CT features and disease spread patterns in ROS1-rearranged lung adenocarcinomas: comparison with those of EGFR-mutant or ALK-rearranged lung adenocarcinomas

The purpose of this study was to investigate the differences in CT characteristics and disease spread patterns between ROS1-rearranged adenocarcinomas and epidermal growth factor receptor (EGFR)-mutant or anaplastic lymphoma kinase (ALK)-rearranged adenocarcinomas. Patients with stage IIIb/IV adenocarcinoma with ROS1 rearrangement, EGFR mutations, or ALK rearrangement were retrospectively identified. Two radiologists evaluated CT features and disease spread patterns. A multivariable logistic regression model was applied to determine the clinical and CT characteristics that can discriminate between ROS1-rearranged and EGFR-mutant or ALK-rearranged adenocarcinomas. A cohort of 169 patients was identified (ROS1 = 23, EGFR = 120, and ALK = 26). Compared to EGFR-mutant adenocarcinomas, ROS1-rearranged adenocarcinomas were less likely to have air-bronchogram (p = 0.011) and pleural retraction (p = 0.048) and more likely to have pleural effusion (p = 0.025), pericardial metastases (p < 0.001), intrathoracic and extrathoracic nodal metastases (p = 0.047 and 0.023, respectively), and brain metastases (p = 0.017). Following multivariable analysis, age (OR = 1.06; 95% CI: 1.01, 1.12; p = 0.024), pericardial metastases (OR = 10.50; 95% CI: 2.10, 52.60; p = 0.005), and nodal metastases (OR = 8.55; 95% CI: 1.14, 62.52; p = 0.037) were found to be more common in ROS1-rearranged tumors than in non-ROS1-rearranged tumors. ROS1-rearranged adenocarcinomas appeared as solid tumors and were associated with young age, pericardial metastases and advanced nodal metastases relative to tumors with EGFR mutations or ALK rearrangement.


Scientific RepoRtS
| (2020) 10:16251 | https://doi.org/10.1038/s41598-020-73533-y www.nature.com/scientificreports/ demonstrated an overall response rate of 72% as well as a median progression-free survival of 19.2 months in patients with ROS1-rearranged lung cancer 7 and was approved as front-line therapy for ROS1-rearranged NSCLC in 2016. The most recent National Comprehensive Cancer Network (NCCN) guidelines recommend genetic testing in all patients with advanced NSCLC before initial treatment 8 . However, molecular testing may not be feasible because of insufficient tissue samples from small biopsies or be inaccurate owing to intra-and intertumoral heterogeneity 9,10 . In addition, rebiopsy for genomic evaluation during treatment may not be feasible in some patients with advanced disease. Recent studies have shown that imaging features suggest certain molecular alterations in NSCLC, such as EGFR mutations and ALK rearrangement [11][12][13] . However, to date, limited studies have evaluated the imaging features of ROS1-rearranged lung cancer 14,15 . Therefore, the purpose of our study was to investigate the differences in CT characteristics and disease spread patterns between patients with lung adenocarcinoma who have ROS1 rearrangement and those with EGFR mutations or ALK rearrangement. Results patient characteristics. Twenty-three patients who had lung adenocarcinoma with ROS1 rearrangement [5 men and 18 women; mean age of 56 years (range of 31-76 years)] were identified. For the control groups, 120 patients with EGFR-mutant lung adenocarcinoma [40 men and 80 women; mean age of 62 years (range of 28-83 years)] were randomly chosen based on the prevalence of genetic mutations in the lung cancer population study 16 . 26 patients with ALK-rearranged adenocarcinoma [9 men and 17 women; mean age of 56 years (range of 30-83 years)] were also included in this study. The mean age of the 169 patients was 59.4 years (range of 28-83 years). Clinicopathologic characteristics of these patients are summarized in Table 1 correlation between the predictors of ROS1-rearranged tumors and response to crizotinib. Among

Discussion
Our study showed that in a cohort of patients with advanced adenocarcinomas, patients with ROS1-rearranged tumors exhibit characteristic clinical and radiologic features compared to those with EGFR mutations or ALK rearrangement. As per our findings, it is proposed that young age and disease spread patterns, including pericardial metastasis and nodal metastasis, are important predictors of ROS1-rearranged tumors. Current guidelines recommend that all patients with adenocarcinomas be tested for routine biomarkers, including EGFR mutations, ALK rearrangement, and ROS1 rearrangement, because FDA-approved agents for lung cancer are available for these biomarkers 8,18 . However, in clinical practice, molecular testing in patients with advanced lung cancer may not always be feasible for several reasons, such as nondiagnostic or inconclusive results from small biopsy specimens, inconsistency among the various molecular tests, or intra-and intertumoral heterogeneity of genetic mutations [19][20][21] . Recent studies have demonstrated that imaging features may suggest certain molecular alterations in NSCLC, such as EGFR mutations and ALK rearrangement [11][12][13] . Therefore, these specific imaging features combined with clinical features may help identify patients who could benefit from expedited testing for genetic mutations or rebiopsy after nondiagnostic results.
