Introduction

Adenocarcinoma is the most common histological type of lung cancer [1]. In 2011, a new classification of lung adenocarcinoma was purposed by an international multidisciplinary committee sponsored by the International Association for the Study of Lung Cancer (IASLC), the American Thoracic Society (ATS), and the European Respiratory Society (ERS) [2]. This classification addresses recent advances in oncology, molecular biology, pathology, radiology, and surgery of lung adenocarcinoma. For resection specimens, new concepts are introduced such as adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) for small solitary adenocarcinomas with either pure lepidic growth (AIS) or predominant lepidic growth with ≤ 5 mm invasion (MIA) to define patients who, if they undergo complete resection, will have 100 % or near 100 % disease-specific survival, respectively. AIS and MIA are usually nonmucinous but rarely may be mucinous. Invasive adenocarcinomas are classified by a predominant pattern after using comprehensive histologic subtyping with lepidic (formerly most mixed subtype tumours with nonmucinous BAC), acinar, papillary, and solid patterns. Micropapillary has been added as a new histological subtype. Variants include invasive mucinous adenocarcinoma (formerly mucinous BAC), colloid, fetal, and enteric adenocarcinoma. This new classification especially affects computed tomographic (CT) diagnosis of lung adenocarcinoma. Lee and Godoy et al. [3, 4] reported that the most common CT findings of the new classification of lung adenocarcinoma as follows: AAH: pure ground-glass nodule (GGN); nonmucinous AIS: typically a pure GGN, sometimes a part-solid; nonmucinous MIA: a part-solid nodule with a predominant ground-glass component; LPA: a part-solid mass with variable proportion of ground-glass components or solid nodule; and another subtype predominant in adenocarcinoma: usually a solid mass or part-solid nodule (with predominant solid component). Those CT characteristics were determined mostly dependent on the studies before the new purposed classification. The radiologic-pathologic correlation under the 2011 classification was not fully established. The aim of this study was to clarify the correlation between CT findings of early stage lung adenocarcinoma(T1N0M0)with pure ground-glass opacity (GGO) and the new IASLC classification.

Materials and methods

Patients

From January 2010 to December 2012, nearly 3,500 consecutive patients with lung adenocarcinoma in our hospital, who underwent a complete surgical resection, were reviewed. Only lesions presenting pGGO on CT and with clinical stage of T1N0M0 were selected. Ninety-four lesions in 88 patients (82 patients had a single lesion and six had two lesions) met the inclusion criteria. There were 33 male and 55 female patients (average age, 57.13 years; range, 26 to 78 years).

Image acquisition and analysis

CT was performed by using Somatom Sensation-64 (Siemens Medical Systems, Germany) with 120kVp and 100mAs. All images were reconstructed with a section thickness of 1.5 mm and displayed at a lung window width of 1600HU, window level of -600HU, and a mediastinal window width of 350HU, with a window level of 35HU. CT characteristics were analyzed by the consensus of two experienced thoracic radiologists (Zhao SH, Wu CC, with 20 and 16 years of experience in chest CT, respectively) who worked together on PACS with measurement done by Dr. Zhao or Wu. The location, size, CT attenuation, uniformity, shape, margin, internal and surrounding malignant signs of the nodules, and tumour-lung interface were recorded. The nodule size was the maximal diameter of the lesion on axial images. The average CT attenuation (HU) was measured by placing a region of interest (ROI) of 15 mm2 in three different sites within the nodule barring the vessels and bronchioles. The uniformity was described as homogeneous, less homogeneous, and of heterogeneous density. No bubble-like lucency within the nodule was defined as homogeneous density, more than three lucency areas was defined as heterogeneous density and less homogeneous density was defined between the above two situations. The shape of nodules was described as round, oval, polygonal, and irregular. The margin was described as smooth, lobulated, spiculated, and lobulated with spiculation. Tumour-lung interface was recorded as well-demarcated or ill-demarcated. The internal and surrounding malignant signs included bubble sign, air bronchogram, and pleural tags. Multiple GGO nodules in one patient were considered as independent primary lesions. According to the IASLC/ATS/ERS classification, all lesions were categorized as preinvasive lesions: atypical adenomatous hyperplasia (AAH) and adenocarcinoma in situ (AIS); minimally invasive adenocarcinoma (MIA); and invasive adenocarcinoma (including lepidic, acinar, papillary, solid and micropapillary predominant adenocarcinomas)

