Invasive Pulmonary Adenocarcinomas versus Preinvasive Lesions Appearing as Ground-Glass Nodules: Differentiation by Using CT Features 1

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights. Sang Min Lee, MD Chang Min Park, MD Jin Mo Goo, MD Hyun-Ju Lee, MD Jae Yeon Wi, MD Chang Hyun Kang, MD Invasive Pulmonary Adenocarcinomas versus Preinvasive Lesions Appearing as Ground-Glass Nodules: Differentiation by Using CT Features 1 Purpose: Materials and Methods: To retrospectively investigate the differentiating computed tomographic (CT) features between invasive pulmonary adenocarcinoma (IPA) and preinvasive lesions appearing as ground-glass nodules (GGNs) in 253 patients. This study was approved by the institutional review board. From January 2005 to October 2011, 272 GGNs were pathologically confirmed (179 IPAs and 93 preinvasive lesions) in 253 patients and were included in this study. There were 64 pure GGNs and 208 part-solid GGNs. Preinvasive lesions consisted of 21 atypical adenomatous hyperplasias and 72 adenocarcinomas in situ. To identify the differentiating CT features between IPAs and preinvasive lesions and to evaluate their differentiating accuracy, logistic regression analysis and receiver operating characteristic (ROC) curve analysis were performed, respectively. Original Research n Thoracic Imaging 1 From the Departments of Radiology (S.M.L., C.M.P., J.M.G., H.J.L., J.Y.W.) and Thoracic and Cardiovascular Surgery (C.H.K.), Seoul National University College of Medicine, and Institute of Radiation Medicine, Seoul National University Medical Research Center, 101 Daehangno, Jongno-gu, Seoul 110-744, Korea. Received May 7, 2012; revision requested June 27; revision received October 9; accepted December 14; final version accepted December 28. Supported by a research grant from the Korean Foundation for Cancer Research (grant CB-2011-02-01). Address correspondence to C.M.P. (e-mail: cmpark@ radiol.snu.ac.kr). q RSNA, 2013 Results: Conclusion: In pure GGNs, preinvasive lesions were significantly smaller and more frequently nonlobulated than IPAs (P,.05). Multivariate analysis revealed that lesion size was the single significant differentiator of preinvasive lesions from IPAs (P =.029). The optimal cut-off size for preinvasive lesions was less than 10 mm (sensitivity, 53.33%; specificity, 100%). In part-solid GGNs, there were significant differences in lesion size, solid portion size, solid proportion, margin, border, and pleural retraction between IPAs and preinvasive lesions (P,.05). Multivariate analysis revealed that smaller lesion size, smaller solid proportion, nonlobulated border, and nonspiculated margin were significant differentiators of preinvasive lesions (P,.05), with excellent differentiating accuracy (area under ROC curve, 0.905). In pure GGNs, a lesion size of less than 10 mm can be a very specific discriminator of preinvasive lesions from IPAs. In part-solid GGNs, preinvasive lesions can be accurately distinguished from IPAs by the smaller lesion size, smaller solid proportion, nonlobulated border, and nonspiculated margin. q RSNA, 2013 Radiology: Volume 268: Number 1 July 2013 n radiology.rsna.org 265

Recently, the International Association for the Study of Lung Cancer, the American Thoracic Society, and the European Respiratory Society introduced a new classification of lung adenocarcinoma (1). In this classification, adenocarcinoma in situ (AIS) defined as a small ( 3 cm) adenocarcinoma with growth restricted to neoplastic cells along preexisting alveolar structures without stromal, vascular, or pleural invasion (1) is newly recognized as another preinvasive lesion for pulmonary adenocarcinoma in addition to atypical adenomatous hyperplasia (AAH), replacing the former term of bronchioloaveolar carcinoma (BAC). This conceptual change from BAC to AIS was due to the observation that BACs showed 100% disease-free survival with proper surgical intervention (2 5). The introduction of AIS implies the possibility of a new surgical approach or sublobar resection for these lesions (6). Several studies (3,5,7 9) have shown that sublobar resection could be Advances in Knowledge nn In pure ground-glass nodules (GGNs), lesion size was a significant differentiating feature of preinvasive lesions from invasive pulmonary adenocarcinoma (IPA), with an optimal cut-off value of less than 10 mm (53.33% sensitivity [24 of 45]; 100% specificity [19 of 19]). nn In part-solid GGNs, smaller lesion size, smaller solid proportion, nonlobulated border, and nonspiculated margin were significantly associated with the presence of a preinvasive lesion (P,.05). nn When lesion size, solid proportion, lesion border, and lesion margin were used as input variables in a logistic regression