Factors predicting organ-specific distant metastasis in patients with completely resected lung adenocarcinoma

The aim of the study is to demonstrate the relationship between clinicopathological variables and organ sites of metastasis in resected lung adenocarcinoma. The clinicopathological characteristics of 748 patients of resected lung adenocarcinoma at Taipei Veterans General Hospital between 2004 and 2012 were retrospectively reviewed. The prognostic value of clinicopathological variables for specific organ site metastasis-free survival was demonstrated. Among the 182 patients with distant metastasis, 93 (51.1%) patients developed contralateral lung metastasis, 81 (44.5%) had brain metastasis, 71 (39.0%) had bone metastasis, and 18 (8.9%) had liver metastasis during follow-up. Acinar predominant (Hazard ratio [HR], 0.468; 95% confidence interval [CI]: 0.250 to 0.877; P = 0.018) was significantly associated with less contralateral lung metastasis in multivariate analysis. Micropapillary predominant (HR, 2.686; 95% CI, 1.270 to 5.683; P = 0.010) was significantly associated with brain metastasis. Acinar predominant (HR, 0.461; 95% CI, 0.216 to 0.986; P = 0.046) was a significant prognostic factor for better contralateral lung metastasis-free survival in multivariate analysis. Micropapillary predominant (HR, 2.186; 95% CI, 1.148 to 4.163; P = 0.017) and solid predominant (HR, 4.093; 95% CI, 1.340 to 12.504; P = 0.013) were significant prognostic factors for worse brain metastasis-free survival and liver metastasis free-survival, respectively. There are significant differences in metastatic behavior between predominant pathological subtypes of lung adenocarcinoma. This information is important for patient follow-up strategy and identification of organ-specific distant metastasis. Prospective multi-institutional studies are mandatory for further validation.

Respiratory Society (ERS) proposed a new classification system of lung adenocarcinoma [17]. They recommended the use of comprehensive histological subtyping to assess histologic patterns semiquantitatively in 5% increments to choose a single predominant pattern (lepidic, acinar, papillary, micropapillary or solid) for invasive adenocarcinomas [17]. The significant prognostic value of the new classification on death and recurrence in lung adenocarcinoma has been reported and validated in many studies [18][19][20][21].
In our previous studies [22,23], we have demonstrated that patients with micropapillary/solid predominant lung adenocarcinoma had significantly worse prognosis. We have also demonstrated that patients with micropapillary or solid predominant adenocarcinoma had a significantly higher possibility of developing initial extrathoracic-only recurrence than other subtypes [23]. However, whether the new classification predicts organspecific metastasis of resected lung adenocarcinoma has not been investigated and reported in the literature. The study aims to demonstrate the predictors of organ-specific metastasis, including the new lung adenocarcinoma classification, in patients with completely resected lung adenocarcinoma with distant metastasis.

Association between organ sites of distant metastasis and clinicopathological variables
We first performed χ 2 test and the paired independent sample t-test to investigate the relationship between specific organ sites of distant metastasis and clinicopathological variables ( Table 2). Acinar subtype predominant (P = 0.005) was associated with less contralateral lung metastasis. Micropapillary (P = 0.002) and solid (P = 0.018) subtype predominant were associated with more contralateral lung metastasis. Micropapillary (P < 0.001) and solid (P = 0.024) subtype predominant were associated with brain metastasis. Micropapillary (P = 0.012) and solid (P < 0.001) subtype predominant were also associated with bone metastasis. Solid subtype predominant (P < 0.001) was associated with liver metastasis.

Application of the number of risk factors in predicting organ-specific metastasis in patients with stage I lung adenocarcinoma
To examine their cumulative predictive value on brain metastasis in stage I lung adenocarcinoma, micropapillary subtype predominant and angiolymphatic invasion were used as risk predictors for brain metastasis. All stage I patients (n = 556) were divided into two groups according to number of positive risk predictors: none positive or one positive (group 1) (n = 541), and two positive (group 2) (n = 15). Kaplan-Meier analysis showed that patients with both micropapillary subtype predominant and angiolymphatic invasion had significantly worse BrMFS than others (P < 0.001) (Figure 2A).
To examine their cumulative predictive value on liver metastasis in stage I lung adenocarcinoma, solid subtype predominant and angiolymphatic invasion were used as risk predictors for liver metastasis. All stage I patients (n = 556) were divided into two groups according to number of positive risk predictors: none positive or one positive (group 1) (n = 540), and two positive (group 2) (n = 16). Kaplan-Meier analysis showed that patients with both solid subtype predominant and angiolymphatic invasion had significantly worse LMFS than others (P < 0.001) ( Figure 2B).