Statistical analysis

Statistical analysis was performed with SPSS software (version 18.0). The correlation of nodule size and CT value with histological invasiveness were analyzed by using one-way analysis of variance (ANOVA). The Student- Newman-Keuls test and Kruskal-Wallis test were used depending on the homogeneity of variance. The correlation of patient’s sex, nodule location, CT findings (shape, margin, tumour-lung interface, internal and surrounding characteristics) with histological invasiveness were analyzed by using the Pearson χ2 test and Fisher exact test. The optimal cut-off value of lesion size between preinvasive lesions and invasive lesions (MIA and invasive adenocarcinoma) was calculated by using a receiver operating characteristic (ROC) curve. Correlation analysis was used for the relationship between lesion size and lesion uniformity. A P value of less than 0.05 was considered to be a significant difference.

Results

Of the 94 nodules, there were 21preinvasive lesions (seven AAHs and 14 AISs) (M:F = 6:15), 35 MIAs (M:F = 14:21), 38 invasive adenocarcinomas (M:F = 14:24) including 27 lepidic predominant, four acinar predominant, three papillary predominant, one micropapillary predominant, and three solid predominant lesions. There were no significant differences in sex ratio and lesion location among preinvasive lesions, MIAs and invasive adenocarcinomas (P = 0.69 and 0.37,P > 0.05) (Table 1). The mean diameter of preinvasive lesions was 12.4±6.8 mm (peak point: 7.0 mm), that of MIAs was 16.0±5.2 mm (peak point: 10.0 mm), and that of invasive adenocarcinomas was 17.5±5.8 mm (peak point: 20.0 mm) (P = 0.008,P < 0.05) (Fig. 1). ROC analysis revealed that the AUC for lesion size was 0.710 (95 % CI: 0.561-0.860). The optimal cut-off value for lesion size in differentiating preinvasive lesions from invasive lesions was 10.5 mm, with a sensitivity of 86.30 %, a specificity of 61.90 %, a positive predictive value of 88.73 %, a negative predictive value of 56.52 %, and an accuracy of 80.85 %, respectively (Fig. 2). The mean CT value was (-531.0±84.6) HU for preinvasive lesions, (-492.7±127.2) HU for MIAs, and (-512.4±109.4) HU for invasive adenocarcinomas. There were no significant differences among preinvasive lesions, MIAs, and invasive adenocarcinomas (P = 0.39,P > 0.05). There was 38.10 % of homogeneity, 52.38 % of less homogeneity and 9.52 % of heterogeneity in preinvasive lesions; 42.86 % of homogeneity, 45.71 % of less homogeneity, and 11.43 % of heterogeneity in MIAs; 7.89 % of homogeneity, 47.37 % of less homogeneity and 44.74 % of heterogeneity in invasive adenocarcinomas. The uniformity of lesions in three groups showed significant differences (P = 0.01,P < 0.05). The lesion size was closely related with lesion uniformity, correlation index was r = 0.45 (P = 0.00,P < 0.05).

Table 1 Lesion location of all nodules*
Full size table
Fig. 1
figure 1

Distribution histogram of lesion size in three groups (A) Preinvasive lesions: main distribution range was 5.0-10.0 mm, peak point was 7.0 mm; (B) MIAs: main distribution range was 10.0-15.0 mm, peak point was10.0 mm; (C) invasive adenocarcinomas: main distribution range was 20.0-22.5 mm, peak point was 20.0 mm