model to differentiate preinvasive lesions from IPAs appearing as part-solid GGNs, the sensitivity and specificity of the logistic regression model were 93.75% (45 of 48) and 77.50% (124 of 160), respectively. acceptable for AIS (previously BAC), as well as AAH, although AIS is one type of pulmonary adenocarcinoma. Thus, it is clearly important to accurately diagnose preinvasive lesions for pulmonary adenocarcinoma (AIS and AAH) prior to or during surgery. Unfortunately, it has been challenging to diagnose preinvasive lesions accurately even through intraoperative histologic evaluation by using frozen specimens, as a definitive diagnosis of a preinvasive lesion requires thorough histologic evaluation of the entire tumor specimen to rule out the possibility of invasive components. Therefore, if we can accurately diagnose invasive pulmonary adenocarcinomas (IPAs) and preinvasive lesions through imaging prior to surgery, we may be able to accurately select patients suitable for sublobar resection. Today, computed tomographic (CT) features of IPAs containing lepidic components and preinvasive lesions including AIS and AAH have been reported to be pure or part-solid groundglass nodules (GGNs); however, there have been reports of considerable overlap between IPA, AIS, and AAH (10 16). Until now, most radiologic research for GGNs has been focused on the differentiation of malignant (IPAs and AIS) from benign (AAHs) GGNs (13,17) or the differentiation among AAHs, BACs, and IPAs on CT scans (18 20). Consequently, the results of previous studies were not able to be directly applied to the differentiation of preinvasive lesions (AAH and AIS) from IPAs in the new classification for pulmonary adenocarcinoma. To our knowledge, however, there have been no studies as of yet that have attempted to differentiate preinvasive lesions from IPAs. Therefore, the purpose of our study was to retrospectively investigate the differentiating CT features between IPAs and preinvasive lesions appearing as GGNs in 253 patients. Implication for Patient Care nn CT features can be used to accurately differentiate preinvasive lesions from IPAs appearing as GGNs. Materials and Methods This retrospective study was approved by the institutional review board of Seoul National University Hospital, which waived the requirement for patients informed consent. Study Population From January 2005 to October 2011, a search of the electronic medical records and the radiology information systems of our hospital by one radiologist (S.M.L.) for patients with pulmonary GGNs identified on chest CT scans was performed. Our study population was selected in the following steps: First, we selected all CT scans for which the reports included the words GGO, GGN, part-solid nodule, nonsolid nodule, ground-glass opacity or ground-glass nodule. A total of 16720 CT scans in 9942 patients were found. Of the 16720 CT scans, only 7759 scans in 4010 patients with thin-section images (section thickness 2.5 mm) were included. Second, two radiologists (H.J.L. and S.M.L., with 13 and 5 years of experience in chest CT, respectively) reviewed all of the CT scans and excluded diffuse GGOs, very small GGNs less than 5 mm, large lesions greater than 3 cm in size, and transient GGNs, leaving 303 patients Published online before print 10.1148/radiol.13120949 Content code: Radiology 2013; 268:265 273 Abbreviations: AAH = adenomatous hyperplasia AIS = adenocarcinoma in situ AUC = area under the ROC curve BAC = bronchioloaveolar carcinoma GGO = ground-glass nodule IPA = invasive pulmonary adenocarcinoma ROC= receiver operating characteristic Author contributions: Guarantors of integrity of entire study, S.M.L., C.M.P., J.Y.W.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, S.M.L., C.M.P., J.M.G., C.H.K.; clinical studies, all authors; statistical analysis, S.M.L.; and manuscript editing, all authors Conflicts of interest are listed at the end of this article. 266 radiology.rsna.org n Radiology: Volume 268: Number 1 July 2013

Figure 1 with 327 GGNs. Among the 327 GGNs, an additional 55 GGNs that were not pathologically confirmed were excluded. Finally, 253 patients (mean age, 59.4 years 6 9.3 [standard deviation]; range, 29 80 years) with 272 pathologically proved GGNs constituted our study population (Fig 1). There were 89 men (mean age, 60.0 years 6 9.1; range, 36 80 years) and 164 women (mean age, 59.0 years 6 9.4; range, 29 77 years). Of 272 GGNs, 199 were identical to the study population of our previous reports (81 GGNs in study [17] and 160 GGNs in study [21]). Preinvasive lesions included AAHs and AISs, which were defined as lesions showing no stromal, vascular, or pleural invasion (1). IPAs were defined as adenocarcinomas containing an invasive