Adenocarcinoma is the predominant type of lung cancer today, and its incidence has been reported to be increasing worldwide [1]. This rising prevalence has led to an evolution in our understanding of lung adenocarcinoma and related lesions, many of which appear as ground-glass opacities (GGOs) on chest CT.

One of the first important studies on pulmonary adenocarcinomas was published by Noguchi et al. [2] in 1995, in which they categorized small pulmonary adenocarcinomas into six groups based on histology. In their study, patients with bronchioloalveolar carcinoma (BAC) (type A) and BAC with foci of collapsed alveolar structures (type B) had a better 5-year survival rate (100%) than patients with BAC with a foci of active fibroblastic proliferation (type C, 75% survival rate) or patients with pure adenocarcinomas (types D–F, 52% survival rate). Because this classification proved to have good correlation with prognosis, in 1999 the World Health Organization (WHO) revised the classification of BAC as a noninvasive carcinoma with no evidence of stromal, vascular, or pleural invasion. Then, in the 2004 WHO classification, the term “mixed subtype adenocarcinoma” was introduced for tumors with invasion in association with a portion of the tumor having a replacement or lepidic growth pattern (BAC component).

Jang et al. [3] first reported that BACs may be seen as GGOs, which are defined as hazy increased opacities of the lung with preservation of bronchial and vascular margins [4]. Before this study, GGOs had traditionally been regarded as areas of inflammation, hemorrhage, or fibrosis, but many studies have since confirmed that the GGO portion of the nodule corresponded to the BAC component of adenocarcinomas [5–7]. With the introduction of MDCT allowing volumetric acquisition of the whole lung using thin-section CT along with the increase of CT screening for lung cancer in asymptomatic subjects, the detection of ground-glass nodules (GGNs) [5, 8, 9], which are GGOs in nodular shape, was made with increasing frequency. GGNs detected on CT have been reported to comprise a wide range of diagnoses, from premalignant lesions, such as atypical adenomatous hyperplasia, or malignant lesions, such as BAC or adenocarcinoma. In a study dealing with the malignant potential of GGNs, the Early Lung Cancer Action Project [9] reported that the malignancy rate of GGNs (34%) was higher than that of solid nodules (7%) and that the malignancy rates of mixed GGNs (part-solid nodules) and pure GGNs (nonsolid nodules) were 64% and 18%, respectively.

Individuals who have BACs on pathologic examination and who have GGNs on CT share many demographic findings; a higher percentage of women, a younger age distribution, and a higher incidence in never-smokers [10, 11]. The histologic features of BACs, which exhibit a lepidic growth pattern with proliferation of cells along the alveolar walls [12], also corresponded well with the CT features of GGOs. Thus, in this article, we review the various features of GGNs as seen on chest CT, with a special focus on their relationship with lung adenocarcinoma (Table 1).

One of the first technical considerations in CT acquisition is slice thickness. It has been reported that with thick-section CT, solid nodules or even calcified nodules may mimic GGNs because of the partial volume effect [8] (Figs. 1A and 1B). In a comparison study performed between 5- and 1-mm section CT images reconstructed from the same raw data, the assessment of nodule consistency was shown to be significantly different between the two scans [13]. In the assessment of GGNs, therefore, obtaining CT with a section thickness of 1.0–1.5 mm or less is essential.

Another consideration is tube current. According to a recent study by Funama et al. [14], a low tube current may hinder the visibility of GGNs because of excessive noise. When the detectability of GGNs at low-dose CT was evaluated with the reference obtained at a standard dose of 180 mAs, 39.5% and 25.8% of GGNs were missed at 21 and 45 mAs, respectively. Therefore, a tube current of 50 mAs or greater is recommended in the detection and follow-up of GGNs.

Pathologically, GGNs can represent focal inflammation, focal fibrosis, atypical adenomatous hyperplasia, and BAC as well as adenocarcinoma [8, 10] (Table 2). Among these, atypical adenomatous hyperplasia, which is usually less than 5 mm in diameter and accompanied by the proliferation of type 2 pneumocyte-like or Clara cell-like cells with varied cellular atypia, is classified as a precursor lesion for lung adenocarcinoma according to the WHO classification. On CT, atypical adenomatous hyperplasia appears as a well-defined round or oval pure GGN [15] (Figs. 2A and 2B) and is frequently found to be associated with adenocarcinoma, particularly BAC [5, 15].