logistic regression analysis for organ sites of distant metastasis in the validation cohort
The median follow-up time for all the 438 patients in the validation cohort was 23.5 months (range, 0.1 to 37.8 months). For all these patients, the 3-year overall survival and disease-free survival rates were 96.0% and 94.0%, respectively. Twenty-two (5.0%) of the 438 patients developed distant metastasis during follow-up. Among the 22 patients with distant metastasis, there were 6 (27.3%) patients with acinar predominant, 2 (9.1%) with papillary predominant, 4 (18.2%) with micropapillary predominant, and 10 (45.5%) with solid predominant adenocarcinoma. The median time to recurrence for the 22 patients was 12.0 months (range, 3.3 ± 29.3 months). Among the 22 patients, 9 (40.9%) patients developed contralateral lung metastasis, 7 (31.8%) had brain metastasis, 8 (36.4%) had bone metastasis, and 3 (13.6%) had liver metastasis during follow-up. Eleven (50.0%) of the 22 patients had local recurrence.
We first performed univariate logistic regression analyses for specific organ sites of distant metastasis in the validation cohort (Supplementary Table 1). The backward stepwise regression procedure was used for multivariate analysis (Supplementary Table 2). N2 (vs. N0 or N1) (P < 0.001) was significantly associated with more contralateral lung metastasis in multivariate analysis. N2 (vs. N0 or N1) (P = 0.003) and micropapillary subtype predominant (HR, 8.035; 95% CI, 1.025 to 63.005; P = 0.047) were significantly associated with more brain metastasis. Stage II or III (vs. stage I) (P = 0.014) was significantly associated with more bone metastasis. Solid subtype predominant (P = 0.025) was significantly associated with more liver metastasis in univariate analysis. Multivariate analysis was not performed for liver metastasis due to the small number of patients with liver metastasis (n = 3).

