Several studies have analyzed relationships between CT features of pGGNs and pathological type [15–18]. To the best of our knowledge however, no published study has identified predictors of lung IACs in pGGNs by analyzing their clinical and imaging characteristics. In the present study 659 adenocarcinomas manifesting as pGGNs on CT were investigated, making the study larger than most recent studies investigating pathological classification of pGGNs [11, 15, 17, 19–20].
Kitami et al.  reported cutoffs of − 600 HU for CT density and 10 mm for maximum diameter for distinguishing between IAC and non-IAC pGGNs of the lung. The corresponding cutoffs identified in the current study are similar to those values, and other previously reported values [11, 15]. Because integers are more convenient, we used published cutoff values in our univariable and multivariable analyses (age > 55 years, maximal diameter > 1 cm, CT density > − 600 HU). In the present study, model 2 suggested that maximal and mean diameters could reliably and independently discriminate IACs in patients with pGGNs with diameters of < 3 cm. Lim et al.  reported that pGGN size was a significant indicator when differentiating IACs from AISs and MIAs when the nodules had diameters > 10 mm. Kitami et al.  identified a maximal diameter cutoff of 10 mm for distinguishing between lung IACs and non-IACs presenting as pGGNs. Although in the current study a tumor diameter of > 10 mm was a significant independent predictor of IAC, that result is not readily apparent from the scatter plot shown in Fig. 3. Perhaps the best predictor of IAC may be a diameter of > 20 mm, because only one of the pre-IACs shown in Fig. 3 had a diameter > 20 mm. However, although that threshold would yield a high positive predictive value the corresponding negative predictive value would be abysmal. This would result in a high rate of missed diagnoses in clinical practice. Notably, the observation in the present study that a tumor diameter of > 10 mm was an independent predictor of IAC is consistent with results reported in similar previous studies [17–18, 21–23]. As depicted in the scatter plot of pGGN diameters determined via CT by pathological type shown in Fig. 3, in the current study the more aggressive the histological subtype was the larger the pGGN’s diameter tended to be. Thus, tumor size is the most important variable on which to base decisions pertaining to the management of pGGNs [12, 24–25].
Although nodules > 10 mm in diameter are more likely to be IACs, smaller nodules may also be IACs. In the present study 29 nodules with diameters of < 10 mm were found to be IACs. Thus, nodules with diameters of < 10 mm must also be managed and followed up diligently to ensure that malignant changes do not go undetected. We suggest that it is appropriate to monitor nodules with diameters < 10 mm independently. With regard to these smaller nodules, further research is needed to establish a basis for clinical planning.
Although CT-determined density of pGGNs of > − 600 HU was a significant predictor of IAC in the present study, the scatter plot shown in Fig. 4 does not clearly depict this result. We think that using a single factor to predict IACs is not good clinical practice, and the assessment of multiple factors simultaneously would be more accurate. Notably, Kitami et al.  reported that the CTdetermined density of pGGNs can distinguish IACs, and Lim et al.  reported that pGGN density was a significant predictor of tumor invasiveness. In contrast, Heidinger et al.  reported that pGGN density was not significantly associated with pathological diagnosis, and several other groups have also reported no significant differences in nodule density between AISs, MIAs, and IACs manifesting as pGGNs on CT [20–21]. Thus, whether CT density is a valid parameter for distinguishing IACs remains controversial. CT density should be combined with other indicators such as size, patient age, and certain CT signs when predicting the nature of a lesion preoperatively.
Liu et al.  reported that the presence of signs of spiculation is suggestive of a diagnosis of IAC. In univariable analysis in the present study spiculation was a significant predictor of IAC; however, this was not confirmed in multivariable analysis. Spiculation is considered to be evidence of malignancy and to represent invasiveness. In one study spiculation was the strongest predictor of invasion .
In the present study age was a significant predictor of IAC. To the best of our knowledge no previous studies have identified this correlation. This hitherto unreported result of the current study warrants further investigation in prospective studies. In clinical practice, whether surgical resection is recommended in patients with pGGN who are aged > 55 years is determined with reference to a combination of other additional factors such as nodule diameter and nodule density. IACS presenting as pGGNs have a good prognosis. In the present study, no pGGN IACs recurred after surgical resection during a median follow-up period of 41 months (range 7–52 months), a finding that is consistent with some previous research [27–29] .
It is well established that IACs and pre-IACs require treatment via different surgical procedures. At our hospital surgery is performed if the diameter of the pGGN is > 8 mm. Most patients are treated via limited resection and frozen section, and lobectomy is performed if an invasive component is detected via frozen section. Unfortunately, the accuracy of frozen section is not satisfactory . Moreover, the distinction between MIA and IAC is still difficult even if invasive components are observed. Because the accuracy of intraoperative frozen sections is still unclear, it is easier to determine the appropriate treatment if IAC can be predicted accurately. Although limited resection is the preferred treatment for pGGN, it may not be possible to remove all lesions. In the present study 20.6% of pGGNs were eventually diagnosed as IACs, therefore the presence of a pGGN should not be used as an indication for limited resection.
The current study had some limitations. One is that it was retrospective, and another is that all the data were derived from a single institution. All the data in the study were derived from 2016 however, and all patients were managed in accordance with the same protocol; thus, there was conceivably relatively minimal bias. Another potential limitation is that only patients in whom a diagnosis had been established via resection were included, but some of the unresected pGGNs may also have been adenocarcinomas. pGGNs can have non-cancerous origins, most notably infection, among others. These factors may have contributed to a selection bias in the present study. A prospective study would be required to minimize these potential sources of bias. An additional study limitation is that distinguishing between pGGNs and other nodules is subjective. Two reviewers (a radiologist and a surgeon) evaluated all CT scans independently to minimize this source of bias. Lastly, it is not always possible to completely exclude all blood vessels and bronchioles when delineating the boundaries of lesions, and this may have contributed to variations between observers in nodule measurements and the characterization of lesions.
In conclusion, patients with pGGNs < 3 cm in diameter on CT are more likely to have IACs if they are aged > 55 years, exhibit a nodule diameter > 1 cm, and the CT-determined density of the nodule is > − 600 HU. We suggest that surgical treatment and lobectomy is preferable to limited resection in these patients. Postoperatively, IACs initially identified as pGGNs have a good prognosis, and there were no recurrences during a median follow-up period of 41 months in the current study. The results of this study may assist decisions pertaining to the selection of surgical procedures in patients with pGGNs identified as being at high risk of malignant disease.