CT features of GGNs correlate well with the histologic classification of adenocarcinoma proposed by Noguchi et al. [2], and the proportion of GGO in the GGN is significantly correlated with the BAC component of the adenocarcinoma-mixed subtype [5, 6, 16]. According to a study by Yang et al. [7], 94% of pure BACs (type A) were visualized as pure GGNs on CT (Figs. 3A and 3B), whereas 71% of BACs with some alveolar wall collapse (type B) appeared as heterogeneous nodules with low attenuation. In addition, BACs with active fibroblastic foci (type C) appeared mostly as nodules with a GGO component in the periphery (29%) (Figs. 4A and 4B) or as a solid nodule (50%). The solid component of a mixed GGN corresponded to areas of structural collapse of alveoli or fibroblastic proliferation [7].

BACs usually manifest as GGNs, whereas adenocarcinomas manifest as both solid nodules and GGNs [5] (Figs. 5A and 5B). However, not all pure GGNs are BACs histologically, and they can have a component of invasive adenocarcinoma [17, 18]. In a study that evaluated the CT findings of 59 small adenocarcinomas, 94% of pure GGNs were of type A or B (BAC in the WHO classification), and 6% were type C lesions (adenocarcinoma with mixed subtype in the WHO classification) on the basis of the Noguchi classification [7] (Table 1).

With an increase in the number of biopsies performed for GGNs, focal interstitial fibrosis is increasingly being detected as a cause. Radiologically, it appears as a persistent GGN and often shares the imaging features of neoplastic GGNs [8, 19, 20] (Figs. 6A and 6B). On histologic examination, this lesion shows focal interstitial thickening with collagen fiber deposition and type 2 pneumocyte proliferation as well as macrophage collection in the airspace [19, 20].

The clinical significance of GGNs in patients with extrapulmonary cancers was evaluated in a study by Park et al. [21]. In their report, 82.4% of patients and 67.8% of GGNs were diagnosed as malignancies, none of which were metastatic. However, because primary adenocarcinomas and BACs manifesting as GGNs typically show slow growth, when GGNs show rapid growth in patients with extrapulmonary cancers, metastasis or inflammatory lesions should be considered.

The differentiation of benign from malignant persistent GGNs can often be difficult [22] because focal interstitial fibrosis shares many morphologic features with malignant GGNs. However, the presence and proportion of a solid portion within a GGN can be an important predictor of malignancy [9, 21, 23]. In the differentiation of BAC from atypical adenomatous hyperplasia, which sometimes manifests as pure GGN, one study showed that sphericity (visually assessed roundness in this study) was significantly associated with atypical adenomatous hyperplasia, whereas air bronchograms were associated with BAC [24]. In addition, it was suggested by Nakata et al. [23] that a threshold of 10 mm could be a significant sign of malignancy, with more recent studies suggesting a threshold of 8 mm in differentiating benign from malignant lesions for pure GGNs through receiver operating characteristic analysis [25, 26]. However, some overlaps between the two entities still remain. According to the results of Lee et al. [25], when pure GGNs and mixed GGNs were assessed separately, significant predictive CT findings of malignancy were a size of > 8 mm for pure GGNs and a lobulated border for both pure and mixed GGNs.

Molecular changes in lung carcinoma include mutations within the tyrosine kinase domain of the epidermal growth factor receptor (EGFR) gene, amplifications of the EGFR genes, and activation of mutations of the K-ras gene. A novel therapeutic approach targeting EGFR, such as EGFR-tyrosine kinase inhibitors (TK inhibitors) including gefitinib and erlotinib has been extensively studied. In these studies, non–small cell lung cancer (NSCLC) patients with EFGR mutations were reported to show excellent response to EGFR-TK inhibitors, and EGFR mutations were more frequently found in patients with adenocarcinoma histology, female sex, never-smokers, and East Asian ethnicity [27, 28]. When EGFR mutation status was correlated with the types of lung adenocarcinoma according to the Noguchi classification, the mutation was preferentially found in adenocarcinoma with a replacing growth (types A–C) compared with adenocarcinomas of types D–F [27]. These results suggest that adenocarcinomas of types A–C may arise through different mechanisms. Furthermore, although some reports have indicated that EGFR mutations are associated frequently with adenocarcinomas with BAC features [27, 29, 30], another report has suggested that BAC features are not absolute determinants for EGFR mutations [31].