component, that is, minimally invasive adenocarcinoma and invasive adenocarcinoma with several subtypes such as lepidic predominant, acinar predominant, or papillary predominant adenocarcinoma (1). In our study, among the 272 pathologically proved GGNs, 93 were preinvasive lesions and 179 were IPAs. Among the 93 preinvasive lesions, 21 were AAH and the remaining 72 were AIS. There were 64 pure GGNs and 208 part-solid GGNs. Among the 253 patients, 208 patients underwent lobectomy, 37 underwent wedge resection or segmentectomy, six underwent lobectomy and wedge resection, and two underwent biopsy Figure 1: Flowchart of study population. Numbers in parentheses are the number of patients. with confirmation of IPA. Among the 272 GGNs, solitary GGNs were found in 240 patients, two GGNs were found in eight, three GGNs were found in four, and four GGNs were found in one. CT Examinations Chest CT was performed by using the following scanners: Somatom Definition, Sensation-16 (Siemens Medical Solutions, Forchheim, Germany), Brilliance-64 (Phillips Medical Systems, the Netherlands), and Lightspeed Ultra (GE Medical Systems, Milwaukee, Wis) with 120 kvp, 100 200 mas, pitch of 0.875 1.5, and collimation of 1 2.5 mm. Images were reconstructed by using a medium sharp reconstruction algorithm with a thickness of 1 2.5 mm. CT scans were obtained in all patients in the supine position at full inspiration. A total of 253 chest CT scans in the 253 patients were evaluated for CT morphologic analysis of GGNs; 85 patients underwent only one chest CT and 168 underwent more than one chest CT prior to surgical resection of GGNs. When several CT scans were available for CT morphologic analysis, we selected the last CT scan prior to surgery or biopsy. The mean interval between CT and surgery or biopsy was 19.0 days (range, 1 87 days; median, 12 days). Of the 253 chest CT examinations, 141 were non contrast material enhanced and 112 were contrast enhanced. In the case of contrast-enhanced CT, 100 ml of contrast medium was injected at a rate of 2 ml/sec. Analysis of CT Features Two chest radiologists (C.M.P. and J.M.G., with 11 and 19 years of experience in chest CT, respectively) who were blinded to the pathologic results evaluated the CT scans in consensus. The CT findings were analyzed in the lung window setting (window level, 2700 HU; width, 1500 HU) by using a picture archiving and communication system (Marotech, Seoul, Korea) and flat-panel monochrome 3-megapixel monitors (ME315; Totoku, Tokyo, Japan). All GGNs were divided into pure and part-solid category according to the existence of an internal solid component at CT. An internal solid component was defined as any opacity that completely obscured the lung parenchyma. CT findings that were analyzed for each lesion included (a) lesion location, (b) lesion size, (c) lesion multiplicity (solitary, multiple), (d) margin (spiculated, nonspiculated), (e) border (lobulated, nonlobulated), (f) solid portion size, (g) solid proportion, (h) presence of bubble lucency, and (i) presence of pleural retraction. To determine the lesion size and the size of the solid portion, one author (C.M.P.) recorded the greatest diameter of the entire nodule and internal solid portion on axial images. The solid proportion of the lesion was calculated by dividing the solid portion size by the lesion size. We did not analyze the attenuation value of GGNs as the diversity in CT protocols in terms of section thickness and contrast enhancement precluded reliable and exact evaluation of attenuation of the lesions. In addition, one chest radiologist (H.J.L., with 13 years of experience in chest CT) blinded to the pathologic results of GGNs and CT morphologic analysis by the two observers (C.M.P. and J.M.G.) analyzed the CT features of GGNs independently to evaluate interobserver agreement for CT morphologic features of GGNs. The analyzed items included (a) nodule type (pure, part-solid GGN), (b) lesion location, (c) lesion size, (d) lesion multiplicity (solitary, multiple), (e) margin Radiology: Volume 268: Number 1 July 2013 n radiology.rsna.org 267

