| ObjectivePrecise staging in non-small cell lung cancer (NSCLC) patients is important for determining choice of treatment and prognosis,and mediastinal lymph node (MLN) staging before surgery is essential but difficult.Integrated 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) is widely used for mediastinal lymph node staging in patients with NSCLC for its high sensitivity and specifity.However,many studies demonstrate that the quality of staging varies with different interpretion methods.The aims of the present study was :(1)to evaluate the diagnostic performance of visual interpretation of 18F-FDG PET/CT in diagnosing MLN metastasis before operation for patients with NSCLC;(2)To compare the preoperative diagnosing efficiency of MLN metastasis between SUVmax and SUVmean with the receiver operator characteristic(ROC) analysis;furthermore, when taking the optimal cut-off values, we evaluated the diagnostic accuracy of SUVmax and visual interpretation.Meanwhile, the capability for different size of mediastinal nodes was assessed.(3) To determin the risk factors for mediastinal lymph node metastasis were using the univariate and multivariate analysis.Methods213 consecutive patients (134 men, 79 women; mean age at diagnosis, 60±11 years; range, 18–84 years) with NSCLC from January 2006 to March 2011 underwent 18F-FDG PET/CT before curative surgical resection (tumor resection and lymph node dissection).Two experienced physicians retrospectively assessed all examination results. Final diagnosis based on each of the methods was made by consensus of two readers.Nodes showing greater 18F-FDG uptake than mediastinum at PET without calcification or high attenuation >70 household unit (Hu) at unenhanced CT were regarded as being positive for malignancy. ROC analysis was used to compare the capability for mediastinal nodes assessment among SUVmax and SUVmean on a per-nodal station basis. The histologic nodal assessment results were used as reference standards. Sensitivity, specificity, and accuracy were compared with the McNemar test; positive predictive value(PPV) and negative predictive value(NPV) were compared with the Pearsonχ2 test. Risk factors for mediastinal lymph node metastasis were defined by univariate and multivariate analysis. A probability value of less than 0.05 was considered statistically significant.Results(1) Of 705 mediastinal nodal stations evaluated in 213 patients, 139 (19.7%) stations in 67 (31.5%) patients proved to be malignant.For the depiction of malignant nodes, the sensitivity, specificity, accuracy,positive predictive value(PPV) and negative predictive value(NPV) using visual interpretation of 18F-FDG PET/CT were 71.9% (100 of 139 nodal groups), 90.1% (510 of 566),86.5%(610 of 705),64.1%(100 of 156) and 92.9% (510 of 549), respectively, whereas those of CT were 57.6% (80 of 139), 85.0% (481 of 566),79.6%(561 of 705),48.5%(80 of 165)and 89.1% (481 of 540) (P=0.000, P=0.002, P=0.000, P=0.005 and P=0.028, respectively). On a per-person basis, the overall sensitivity, specificity, accuracy,PPV and NPV of 18F-FDG PET/CT for mediastinal nodal staging were 82.1% (55 of 67), 80.1% (117 of 146), 80.8%(172 of 213),65.5%(55 of 84)and 90.7% (117 of 129), respectively, whereas those of CT were 70.1% (47 of 67), 69.9% (102 of 146),70.0%(149 of 213),51.6%(47 of 91)and 83.6% (102 of 122) (P=0.021, P=0.017, P=0.001, P=0.064 and P=0.000, respectively).(2) Area under ROC curve (AUC) values of SUVmax (AUC=0.846, P=0.0001) was larger than that of SUVmean (AUC=0.839,P=0.0001), between these two quantitative variables was no significant difference(P=0.299). The optimal cutoff value was 2.3 and 1.8 for SUVmax and SUVmean,respectively.When taking this cutoff value, the sensitivity, specificity, accuracy,PPV and NPV of SUVmax for the depiction of malignant nodes were 71.2%(99 of 139),85.2%(482 of 566),82.4%(581 of 705),54.1%(99 of 183) and 92.3% (482 of 522) respectively,lower than the visual interpretation.But statistical significancy was got only in specificity and accuracy(P=0.000,P=0.000, respectively).The sensitivity, specificity, accuracy,PPV and NPV of SUVmax for mediastinal nodal staging were 79.1%(53/67),71.2%(104/146),73.7%(157/213),55.8%(53/95) and 88.1% (102/122),also lower than the visual interpretation. Much to our regret,we did not get a statistical significancy except for specificity and accuracy(P=0.000,P=0.000, respectively).(3) Using ROC curve analysis, we determined the AUC values of SUVmax was 0.893 (P=0.0001),and the optimal cut-off value was 3.4 for enlarged nodes with the short-axis diameter≥10mm. The sensitivity, specificity, accuracy,PPV and NPV of SUVmax were 82.5%(66 of 80),85.9%(73 of 85),84.2%(139 of 165),84.6%(66 of 78) and 83.9%(73 of 87),while these indicies were 95.0%(66 of 80),65.9%(73 of 85),80.0%(139 of 165),72.4%(66 of 78) and 93.3%(73 of 87) of visual diagnosis.For sensitivity,visual diagnosis was superior to SUVmax (P=0.002),as for specificity,the latter was better(P=0.000).There was no statistical significance among accuray,PPV,and NPV.For normal size nodes(short-axis diameter <10mm),the AUC of SUVmax was 0.722(P=0.0001),with the optimal cut-off value of 2.1. The sensitivity, specificity, accuracy, and PPV of visual diagnosis were 94.4%,88.5%,47.1%,84.4%,80.6%,27.9% of SUVmax(P=0.000,P=0.000,P=0.018).However,as for sensitivity and NPV,significance were not obviously seen.(4) Defined by univariate analysis ,the differentiation of tumor((P=0.006),SUVmax of the primary leision(P=0.024),visual dignosis of PET/CT for positive N1(P=0.000) and enlarged mediastinal nodes on CT(P=0.000) were probable predictors for mediastinal lymph node metastasis. Multivariate analysis demonstrated visual dignosis of PET/CT for positive N1(P=0.002) and enlarged mediastinal nodes on CT(P=0.000) had statistical significance. Conclusions(1) Visual interpretation of PET/CT in diagnosing metastasis of mediastinal nodes was superior to located CT with a high specifity,NPV and a relative high accuracy,while the sensitivity and PPV were not so valuable.Visual interpretation had better specifity and accuray than SUVmax,with equal sensitivity,PPV and NPV between the two methods.In a word, SUVmax can be replaced by visual diagnosis in mediastinal nodes staging.(2) SUVmax and SUVmean had the same diagnostic capability with an optimal cut-off value of 2.3 and 1.8.(3) For enlarged lymph nodes, 3.4 was thought to be the optimal value of SUVmax,with a higher specifity and lower sensitivity compared with visual interpretation.For those nodes with normal size, the best cutoff value should be defined as 2.1.The diagnostic performance of visual diagnosis was superior to SUVmax,but the sensitivity and PPV was low.(4) Univariate analysis demonstrated that,SUVmax of the primary leision,visual dignosis of PET/CT for positive N1 and enlarged mediastinal nodes on CT were high-risk factors for mediastinal lymph node metastasis,while the differentiation of tumor was considered to be a protective factor.Logistic multivariate analysis showed that visual dignosis of PET/CT for positive N1 and enlarged mediastinal nodes on CT were independent predictive factor. |
PDF Full Text Request