Because BACs are closely related with GGNs on CT, the association of GGNs on CT with EGFR and K-ras mutations has also been investigated. However, according to a recent study by Glynn et al. [32], the presence of GGO within a nodule was not shown to be associated with EGFR or K-ras mutations.

Increased EGFR gene copy number, as assessed by a fluorescence in situ hybridization (FISH) assay, on the other hand, was associated with lymph node metastasis, more advanced pathologic stage, and poor prognosis [33]. In a study in which FISH-positive was defined as a tumor showing high polysomy or gene amplification, a high maximum standardized uptake value (SUV) on ¹⁸F-FDG PET was significantly related to FISH-positive results, whereas a high proportion of GGO, small tumor diameter on CT, and well-differentiated histopathology were more frequent in FISH-negative adenocarcinomas [34]. This field is under active research, and understanding molecular biomarkers in combination with imaging findings would be essential in the management of lung adenocarcinoma.

Most inflammatory lesions and focal hemorrhage, which manifest as GGNs, disappear at short-term follow-up (Figs. 7A and 7B). At screening settings, it has been reported that 37–70% of GGNs can resolve spontaneously or after appropriate treatment over several weeks or months [35–37]. Patient and nodule characteristics of these transient GGNs include young patient age, detection of lesion at follow-up, blood eosinophilia, lesion multiplicity, large solid portion, ill-defined border, and polygonal shape [35–37].

Persistent GGNs, on the other hand, can signify a wider range of diagnoses, including focal interstitial fibrosis, atypical adenomatous hyperplasia, BAC, or adenocarcinoma [22]. The natural progression of GGNs has been shown in several serial CT studies with follow-up. The evolving features of GGNs on serial CT that suggest malignancy include an overall increase in the size of a GGN, development of a solid component within a GGN, or an increase in the solid component of a GGN (Figs. 8A and 8B). Stable size with increasing attenuation of a GGN could also be a sign of malignancy [38]. A study by Takashima et al. [39] showed that lung adenocarcinomas that initially presented as GGNs subsequently increased in size in 75% of cases, developed solid components within the nodule in 17%, and showed an increase in the solid portion in 23% of cases.

Some GGNs, which are proven to be malignancies on histologic examination, do not change in size [17, 40]. In one study involving 19 patients, the size of pure GGNs did not change in eight of the patients during a follow-up of 2 years or more, and the authors suggested that some pure GGNs would never progress to a clinical disease [40]. In another study, Kakinuma et al. [38] reported a rare finding of decreasing size with the appearance of a solid component in one BAC and one adenocarcinoma with mixed subtype. This decrease in size in the course of follow-up, however, was explained by the collapse or severe narrowing of alveolar space or fibrosis [38].

When GGNs are persistent at 3-month follow-up, GGNs larger than 10 mm and patients with a history of lung cancer are reported to be higher risk factors for GGN growth [41]. In addition, compared with pure GGNs, mixed GGNs show interval growth more frequently and usually turn out to be adenocarcinomas [17].

The multistep progression of lung adenocarcinoma in which a precursor or premalignant lesion of atypical adenomatous hyperplasia progresses to BAC followed by invasive adenocarcinoma has previously been proposed [2], and more recently, the molecular background of this stepwise progression has been suggested [31]. In this concept, EGFR mutations contribute in the early stage of lung adenocarcinoma development [33], and EGFR amplification is superimposed on the progression to invasive cancer.

The proportion of the BAC component in lung adenocarcinoma is known to be a positive prognostic factor [16]. A high non-BAC component has been strongly associated with both recurrence and poor prognosis in patients with lung adenocarcinoma [42]. It was further suggested that adenocarcinoma with greater than 90% BAC component might be considered minimally invasive [16]. As expected from the correlation between the proportion of BAC component and CT consistency, the relative proportion of GGO correlates well with recurrence, vascular invasion, nodal metastasis, and survival [6, 43–48]. When various clinical, CT, and pathologic parameters were evaluated for the determination of prognostic significance in lung adenocarcinoma, lesion size, percentage of GGO area, and tumor subtypes were significant on univariate analyses, whereas the percentage of GGO area was the only independent prognostic factor according to multivariate analyses [49].