(spiculated, nonspiculated), (f) border (lobulated, nonlobulated), (g) solid portion size, (h) solid proportion, (i) presence of bubble lucency, and (j) presence of pleural retraction. Statistical Analysis Statistical analysis was performed separately for 64 pure and 208 part-solid GGNs as the malignancy probability of pure and part-solid GGNs is distinctly different (10), with different management strategies for pure and part-solid GGNs (22). Statistical differences between preinvasive lesions and IPAs were analyzed by using the independent sample t test for differences in mean patient age, lesion size, and solid portion size. Statistical differences between patient s sex and CT findings were analyzed by using the Pearson x 2 test and Fisher exact test, as appropriate. In our study, we did not take into account within-patient correlation among multiple GGNs in a patient since multiple GGNs in a patient have been considered as multiple independent primary lesions (23,24). For the differentiation of preinvasive lesions from IPAs, the optimal cut-off values of lesion size in pure GGNs, lesion size and solid proportion in part-solid GGNs were calculated by using receiver operating characteristic (ROC) curve analysis. The optimal cut-off values were determined as the point closest to the upper left hand corner of the ROC curve. To identify variables that could be used in differentiating preinvasive lesions from IPAs, logistic regression analysis was conducted. Characteristics with a P value of less than.10 at univariate analysis were used as the input variables for multiple logistic regression analysis. In multiple logistic regression analysis, a backward stepwise selection mode was used, with iterative entry of variables on the basis of test results (P,.05). The removal of variables was based on likelihood ratio statistics with a probability of.10. A C statistic and ROC analysis were also performed to evaluate the differentiating performance of multiple logistic regression models in discriminating preinvasive lesions from IPAs. Table 1 Patient Demographics Characteristic Preinvasive Lesions IPAs P Value Pure GGNs (n = 56) Age (y)* 57.9 6 8.6 58.6 6 5.5.700 Male-to-female ratio.765 No. of men 15 7 No. of women 25 9 Lesion multiplicity.186 Solitary 27 14 Multiple 13 2 Part-solid GGNs (n = 197) Age (y)* 57.5 6 9.5 60.4 6 9.7.078 Male-to-female ratio..99 No. of men 16 51 No. of women 30 100 Lesion multiplicity.160 Solitary 36 131 Multiple 10 20 Note. Except where indicated, data are the number of patients. * Data are means 6 standard deviation. Independent sample t test. Fisher exact test. Pearson x 2 test. Interobserver agreement for CT findings of GGNs was investigated by using weighted k statistic for qualitative features and linear regression analysis for continuous variables, including lesion size and solid portion size. According to Landis and Koch (25), a k value of less than 0.00 indicates poor agreement; k value of 0.00 0.20, slight agreement; k value of 0.21 0.40, fair agreement; k value of 0.41 0.60, moderate agreement; k value of 0.61 0.80, substantial agreement; and k value of 0.81 1.00, almost perfect agreement. All statistical analyses were performed by using SPSS, version 18.0 (SPSS, Chicago, Ill) and MedCalc, version 12.2.1 (MedCalc, Mariakerke, Belgium) software. A P value of less than.05 was considered to indicate a significant difference. Results Demographic and CT Findings of Pure GGNs Among the demographic findings, there were no significant differences between preinvasive lesions and IPAs appearing as pure GGNs (P..05) (Table 1). However, among CT findings, preinvasive lesions were significantly more frequently nonlobulated (P =.005) and smaller (10.3 mm 6 4.4 vs 14.7 mm 6 5.1, P =.001) in comparison to IPAs. Table 2 summarizes the comparison of CT findings between preinvasive lesions and IPAs appearing as pure GGNs. Logistic Regression Analysis and ROC Analysis of Pure GGNs Lesion size and nonlobulated border were used as input variables for multiple logistic regression analysis. Multivariate analysis revealed that lesion size was the sole significant differentiator of preinvasive lesions from IPAs (P =.029; 95% confidence interval: 0.754, 0.985; adjusted odds ratio, 0.862). The P value of nonlobulated border was.80 (odds ratio, 4.289). ROC analysis showed that the area under the ROC curve (AUC) for lesion size was 0.773 (95% confidence interval: 0.651, 0.868) and that the optimal cutoff value of lesion size for differentiating 268 radiology.rsna.org n Radiology: Volume 268: Number 1 July 2013

Table 2 Figure 2 CT Findings of Preinvasive Lesions and IPAs in Pure GGNs Characteristic Preinvasive Lesions (n = 45) IPAs (n = 19) P Value Lesion size (mm)* 10.3 6 4.4 14. 