Full size image
Fig. 2
figure 2

A receiver operating characteristic curve of lesion size

Full size image

The morphologic characteristics of nodules regarding the margin, tumour-lung interface, and air bronchogram sign were significantly different among preinvasive lesions, MIAs, and invasive adenocarcinomas (P = 0.02,P = 0.00,P = 0.048,respectively, P < 0.05). The spiculation sign was demonstrated in 52.38 % of preinvasive lesions, 51.43 % of MIAs, and 47.37 % of invasive adenocarcinomas. There were no differences among the three groups. The preinvasive lesions often presented as a smooth margin(70.00 %), while MIAs and invasive adenocarcinomas presented as a lobulated and spiculated margin (82.35 %, 90 %, respectively) (Figs. 3 and 4). Tumour-lung interface in three groups was shown to be as follows: well-defined (38.10 %) and ill-defined (61.90 %) in preinvasive lesions, well-defined (77.14 %) and ill-defined (22.86 %) in MIAs, and well-defined (86.84 %) and ill-defined (13.16 %) in invasive adenocarcinomas (Fig. 5). The air bronchogram was demonstrated in 9.52 % of preinvasive lesions, 20.00 % of MIAs, and 36.84 % of invasive adenocarcinomas (Fig. 6). Lesion shape, bubble sign and plural tag were not significantly different among the three groups (P = 0.17, P = 0.13, P = 0.66, respectively, P > 0.05) (Table 2). Some other subtypes of invasive adenocarcinoma are shown in Fig. 7.

Fig. 3
figure 3

Bar chart of lesion marginal characteristics in different pathological subtypes

Full size image
Fig. 4
figure 4

Marginal characteristics of pGGO nodules (A) AIS in a 64-year-old woman, 1.5 mm thin-section CT image showed a smooth nodule located in the right upper lobe (B) Histopathologic photomicrograph (haematoxylin-eosin stain; magnification, Å ~ 100) showed tumour cell growing restricted to alveolar walls (arrow) without stromal, vascular, and pleural invasion. (C) LPA in a 50-year-old woman, 1.5 mm thin-section CT lung window showed a lobulated and spiculated nodule in the left lower lobe. (D) Histopathologic photomicrograph (haematoxylin-eosin stain; magnification, ×100) showed tumour cell growing along alveolar and with a infiltrative growth pattern (arrow), invasion > 5 mm

Full size image
Fig. 5
figure 5

Tumour-lung interface features of nodules (Transverse lung-window, thin-section 1.5 mm) (A) AAH lesion in a 59-year-old woman, an ill-defined nodule located in the left upper lobe. (B) LPA in a 41-year-old woman, a nodule with well-defined tumour-lung interface located in the right middle lobe

Full size image
Fig. 6
figure 6

(A) CT (Transverse lung-window, thin-section 1.5 mm) showed a pGGO nodule with air bronchogram in the right lower lobe in a 49-year-old man (B) Histopathologic photomicrograph proved to be LPA with a tumour cell growing infiltrate stroma, invasion > 5 mm (haematoxylin-eosin stain; magnification, ×100)

Full size image
Table 2 CT Findings of all pGGO nodules
Full size table
Fig. 7
figure 7

Other subtypes of invasive adenocarcinoma, CT (transverse lung-window, thin-section 1.5 mm), and histopathologic photomicrograph (haematoxylin-eosin stain; magnification, B: ×40; D,F: ×100 ) (A) 1.5 mm thin-section CT scan of a 60-year-old woman showed a nodule with spiculation located in the right upper lobe. (B) Histopathology proved to be acinar predominant type, Haematoxylin-eosin stain (magnification, ×40) showed a central luminal space surrounded by tumour cells. (C) Thin-section CT of a 69-year-old man showed a nodule with spiculation and central large lucency located in the left lower lobe. (D) Histopathology proved to be a papillary predominant type, Haematoxylin-eosin stain (magnification, ×100) showed malignant cuboidal to columnar tumour cells growing on the surface of fibrovascular cores. (E) A 6 mm nodule with bubble sign located in the left upper lobe in a 59- year-old man. (F) Histopathology proved to be poorly differentiated solid predominant type with cell degeneration

Full size image

Discussion

In our study, there were 38 (40 %) invasive adenocarcinomas presented as pGGO. Honda T et al. [5] reported that based on the new classification, invasive adenocarcinoma always presented as a solid component, but Lim HJ [6] and Lee SM [7] et al., both reported invasive adenocarcinoma as pGGO nodules, and 40 % of invasive adenocarcinoma appearing as pGGO in our study correlated very well with the results of Lim’ s report. Therefore, the CT characteristics for distinguishing preinvasive lesions from invasive adenocarcinoma appearing as pGGO nodules was of the most important results of the present study. This finding may potentially help the surgeon make a better selection of management of pGGO.