Tumor size is considered one of the determinants in staging; however, tumor size alone is not able to indicate the invasive or noninvasive nature of lung adenocarcinomas. Indeed, it was shown that small adenocarcinomas even 1 cm or less in diameter can show lymph node metastasis. In a study by Ohta et al. [50], four of 11 patients with adenocarcinomas ≤ 1 cm in diameter showed lymph node metastasis. A better indicator for the prediction of the noninvasive nature of lung adenocarcinoma may be the GGO proportion of a tumor. When the combined effects of GGO proportion and tumor size were compared, stage IB disease with a GGO proportion ≥ 50% showed a higher 5-year relapse-free survival rate than stage IA disease with a GGO proportion of < 50% [51].

As to the question of whether GGNs detected on chest CT can be safely followed up without immediate biopsy or surgical resection, Sawada et al. [52] reported that it can. In their study, follow-up CT examinations were performed at an interval of 3–12 months for both pure and mixed GGNs until the authors had detected either lesion growth or a new solid portion. They found that such follow-up did not lead to either treatment delays or negative influence on patient outcome.

Pathology for persistent GGNs can be established either through CT-guided biopsy or surgical biopsy. The diagnostic yield for CT-guided fine-needle biopsy for GGNs was reported to be 64.5% and that for GGNs with a GGO proportion > 50% was 51.2% [53]. CT-guided core biopsy showed better results: sensitivity of 92% and specificity of 90% after seven of 50 lesions were excluded from the calculation because of loss to follow-up or nondiagnostic samples [54]. It is impossible to make an unequivocal diagnosis of BAC in small biopsy specimens because only the BAC component may be biopsied in patients with adenocarcinoma with mixed BAC. If the suspicion for malignancy of a persistent GGN is high on CT or if there is any increase in size or solid portion of GGN, CT-guided biopsy may be bypassed and surgical biopsy can be performed to obtain a pathologic diagnosis.

On the basis of a randomized trial by the Lung Cancer Study Group, the standard procedure for T1N0 NSCLC is lobectomy with mediastinal lymph node dissection [55]. For the surgical treatment of peripheral lung adenocarcinoma manifesting as a pure GGN, however, limited wedge resection without lymph node dissection rather than standard lobectomy has been suggested and conducted by many Japanese investigators. For the mean follow-up period of 2–5 years, when limited resections were performed in patients with pure or predominantly pure GGNs, no intrathoracic recurrence or distant metastasis was observed [56–58]. The sample sizes were relatively small, the patients were not randomized, and invasive cancers cannot be excluded with CT findings alone; however, the studies do suggest that the favorable prognosis of GGNs may justify limited resection.

In comparison with solid nodules, FDG PET has limited value in the evaluation of GGNs for determining nodule malignancy and staging. Kim et al. [59] reported a high false-negative rate of FDG PET for identifying BAC (Figs. 9A and 9B). In another study, Chun et al. [60] compared the maximum SUV of malignancy and inflammation that manifested as GGN on CT. Interestingly, in mixed GGNs, the maximum SUV was significantly higher in inflammatory lesions (2.00 ± 1.18) than in malignancies (1.26 ± 0.71), but in pure GGNs, both inflammation and malignancy showed an SUV less than 1.0 and did not show a difference.

Detecting GGNs is not an easy task because of their low contrast to the background lung. In a study by Li et al. [61], 69% of lung cancers missed by radiologists at screening CT were GGNs. Thus, automated computer-aided detection (CAD) schemes for GGNs have been investigated and the first CAD scheme for automated detection of GGNs was reported by Kim et al. [62] using texture features and gaussian curve fitting features. More recently, Yanagawa et al. [63] evaluated the performance of a commercial CAD system in the detection of GGNs, but unfortunately, to date, the CAD system can only play a complementary role in the detection of GGNs.