6 5.1.001 Lesion margin.509 Speculated 1 1 Nonspiculated 44 18 Lesion border.005 Lobulated 3 7 Nonlobulated 42 12 Bubble lucency 4 5.111 Pleural retraction 6 3..99 Note. Except where indicated, data are the number of nodules. * Data are means 6 standard deviation. Independent sample t test. Fisher exact test. Table 3 CT Findings of Preinvasive Lesions and IPAs in Part-Solid GGNs Characteristic Preinvasive Lesions (n = 48) IPAs (n = 160) P Value Lesion size (mm)* 12.6 6 5.0 18.1 6 5.4,.001 Size of solid portion (mm)* 3.5 6 2.6 10.2 6 5.8,.001 Solid proportion (%)* 29.6 6 18.1 56.7 6 26.6,.001 Lesion margin,.001 Spiculated 1 54 Nonspiculated 47 106 Lesion border.017 Lobulated 21 101 Nonlobulated 27 59 Bubble lucency 6 33.291 Pleural retraction 20 107.002 Note. Except where indicated, data are the number of nodules. * Data are means 6 standard deviation. Independent sample t test Fisher exact test. Pearson x 2 test. preinvasive lesions from IPAs was less than 10 mm (sensitivity, 53.33%; specificity, 100%) (Fig 2). There were no IPAs among pure GGNs smaller than 10 mm in the present study. Demographic and CT Findings of Part- Solid GGNs Among the demographic findings, there were no significant differences between preinvasive lesions and IPAs appearing as part-solid GGNs (P..05) (Table 1). Among the CT findings, there were significant differences between preinvasive lesions and IPAs in terms of lesion size (12.6 mm 6 5.0 vs 18.1 mm 6 5.4, P,.001). In addition, the size of the solid portion and solid proportion of preinvasive lesions was significantly smaller than that of IPAs (P,.001). With respect to lesion margin and border, preinvasive lesions were more frequently nonspiculated (P,.001) and nonlobulated Figure 2: CT scans of preinvasive lesion and IPA appearing as pure GGNs. (a) An 8-mm faint pure GGN (arrow) in the left upper lobe in a 55-year-old woman. This nodule was confirmed as atypical AAH at wedge resection. (b) An 18-mm well-defined pure GGN (arrow) in the left upper lobe in a 65-year-old woman. This nodule was confirmed as adenocarcinoma at lobectomy. compared with IPAs (P =.017). Furthermore, preinvasive lesions showed pleural retraction less frequently than did IPAs (P =.002). Table 3 summarizes the comparison of CT findings Radiology: Volume 268: Number 1 July 2013 n radiology.rsna.org 269

Table 4 Logistic Regression Analysis for CT Determination of Preinvasive Lesions and IPAs in Part-Solid GGNs Variable Odds Ratio 95% Confidence Interval P Value Lesion size 0.819 0.747, 0.898,.001 Solid proportion 0.953 0.933, 0.977,.001 Nonlobulated border 2.856 1.156, 7.057.023 Nonspiculated margin 26.797 2.978, 241.106,.001 Figure 3 Figure 3: CT scans of preinvasive lesion and IPAs appearing as part-solid GGNs. (a) A 14-mm part-solid GGN with a small solid portion (arrow) in the left upper lobe in a 68-year-old man. This nodule was confirmed as AIS at lobectomy. (b) A 25-mm part-solid GGN with a lobulated border and air bronchogram (arrow) in the right upper lobe in a 69-year-old man. This nodule was confirmed as adenocarcinoma at lobectomy. (c) A 15-mm partsolid GGN with spiculated margin (arrow) in the right middle lobe in a 57-year-old woman. This nodule showed a large solid proportion and was confirmed as adenocarcinoma at lobectomy. between preinvasive lesions and IPAs appearing as part-solid GGNs. Logistic Regression Analysis and ROC Analysis of Part-Solid GGNs Age, lesion size, solid portion size, solid proportion, lesion border, lesion margin, and pleural retraction were used as input variables for multivariate logistic regression analysis. Multivariate analysis revealed that smaller lesion size, smaller solid proportion, nonlobulated border, and nonspiculated margin were significant independent factors of preinvasive lesions (P,.001, P,.001, P =.023, P,.001, respectively) (Fig 3). The adjusted odds ratio for lesion size, solid proportion, nonlobulated border, and nonspiculated margin was 0.819, 0.953, 2.856, and 26.797, respectively (Table 4). ROC analysis revealed that the AUC for lesion size and solid proportion was 0.765 (95% confidence interval: 0.702, 0.821) and 0.786 (95% confidence interval: 0.724, 0.840), respectively. The optimal cut-off value for lesion size and solid proportion for differentiating preinvasive lesions from IPAs was 14 mm or less (sensitivity, 66.67%; specificity, 73.75%) and 28.6% or less (sensitivity, 68.75%; specificity, 80.00%), respectively. Performance of Logistic Regression Model in Part-Solid GGNs C statistic analysis was conducted to evaluate the performance of the logistic regression model in discriminating preinvasive lesions from IPAs, and the AUC was 0.905 (95% confidence interval: 0.857, 0.941). The sensitivity and specificity of the logistic regression model were 93.75% and 77.50%, respectively. The performance of the logistic regression model by using lesion size, solid proportion, and morphologic CT features was significantly higher than that using only lesion size or solid proportion alone (P,.001) (Fig 4). Interobserver Agreement for CT Findings of GGNs The results of interobserver agreement for various CT features of GGNs are summarized in Table 5. k values of qualitative CT morphologic features of GGNs ranged from 0.460 to 0.944. Lesion multiplicity and nodule type showed almost perfect agreement (k = 0.944 and 0.861, respectively). Other than lesion multiplicity and nodule type, interobserver agreement for CT findings showed moderate to substantial agreement and was highest for 270 radiology.rsna.org n Radiology: Volume 268: Number 1 July 2013