Kim HK and Lee HJ, et al. [8, 9] had shown that 8 mm was the optimal cut-off value for distinguishing benign from malignant pGGOs. Nakata et al. [10] reported that pGGOs of 10 mm or more in diameter were informative signs of malignancy. As for the new adenocarcinoma classification, Lee SM [7] showed that the optimal cut-off value for discriminating preinvasive lesions from IPAs (MIA and invasive adenocarcinoma) appearing as pGGOs was 10 mm and the specificity of the cut-off value was 100 %. We found that the optimal cut-off value for discriminating preinvasive lesions from invasive lesions (MIA and invasive adenocarcinoma) with pGGOs was 10.5 mm, and the positivity was 88.73 %, negativity was 56.52 %, and the specificity was 61.90 %. There were other authors [11, 12] who had reported cases of AAH larger than 10 mm in diameter. Lee SM [7] had mentioned the reason for the cut-off value being higher than in previous studies was that previous studies included BACs as malignant GGNs, whereas BAC, now referred to as AIS, was classified as a preinvasive lesion. In our study, there were several preinvasive lesions larger than 10 mm, and also invasive lesions less than 10 mm. This was why the specificity was lower (61.90 %) than that reported in Lee SM’s report.

Our study showed that there were no significant differences in CT attenuation of pGGOs among the three groups. Previous studies had shown that CT attenuation had been suggested as an approach to differentiating different pathological types of lesions, and the higher the CT value, the more malignant the lesion was [13, 14]. Ikeda K et al .[13] reported AAH showed a one-peak pattern and AIS or MIA showed a two-peak pattern of attenuation. The difference between our study and previous studies can be explained by the fact that previous studies included pure GGO and part-solid GGO nodules, while pure GGO was the only subject in our study. There were significant differences in lesion uniformity between preinvasive lesions and MIAs or invasive adenocarcinomas in our study (P = 0.01,P < 0.05), and the lesion uniformity was correlated with lesion size (r = 0.45). Pathologically, with the tumour growing, dilated bronchiole and enlarged alveolar space were presented more [15], consequently, CT images appeared more heterogeneous. Regarding the morphologic CT features of pGGOs, there were no significant differences of shape among preinvasive lesions, MIAs, and invasive adenocarcinomas in our study, but some studies showed that a pGGO nodule with an oval appearance suggested a malignant nature and a polygonal shape indicated a benign nature [16–18], Kim HY [19] reported that no significant morphologic differences were observed in GGO nodules, so the shape of GGO was not a reliable CT sign for determining GGO natures. Our results demonstrated that the margin, tumour-lung interface, and air bronchogram were useful CT signs for indicating the histological invasiveness of pGGO nodules. A smooth margin was more common in preinvasive lesions, while MIAs and invasive adenocarcinomas appeared more likely as lobulated and spiculated nodules and with more air bronchogram signs, and, thus, our results correlated with previous studies [15, 20–22]. We also found that the more invasive extent of a lesion in pathology, the clearer was the border on CT, which may be explained histopathologically by the thickness of alveolar walls within the lesion. With the development of a lesion, the alveolar walls (including tumour cells and fibrous stroma) got thicker and thicker. Our study was limited by selection bias. We only included the pGGO nodules that underwent surgical resection, and those pGGO nodules may be considered to have the malignant tendency. This may explain why, when compared with previous reports [6, 7], our study contained a relatively greater percentage of pGGO invasive adenocarcinomas (40 %) and had four AAHs larger than 10 mm. Another limitation was the definition of lesion uniformity, which was not a widely accepted standard. The reliability of this definition needs further confirmation in the future. In conclusion, 40 % of resected pGGO lesions were invasive adenocarcinoma in our study. A value of 10.5 mm was the cut-off value for discriminating preinvasive lesions from invasive lesions. The lesion uniformity, tumour-lung interface, and the air bronchogram could indicate the invasive extent of early stage lung adenocarcinoma with pGGO.