DIscUssION
This study demonstrated that pathological subtypes of lung adenocarcinoma are associated with organ-specific metastasis in patients of resected lung adenocarcinoma with distant metastasis. Acinar, micropapillary, and solid predominant adenocarcinomas are significantly associated with contralateral lung metastasis, brain metastasis, and liver metastasis, respectively. Acinar, micropapillary and solid predominant adenocarcinomas are also significant prognostic factors for CLMFS, BrMFS and LMFS, respectively. By combination of risk factors, stage I patients with micropapillary subtype predominant and angiolymphatic invasion have significant worse BrMFS. Those with solid subtype predominant and angiolymphatic invasion have significant worse LMFS.
The new classification of lung adenocarcinoma proposed by IASLC/ATS/ERS in 2011 was a significant prognostic factor for survival and recurrence in lung adenocarcinoma [18][19][20][21]. In our previous study [23], we have demonstrated that patients with micropapillary or solid predominant adenocarcinoma had a significantly higher possibility of developing initial extrathoraciconly recurrence than other subtypes. While breast cancer molecular subtypes have been reported to predispose the site of distant metastases [12][13][14], whether the new classification of lung adenocarcinoma predicts organ-specific metastasis in completely resected lung adenocarcinoma remains unknown. In the current study, we showed that the new classification of lung adenocarcinoma is significantly associated with organ-specific metastasis in patients of resected lung adenocarcinoma with distant metastasis. Furthermore, we demonstrated the prognostic significance of the new classification in specific organ metastasis-free survival. Our study is the first to demonstrate the prognostic value of the new classification of lung adenocarcinoma for organ-specific metastasis in the literature.
The lung, brain and bone are the most common organ sites of metastasis in resected NSCLC [3-5, 15, 16]. In our previous study [5], bone was the most common site of single organ metastasis in patients with resected stage I NSCLC, followed by the brain. In the current study of resected stage I-III lung adenocarcinoma, contralateral lung was the most common site of metastasis, followed by the brain and the bone. We have demonstrated that acinar subtype predominant was associated with less contralateral lung metastasis, and was also a significant prognostic factor for better CLMFS. Approximately 10-25% of lung cancer patients have brain metastases at initial diagnosis [9]. About 40-50% of patients with lung cancer will develop brain metastases during the course of the disease [24,25]. The brain is also one of the most common organ sites of metastasis in patients undergoing completely resected lung cancer [3-5, 12, 13]. Many biomarkers have been reported to be predictive for the development of brain metastases from lung cancer [9][10][11]. Several studies have demonstrated that brain metastases would be more frequent in patients with tumors harboring epidermal growth factor receptor (EGFR) mutations [26][27][28]. However, the numbers of patients were small in these studies, and were far too limited to make any firm conclusions. In our study, micropapillary predominant adenocarcinoma was significantly associated with more brain metastasis, and was a significant prognostic factor for worse BrMFS. For stage I lung adenocarcinoma, patients with both micropapillary subtype predominant and angiolymphatic invasion had significant worse BrMFS. Since our study is the first to demonstrate the association of the new classification of lung adenocarcinoma and organ-specific metastasis in the literature, we have . Patients with both micropapillary subtype predominant and angiolymphatic invasion had significantly worse brain metastasis-free survival than others (P < 0.001). (b) For liver metastasis, the patients were divided into two groups according to number of positive risk predictors (solid subtype predominant and angiolymphatic invasion): none positive or one positive (group 1) (n = 540), and two positive (group 2) (n = 16). Patients with both solid subtype predominant and angiolymphatic invasion had significantly worse liver metastasis-free survival than others (P < 0.001).
conducted a validation cohort for confirmation. Although the number of patients with metastasis is small and the follow-up time was short in the validation cohort, micropapillary predominant adenocarcinoma was still significantly associated with more brain metastasis. The information is important and helpful to identify patients of resected lung adenocarcinoma who are at higher risk developing brain metastases because brain metastasis without neurologic symptoms may be difficult to be early diagnosed.
Although liver was not a common organ site of metastasis in lung cancer, liver metastasis was not rare during follow-up after surgery. Approximately 9% of patients with distant metastasis developed liver metastasis in our study. The predictor of liver metastasis from lung cancer has not been reported. In the study, we have demonstrated that solid predominant adenocarcinoma tended to be significantly associated with a higher frequency of liver metastasis. Solid predominant adenocarcinoma was a significant prognostic factor for worse LMFS. For stage I lung adenocarcinoma, patients with both solid subtype predominant and angiolymphatic invasion had significantly worse LMFS. In the validation cohort, solid subtype predominant adenocarcinoma was significantly associated with more liver metastasis in univariate analysis. Although multivariate analysis was not performed due to the small number of patients with liver metastasis, the association between solid predominant adenocarcinoma and liver metastasis was worthy of further investigation in the future. The results are important because patients with solid predominant adenocarcinoma may undergo more frequently liver imaging study during follow-up for early diagnosis of liver metastasis.
The associations between age, sex, tumor size, N status, tumor stage, visceral pleural invasion, angiolymphatic invasion, or adjuvant therapy and organ site of metastasis from lung cancer have not been well demonstrated. In the current study, female is a significant prognostic factor for better BoMFS. Greater tumor size was associated with more contralateral lung, brain, and bone metastasis. Greater tumor size was also a significant prognostic factor for worse CLMFS and BoMFS. N2 status (vs. N0 or N1) was associated with more contralateral lung and bone metastasis. Stage II or III (vs. stage I) was associated with more contralateral lung and brain metastasis. Angiolymphatic invasion was associated with more brain, bone, and liver metastasis. Angiolymphatic invasion was also a significant prognostic factor for worse BrMFS and BoMFS. No significant association between other conventional clinicopathological variables and organ sites of metastasis was identified.
There are some limitations and biases of this study that should be mentioned. As a retrospective single institute study, patient selection bias and time trend bias were inevitable. Another limitation is the diagnostic bias that conventional imaging would not necessarily detect all metastatic disease. Subclinical metastases may be missed simply because imaging not performed. There was also bias in defining a new primary lung cancer from a recurrent NSCLC. The major strength of this study is that a full and detailed follow-up of organ-specific metastasis made the analyses for specific organ site metastasisfree survival possible. Furthermore, we have conducted a validation cohort for internal validation. However, prospective multi-institutional studies are mandatory to further validate the prognostic value of the predominant pathological subtypes of lung adenocarcinoma on organspecific metastasis after surgical resection.
In conclusion, this study demonstrates significant differences in metastatic behavior between predominant pathological subtypes of lung adenocarcinoma. Acinar predominant adenocarcinoma is associated with less contralateral lung metastasis. Micropapillary and solid predominant adenocarcinomas are significantly associated with brain and bone metastasis, respectively. This information is important for patient follow-up strategy and further study of molecular mechanisms leading to organspecific metastasis in lung adenocarcinoma.