Figure 4 Table 5 Interobserver Agreement for CT Findings of GGNs Characteristic Observer 1 Observer 2 Interobserver Agreement* Figure 4: Graph shows results of C statistic analysis of logistic regression model in discriminating preinvasive lesions from IPAs appearing as partsolid GGNs. The AUC of the logistic model including lesion size, solid proportion, lesion border, and lesion margin was significantly higher (AUC = 0.905) than the AUC of lesion size alone (AUC = 0.765) or solid proportion alone (AUC = 0.786) (P,.001). Nodule type 0.861 Pure 64 70 Part solid 208 202 Lesion multiplicity 0.944 Solitary 208 212 Multiple 45 41 Lesion margin 0.460 Spiculated 57 58 Nonspiculated 215 214 Lesion border 0.532 Lobulated 132 93 Nonlobulated 140 179 Bubble lucency 0.521 Yes 48 42 No 224 230 Pleural retraction 0.654 Yes 136 99 No 136 173 Note. Except where indicated, data are the number of GGNs. * Data are k values. Data are the number of patients. pleural retraction (k = 0.654). Linear regression analysis for lesion size and solid portion size was y = 0.6950 + 0.9442 3 (R 2 = 0.9203, P,.001) and y = 20.07903 + 0.9786 3 (R 2 = 0.8783, P,.001), respectively. Discussion The present study is the first large study to investigate the differentiating CT features of preinvasive lesions from IPAs appearing as GGNs in the light of the new adenocarcinoma classification. We found that the optimal cut-off value for discriminating preinvasive lesions from IPAs appearing as pure GGNs was 10 mm. Previous studies (17,23) had shown that 8 mm was the optimal cut-off value for distinguishing benign from malignant pure GGNs. The difference between our study and the previous studies can be explained by the fact that previous studies (17,23) included BACs as malignant GGNs, whereas BAC, now referred to as AIS, was classified as preinvasive lesions in our study. Thus, the higher cut-off value for preinvasive lesions than that for benign lesions manifesting as pure GGNs is reasonable. The cut-off value of 10 mm for pure GGNs in our study correlates well with the results of previous studies (14,26). In one study, (14) showed that all 11 pure GGNs 10 mm or smaller were preinvasive lesions consisting of six AAHs and five BACs. Nakata et al (26) also showed that 93.0% (53 of 57) of pure GGNs 10 mm or smaller were AAH or BAC. Furthermore, we found that the specificity of the cut-off value of 10 mm in pure GGNs was 100%. This means that a 10-mm cut-off value for pure GGNs can allow complete exclusion of IPAs from preinvasive lesions and that it can be used as an appropriate criterion for selection of patients suitable for sublobar resection. Interestingly, our results showed good correlation with the interim guidelines of pulmonary GGNs suggested by Godoy et al (22), in which conservative management of nodules 5 10 mm in size with pure GGOs and resection of solitary pure GGOs 10 mm or larger was recommended. In our study, there were no other significant CT morphologic differentiators of preinvasive lesions from IPAs appearing as pure GGNs. Although a nonlobulated border was significantly more frequent in preinvasive lesions than IPAs in univariate analysis, as in study (17), a nonlobulated border did not prove to be a significant factor in multivariate analysis. Thus, when pure GGNs are encountered in clinical practice, lesion size would be the most important criterion. As for part-solid GGNs, there were significant differences between preinvasive lesions and IPAs in lesion size, solid portion size, solid proportion, lesion margin, border, and pleural retraction (P,.05). Multivariate analysis revealed that smaller lesion size, smaller solid proportion, nonlobulated border, and nonspiculated margin were significant independent differentiators of preinvasive lesions from IPAs. In pulmonary nodules, it has been known that larger nodules are more likely to be malignant than smaller nodules (16), Radiology: Volume 268: Number 1 July 2013 n radiology.rsna.org 271

