还原