Various approaches have been made to evaluate the change of GGNs over time. One approach of particular note is computer-assisted volumetry, which has shown improved reproducibility in measuring solid nodules. When computer-aided volumetry was applied to GGNs 8 mm or larger, 95% limits of intraobserver agreement were up to 14.9–16.6% in measuring the same nodule and those of interobserver agreement were up to 23.7% [64]. The interscan variability of nodule volumetry for GGNs using a same-day repeat CT study was reported to be up to 18.9% [65]. To reflect the changes in GGNs not only in size but also in attenuation, another approach in determining the growth of GGNs has been proposed. In a phantom study by Lee et al. [66], the calculated amount of soft tissue in a nodule, [volume × (1 + mean attenuation value / 1,000)], showed a strong correlation with the reference standard amount of soft tissue. In a clinical evaluation, de Hoop et al. [67] applied a similar method of GGN mass measurement, and this method showed the smallest measurement variability thus far compared with diameter and volume measurements, meaning that the growth of a GGN can be detected earlier with mass measurements.

Because the GGO component in GGNs is one of the most important prognostic factors, various methods to measure the proportion of GGO have also been investigated through manual measurement of the diameter or area of a nodule and the solid component of a nodule [43, 45, 47, 48]. Another method of measuring GGO proportion is through the vanishing ratio or tumor shadow disappearance rate, which is defined as the ratio between GGN size measured at lung window settings and at mediastinal window settings. This ratio was shown to correlate better with 5-year relapse-free survival than other measurements using either length or area [68]. When a commercial automated program able to segment the solid area from a GGO area was compared with manual measurement methods, the measurement using software showed better interobserver agreements but worse correlation with histology on the basis of the Noguchi classification of adenocarcinoma [69].

In measuring the size and solid portion of a GGN, investigators need to consider the dimension of measurements (length, area, or volume); section thickness of the CT scan; and thresholds for lesion boundary, which can be determined by window settings on visual assessment [70]. Several reports have shown that the tumor shadow disappearance rate had better correlation with the BAC proportion or prognosis than the GGO proportion determined at lung window settings [6, 68]. This indicates that the solid portion in GGNs on CT can be determined better at mediastinal window settings. Therefore, applying an adequate threshold is necessary to produce better parameters that correlate with pathologic examination and to predict the prognosis.

GGNs frequently manifest as multifocal lesions in the lung. In a study by Vazquez et al. [16], nonsolitary cancers were seen in 49 (18%) of 279 resected adenocarcinomas. These synchronous multifocal lung GGNs may be due to either intrapulmonary metastasis or synchronous primary cancers and raise a staging dilemma. The problem also extends to whether systemic chemotherapy or surgical resection should be administered. Furthermore, when a pure GGN is present in patients with otherwise operable NSCLC, controversy exists as to whether these lesions should be resected along with the primary tumor. Several studies have attempted to answer these questions and found that most of the small node-negative multiple carcinomas probably represent multiple primaries rather than intrapulmonary metastasis [16, 26]. Thus, they found that multifocal lesions may be amenable to surgical resection with prolonged survival and that their presence in resected cases did not impact the prognosis [16]. One study by Kim et al. [26] further suggested that pure GGNs ≥ 8 mm should be resected to rule out the possibility of malignancy, whereas pure GGNs < 8 mm could be closely followed using CT.

When clinical, pathologic, and thin-section CT features were compared between multiple GGNs and solitary GGNs [71], atypical adenomatous hyperplasias and BACs were more frequent in multiple GGNs (Fig. 10), whereas adenocarcinomas were more frequent in solitary GGNs. Those authors also found that multiple GGNs were more frequently found in women and nonsmokers and were smaller than solitary GGNs. In addition, on CT, solitary GGNs showed air bronchogram, bubble-lucency, and pleural retraction more frequently.

In a molecular investigation, Chung et al. [72] analyzed the clonal relationship between multiple GGNs using EGFR and K-ras mutational profiles. In this study, 18 of 24 patients had different clonality of multiple GGNs, suggesting that these lesions arise as independent events rather than intrapulmonary spread or metastasis. However, the multifocal GGNs discussed in this section should not be confused with multiple GGOs, which can sometimes occur in combination with airspace pattern malignancy, because such cases have a worse prognosis compared with the cases dealt with in this section.