In our study, CT features of the primary tumor among the three genotypes showed substantial overlap. In terms of lesion density, ROS1-rearranged tumors were mainly solid, similar to EGFR-mutant or ALK-rearranged tumors. Previous studies have reported that ALK-rearranged tumors typically appear as a solid lesion with a lobulated contour and hypoattenuation on a contrast-enhanced CT representing histologic features, such as abundant intra-or extracellular mucin and a solid signet-ring cell pattern 22,23 . Notably, the majority of EGFRmutant tumors in our cohort were also solid in density, although several studies suggested a close association between EGFR-mutant tumors and the presence of ground-glass opacity components 11,12 . These findings suggest that the density of the tumor may not be unique across the genetic mutations, especially in advanced adenocarcinomas. ROS1-rearranged tumors were less likely to have air-bronchogram and pleural retraction, which are well-recognized imaging features favoring EGFR-mutant adenocarcinomas 12,24 . It is also noteworthy that no significant difference was observed in imaging features of the primary tumor between ROS1-and ALKrearranged tumors, which may be attributed to the fact that these tumors have substantial similarities in both clinical attributes and response to crizotinib therapy. www.nature.com/scientificreports/ With regard to imaging features other than the primary tumor, ROS1-rearranged tumors more frequently showed advanced intra-and extrathoracic lymph node metastases, pleural effusion, and pericardial metastases compared to EGFR-mutant tumors, although these differences were not observed between ROS1-and ALKrearranged tumors. This tendency toward lymphangitic spread of ROS1-rearranged tumors, such as advanced lymphadenopathy and pericardial metastases, is also similar to that of ALK-rearranged tumors, which has been reported in a previous study comparing ALK-rearranged and EGFR-mutant advanced adenocarcinomas 11 . In addition, ROS1-rearranged tumors were less likely to be associated with brain metastases compared with other mutation groups, which is corroborated by previous studies 25,26 . The mechanism of lower incidence of brain metastasis in ROS1-rearranged tumors compared with other mutations groups is not yet fully understood but might be partly explained by the propensity for lymphangitic tumor spread rather than hematogenous spread.
Given the results of our study, clinical and imaging features suggest the possibility of ROS1 rearrangement and prioritize appropriate genetic testing in advanced lung cancer. This has substantial clinical implications because the prevalence of ROS1 rearrangement (1-2%) is much lower compared to that of EGFR mutations, which is known to be 20-30% in Western countries and 50-65% in East Asian countries 27 .
Our study has several limitations. First, the number of patients with ROS1-rearranged tumors was small mainly due to the overall rarity of this mutation in lung cancer. Second, our study is a retrospective study based on a single large tertiary referral center, and the findings of our study may not be generalizable. In addition, there may have been a bias in the selection of patients for our study. Additional prospective studies with a large number of patients are needed for further validation of the current results. Third, it is difficult to distinguish ROS1-from ALK-rearranged tumors with clinical and imaging features alone as these tumors have considerable overlaps in clinicoradiologic features as well as treatment regimen. Therefore, appropriate genetic testing should be guaranteed at initial diagnosis for effective personalized treatment. Finally, although our study suggested that imaging features might be helpful in distinguishing ROS1 rearrangement from other mutations, the mechanism underlying the differences still remains to be elucidated.