MAterIAls AND MetHODs
This study has been approved by the Institutional Review Board of Taipei Veterans General Hospital. From January 2004 to December 2012, all patients underwent completely resection for lung adenocarcinomas at Taipei Veterans General Hospital were retrospectively reviewed. Patients undergoing neoadjuvant chemotherapy or with stage IV disease were excluded. Patients with incomplete clinical information and follow-up were also excluded. A total of 801 patients were eligible for the study. Among them, 566 (70.7%) patients were free of tumor recurrence and 235 (29.3%) patients developed recurrence during follow-up. Only two patients with recurrence were diagnosed as lepidic predominant adenocarcinoma. The number of patients of lepidic predominant adenocarcinoma developing recurrence was small as compared with the other four subtypes of adenocarcinoma. Therefore, the 2 patients with lepidic predominant adenocarcinoma were excluded. Fifty-one (21.9%) of the remained 233 patients developed local only recurrence during follow-up, and was excluded for analysis. The remained 182 patients developed distant metastasis during follow-up. The 182 patients with distant metastasis and the 566 patients without tumor recurrence were included for analysis in the current study. The preoperative staging work-up was routinely performed as previously described [22,23]. Mediastinoscopy was performed only when enlarged mediastinal lymph nodes (diameter > 1.0 cm) were shown by computed tomography scan. Complete resection of lung cancer and mediastinal lymph nodes dissection/ sampling were performed as previously described [22,23]. Determination of disease stages was based on the TNM classification (seventh edition) of the American Joint Committee on Cancer and the International Union Against Cancer [29,30].
All resected specimens were formalin fixed and stained with hematoxylin and eosin, and were evaluated microscopically as previously described [23]. Each tumor was reviewed using comprehensive histological subtyping, recording the percentage of each histologic component (lepidic, acinar, papillary, micropapillary, and solid) in 5% increments as previously described [23]. The predominant pattern is defined according to the most dominant pattern.
All patients were followed-up at our outpatient department quarterly in the first 2 years after resection and semi-annually thereafter. The modalities and protocols during follow-up were used as previously described [5,22,23]. Computed tomography scans of chest and upper abdomen were routinely done in every scheduled outpatient department visit for follow-up. Nuclear medicine survey of the bone was arranged every 6 months in the first 2 years after resection and annually thereafter during follow-up. Suspicious bony lesions were confirmed by x-ray or bone biopsy. Computed tomography scan of brain was done when neurological symptoms occurred or when clinical suspicions were raised. Once a metastasis was discovered, a routine investigation was arranged to look for other metastatic sites. After initial diagnosis of recurrence, further examinations were arranged to discover other metastatic sites if symptoms occurred or clinical suspicions were raised. The hospital charts of all patients were reviewed to collect data of patterns of recurrence, organ sites of recurrence, and treatment for recurrence. Data collected from telephone call and correspondence letters during follow-up were also included.
To investigate their impact on specific organ site metastasis-free survival, clinicopathologic factors were examined in univariate and multivariate analyses. Local recurrence was defined as tumor recurrence in contiguous anatomical sites, including the ipsilateral hemithorax and mediastinum after surgical resection. Distant metastasis was defined as tumor recurrence in the contralateral lung or outside the hemithorax and mediastinum after surgical resection. Local only recurrence was defined as only local recurrence identified from initial operation to death or last follow-up. Distant only metastasis was defined as only distant metastasis discovered from initial operation to death or last follow-up. Secondary primary lung cancer was differentiated from recurrent NSCLC in patients undergoing surgical resection or biopsy according to the criteria proposed by Girard et al [31]. For those not undergoing resection or biopsy, judgment was made according to clinical course, eg. progression or aggressive clinical behavior (multiple lesions). The length of specific organ site (contralateral lung, brain, bone, or liver) metastasis-free survival was defined as the interval between the date of surgical resection and the date of the specific organ site metastasis (contralateral lung, brain, bone, or liver, respectively) or the last follow-up. An observation was censored at the last follow-up session when the patient was alive with specific organ metastasisfree status, or had died without specific organ metastasis.
To confirm the findings of the study, we conducted a validation cohort for further verification. From January 2013 to December 2014, all patients underwent completely resection for lung adenocarcinomas at Taipei Veterans General Hospital were retrospectively reviewed. The same exclusion criteria used in the original cohort were applied in the validation cohort. A total of 438 patients were eligible. Among them, 416 (95.0%) patients were free of tumor recurrence and 22 (5.0%) patients developed distant metastasis during follow-up.
The specific organ site metastasis-free survival was calculated by the Kaplan-Meier method [32]. The log-rank test was used to make group comparisons. To compare between groups with respect to categorical and continuous variables, the χ 2 test and the paired independent sample t-test were used as appropriate. To investigate their association with specific organ sites of metastasis, clinicopathological factors were analyzed in univariate and multivariate logistic regression. For specific organ site metastasis-free survival, univariate and multivariate analyses were performed by means of the Cox proportional hazards model using SPSS software (version 20; IBM, Armonk, New York, USA). All variables with P < 0.1 in univariate analysis were entered into multivariate analysis. Age and sex were also entered for mutual adjustment despite P > 0.1. Statistical significance was defined as P < 0.05.