Because most GGNs show a slow growth rate despite their high potential for malignancy, the issue of overdiagnosis, which is the diagnosis of a “disease” that will never cause symptoms or death during a patient's lifetime, remains a problem in the management of patients with GGNs. Cancers detected at baseline screening have been reported to be more slow growing than those detected on repeat examinations, and cancers presenting as GGNs were more frequent at baseline rather than repeat screening [5]. In another study, Hasegawa et al. [73] reported that the mean volume doubling time of pure GGNs was 813 ± 375 days and that of mixed GGNs was 457 ± 260 days, whereas the mean volume doubling time of solid nodules was 149 ± 125 days. Considering the standard of 400 days for overdiagnosis proposed by Yankelevitz et al. [74], 27 of 61 cancers included in the study by Hasegawa et al. can be considered as cases of overdiagnosis. In a study by Lindell et al. [75], 13 of 48 cancers, including 23 GGNs and 25 solid nodules, had a volume doubling time longer than 400 days, although nodule consistency was assessed using 5-mm-thick CT scans in that study. They also stressed that the majority of possibly overdiagnosed cancers occurred in women. In addition, all pure GGNs, which were not only BACs but also adenocarcinomas, showed a 2-year relapse-free survival [18]. Because GGNs, especially pure GGNs, have a potential for overdiagnosis, physicians should consider whether the benefit from surgical resection of malignant GGNs outweighs the potential adverse effects of surgery in older patients, especially those with comorbidity.

Several guidelines have been suggested in the management of GGNs on the basis of CT findings. Recently, Godoy et al. [76] proposed interim guidelines in terms of size and nodule consistency, which can be summarized as follows: Isolated pure GGNs smaller than 5 mm in size do not need follow-up CT studies. Pure GGNs between 5 and 10 mm in size require initial follow-up in 3–6 months to confirm persistence, with continued long-term follow-up for persistent GGNs. Persistent pure GGNs 10 mm or larger in size, any mixed GGNs, and growing GGNs should be resected. Finally, multiple GGNs can be managed through limited surgical resection for dominant lesions.

In another study, Asamura [11] recommended a minimally invasive approach for lung adenocarcinomas that manifested as GGNs: Pure GGNs with a diameter of less than 15 mm can be monitored carefully without intervention, with overt growth or an overt solid portion being an indication for surgical intervention; and pure GGNs larger than 15 mm in diameter or mixed GGNs less than 15 mm in diameter should be managed through wide resection or segmentectomy provided they are located in the outer one third of the lung parenchyma or through standard lobectomy if tumors show overt invasive growth or predominantly solid GGNs. On the basis of previous guidelines and literature regarding the management of incidentally detected GGNs, recommendations for follow-up and intervention of GGNs are presented in Appendix 1.

Conventionally, lung cancer is closely related to smoking and occurs more commonly in older men. However, there appears to be another type of distinctive lung cancer with a relatively slow growth rate that can be characterized clinically as having a more common occurrence in women and nonsmokers. Pathologically, it is often accompanied by a BAC component. GGNs on CT share these characteristics, and thus can serve as imaging biomarkers that represent the BAC component of adenocarcinoma on histology and indicate a better prognosis of lung adenocarcinoma. Yet, there are still discrepancies between nodule consistency on CT and the histologic subtype of adenocarcinoma described in the literature, and therefore prognosis cannot be determined at present with CT features alone. Recent research has indicated that molecular biomarkers, such as EGFR mutations, are closely related to these cancers and newly developed targeted therapy may be able to use this information for personalized treatment. To that end, the molecular background of the multistep progression of adenocarcinoma is under active research today.

In the clinical setting, despite the high probability of malignancy of GGNs, the issue of overdiagnosis should be considered in the management of GGNs. Thus, in addition to size and nodule consistency, growth rate needs to be considered for surgical treatment especially in elderly individuals with comorbid conditions. In the near future, robust quantifying schemes, providing objective standards able to detect the change in size of GGNs and the proportion of the solid portion of GGNs must be developed because pure GGNs as well as mixed GGNs with a predominant GGO component have been shown to provide better prognosis than lung cancer manifesting as solid nodules.

Address correspondence to J. M. Goo ([email protected]).