In summary, despite shared clinical and imaging features, ROS1-, ALK-, and EGFR-positive advanced adenocarcinomas differ in certain imaging features of the primary tumor and disease spread patterns. ROS1-rearranged adenocarcinomas are more likely to be associated with younger age and distribution of metastatic disease, including pericardial and nodal metastases.

Methods
This retrospective study was approved by the Institutional Review Board and the Ethics Committee of Samsung Medical Center. Informed consent was waived from the patient and all methods in the study were performed in accordance with the relevant guidelines and regulations.
patients and data selection. From July 2009 to June 2015, a total of 7033 patients with NSCLC underwent genetic mutation studies at our institution. We selected all patients who had advanced adenocarcinoma (stage IIIb/IV) with ROS1 rearrangement to participate in this study. For comparison with patients with ROS1 rearrangement, we also identified patients who had advanced adenocarcinoma with EGFR mutations and ALK rearrangement during the study time frame. Only the patients who satisfied the following criteria were included in this study: (1) aged 18 years or older; (2) histologically proven adenocarcinoma at clinical or pathological stage IIIb/IV; (3) positive for ROS1 rearrangement, ALK rearrangement, or EGFR mutations; (4) no history of previous treatment; and (5) available for a pretreatment chest CT study. In all patients, the histologic diagnoses were made by a pathologist (with 23 years of experience in thoracic pathology) by means of a percutaneous core needle and/or bronchoscopic biopsy. Chest CT studies were performed within one month prior to lung biopsy. Clinical and pathologic data were obtained from electronic medical records, including age at diagnosis; gender; smoking status (never smoker, ex-smoker, and current smoker); and Tumor, Node, Metastasis (TNM) staging based on the 8th edition of the TNM Classification of Malignant Tumors 28 . Tumor characteristics were evaluated by the two radiologists on the basis of a review of transverse images, including tumor size (maximum axial diameter); density (solid or subsolid); location; border (smooth, lobulated, or spiculated); and the presence or absence of calcification, air-bronchogram, and pleural retraction.
Metastatic lymphadenopathy was confirmed histologically (endobronchial ultrasound-guided lymph node aspiration biopsy) or determined by imaging studies. Lymph nodes that measured more than 10 mm in short axis diameter and/or displayed increased glucose uptake [higher than that of the surrounding tissue and with a www.nature.com/scientificreports/ maximum standardized uptake value (SUV) of more than 3.5 as determined by quantitative analysis] on PET/ CT scans were considered malignant 29 . Intrathoracic metastases were recorded as follows: intrapulmonary, pleural, pericardial, or bone. Intrapulmonary metastases were classified as miliary (< 5 mm), nodular scattered (> 5 mm), or lymphangitic carcinomatosis. Intrathoracic bone metastases were determined by a decrease in tumor size after chemotherapy or target therapy on follow-up imaging studies (5). Extrathoracic metastases were evaluated by CT of the abdomen and/or pelvis as well as a brain MRI for each patient. PET/CT scans were also reviewed for the presence of distant metastases if available.
Statistical analysis. All data were recorded as means ± standard deviations for continuous variables or frequencies and as percentages for categorical variables. To explore discriminative imaging features between the mutation groups, we used the two sample t-test, Fisher's exact test, the Wilcoxon rank sum test, and the chi-square test for univariate analysis. A multivariable logistic regression model was created with the factors that demonstrated a p-value < 0.1 in the univariate analysis. To evaluate the correlation between the predictors of ROS1-rearranged tumors and antitumor activity of crizotinib, the performance of the model in predicting a best overall response of CR or PR to crizotinib was assessed by calculating the area under the receiver operating characteristic (ROC) curve. Best overall response was derived from investigator assessment using RECIST v1.1 criteria. Statistical analyses were performed with SPSS software (version 26.0, SPSS, Chicago, IL, USA). A p-value < 0.05 was considered to indicate a significant difference.