and that the solid component of GGNs usually represents areas of fibroblastic proliferation or invasive components of the tumor (15). Taking into consideration the results of these two studies (15,16), it may be reasonable to infer that smaller GGNs with smaller solid proportions have a higher probability of preinvasive lesions. The morphologic CT features of a nonlobulated border and nonspiculated margin for preinvasive lesions in partsolid GGNs also accord closely with previous studies (17,27), in which a lobulated border and spiculated margin were predictive CT features of malignant lesions. We also found that a morphologic lesion analysis that included lesion margin and lesion border enhanced our performance in differentiating preinvasive lesions from IPAs compared with analysis that included lesion size or solid proportion alone. When lesion size or solid proportion was applied alone, the sensitivity and specificity were 66.67% and 73.75% or 68.75% and 80.00%, respectively. Considering the sensitivity of 93.75% and specificity of 77.50% of the logistic regression model, we believe that morphologic CT features including lesion margin and lesion border can increase sensitivity, especially in comparison to the use of lesion size or solid proportion alone, to differentiate preinvasive lesions from IPAs. This result provides evidence of the value of meticulous CT analysis and the potential inclusion of CT features in future guidelines for management of part-solid GGNs. Interobserver agreement for CT findings of GGNs in our study was moderate to very good. In particular, classification of GGNs into pure and part-solid GGNs, which is one of the most important steps in the management decision of GGNs, showed very good interobserver agreement (k = 0.861). This result can support the reliability and applicability of GGN management guidelines according to nodule type. Lesion size and solid portion size are also of great importance since nodule management is based on lesion size or solid portion size. In our study, lesion size and solid portion size of GGNs showed a squared linear correlation coefficient of more than 0.9 despite the fact that some GGNs had ill-defined or spiculated margins. We believe this result can also provide a sound basis for nodule management with use of lesion size or solid portion size. However, lesion border or lesion margin showed only moderate interobserver agreement, thus a more objective and practical definition is necessary to widely use these findings for differentiation of preinvasive lesions from IPAs. Our study had several limitations. First, our study was of retrospective design. Therefore, there may have been selection bias. In this study we included only pathologically proved GGNs, of which CT features could be more likely to be a malignancy. However, we believe that our results reflected the actual clinical practice and can help radiologists and physicians to differentiate preinvasive lesions from IPAs in their daily clinical practice. Second, we retrospectively searched for individuals with pulmonary GGNs identified at CT using the electronic medical records and radiology information system of our hospital. Thus, there is a possibility that nodules might have been unreported in the original clinical readings and therefore missed using this search method. Third, in our study, CT morphologic assessment for differentiating preinvasive lesions from IPAs was made in consensus of two observers, and we did not analyze differentiating CT features on the basis of each observer s assessment. However, additional CT morphologic assessment by another observer showed the interobserver agreement for these CT features was moderate to very good in our study. In conclusion, in pure GGNs, a lesion size of less than 10 mm can be a very specific discriminator of preinvasive lesions from IPAs and in part-solid GGNs, preinvasive lesions can be accurately distinguished from IPAs by their smaller lesion size, smaller solid proportion, nonlobulated border, and nonspiculated margin. Acknowledgment: We would like to express special thanks to Chris Woo, BA, for his editorial assistance. Disclosures of Conflicts of Interest: S.M.L. No relevant conflicts of interest to disclose. C.M.P. No relevant conflicts of interest to disclose. J.M.G. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: received fees for consultancy from Infinitt Healthcare. Other relationships: none to disclose. H.J.L. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: institution has grants from Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2012R1A1A2006795), research grant 05-2012-0010 (2012-1356) from the SNUH Research Fund. Other relationships: none to disclose. J.Y.W. No relevant conflicts of interest to disclose. C.H.K. No relevant conflicts of interest to disclose. References 1. Travis WD, Brambilla E, Noguchi M, et al. International association for the study of lung cancer/american thoracic society/european respiratory society international multidisciplinary classification of lung adenocarcinoma. J Thorac Oncol 2011;6(2):244 285. 2. Noguchi M, Morikawa A, Kawasaki M, et al. Small adenocarcinoma of the lung: histologic characteristics and prognosis. Cancer 1995; 75(12):2844 2852. 3. Watanabe S, Watanabe T, Arai K, Kasai T, Haratake J, Urayama H. Results of wedge resection for focal bronchioloalveolar carcinoma showing pure ground-glass attenuation on computed tomography. Ann Thorac Surg 2002;73(4):1071 1075. 4. Sakurai H, Dobashi Y, Mizutani E, et al. Bronchioloalveolar carcinoma of the lung 3 centimeters or less in diameter: a prognostic assessment. Ann Thorac Surg 2004; 78(5):1728 1733. 5. Koike T, Togashi K, Shirato T, et al. Limited resection for noninvasive bronchioloalveolar carcinoma diagnosed by intraoperative pathologic examination. Ann Thorac Surg 2009; 88(4):1106 1111. 6. Van Schil PE, Asamura H, Rusch VW, et al. Surgical implications of the new IASLC/ ATS/ERS adenocarcinoma classification. Eur Respir J 2012;39(2):478 486. 7. El-Sherif A, Gooding WE, Santos R, et al. Outcomes of sublobar resection versus lobectomy for stage I non-small cell lung cancer: a 13-year analysis. Ann Thorac Surg 2006;82(2):408 415; discussion 415 416. 8. Okada M, Koike T, Higashiyama M, Yamato Y, Kodama K, Tsubota N. Radical sublobar resection for small-sized non-small cell lung cancer: a multicenter study. J Thorac Cardiovasc Surg 2006;132(4):769 775. 272 radiology.rsna.org n Radiology: Volume 268: Number 1 July 2013

9. Sugi K, Kobayashi S, Sudou M, Sakano H, Matsuda E, Okabe K. Long-term prognosis of video-assisted limited surgery for early lung cancer. Eur J Cardiothorac Surg 2010; 37(2):456 460. 10. Henschke CI, Yankelevitz DF, Mirtcheva R, et al. CT screening for lung cancer: frequency and significance of part-solid and nonsolid nodules. AJR Am J Roentgenol 2002; 178(5):1053 1057. 11. Nakata M, Saeki H, Takata I, et al. Focal ground-glass opacity detected by low-dose helical CT. Chest 2002;121(5):1464 1467. 12. Park CM, Goo JM, Lee HJ, et al. CT findings of atypical adenomatous hyperplasia in the lung. Korean J Radiol 2006;7(2):80 86. 13. Kim HY, Shim YM, Lee KS, Han J, Yi CA, Kim YK. Persistent pulmonary nodular groundglass opacity at thin-section CT: histopathologic comparisons. Radiology 2007;245(1): 267 275. 14. Lee HJ, Goo JM, Lee CH, Yoo CG, Kim YT, Im JG. Nodular ground-glass opacities on thin-section CT: size change during follow-up and pathological results. Korean J Radiol 2007;8(1):22 31. 15. Park CM, Goo JM, Lee HJ, Lee CH, Chun EJ, Im JG. Nodular ground-glass opacity at thin-section CT: histologic correlation and evaluation of change at follow-up. Radio- Graphics 2007;27(2):391 408. 16. Goo JM, Park CM, Lee HJ. Ground-glass nodules on chest CT as imaging biomarkers in the management of lung adenocarcinoma. AJR Am J Roentgenol 2011;196(3):533 543. 17. Lee HJ, Goo JM, Lee CH, et al. Predictive CT findings of malignancy in ground-glass nodules on thin-section chest CT: the effects on radiologist performance. Eur Radiol 2009; 19(3):552 560. 18. Nomori H, Ohtsuka T, Naruke T, Suemasu K. Differentiating between atypical adenomatous hyperplasia and bronchioloalveolar carcinoma using the computed tomography number histogram. Ann Thorac Surg 2003; 76(3):867 871. 19. Oda S, Awai K, Liu D, et al. Ground-glass opacities on thin-section helical CT: differentiation between bronchioloalveolar carcinoma and atypical adenomatous hyperplasia. AJR Am J Roentgenol 2008;190(5):1363 1368. 20. Ikeda K, Awai K, Mori T, Kawanaka K, Yamashita Y, Nomori H. Differential diagnosis of ground-glass opacity nodules: CT number analysis by three-dimensional computerized quantification. Chest 2007; 132(3):984 990. 21. Lee SM, Park CM, Paeng JC, et al. Accuracy and predictive features of FDG-PET/CT and CT for diagnosis of lymph node metastasis of T1 non-small-cell lung cancer manifesting as a subsolid nodule. Eur Radiol 2012;22(7): 1556 1563. 22. Godoy MC, Naidich DP. Subsolid pulmonary nodules and the spectrum of peripheral adenocarcinomas of the lung: recommended interim guidelines for assessment and management. Radiology 2009;253(3):606 622. 23. Kim HK, Choi YS, Kim K, et al. Management of ground-glass opacity lesions detected in patients with otherwise operable non-small cell lung cancer. J Thorac Oncol 2009;4(10): 1242 1246. 24. Vazquez M, Carter D, Brambilla E, et al. Solitaryand multiple resected adenocarcinomas after CT screening for lung cancer: histopathologic features and their prognostic implications. Lung Cancer 2009;64(2):148 154. 25. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159 174. 26. Nakata M, Sawada S, Saeki H, et al. Prospective study of thoracoscopic limited resection for ground-glass opacity selected by computed tomography. Ann Thorac Surg 2003; 75(5):1601 1605; discussion 1605 1606. 27. Aoki T, Tomoda Y, Watanabe H, et al. Peripheral lung adenocarcinoma: correlation of thin-section CT findings with histologic prognostic factors and survival. Radiology 2001; 220(3):803 809. Radiology: Volume 268: Number 1 July 2013 n radiology.rsna.org 273