Predictive and Prognostic Factors of Synchronous Colorectal Lung-Limited Metastasis

Aim This study is aimed at investigating predictive and prognostic factors of synchronous colorectal lung-limited metastasis (SCLLM) based on The Surveillance, Epidemiology, and End Results (SEER) database. Methods A multivariate logistic regression model was constructed to identify independent predictors of SCLLM. A multivariate Cox proportional hazards regression model was used to distinguish independent prognostic factors. Results This study enrolled 168,007 colorectal cancer (CRC) patients without metastatic diseases and 1,298 cases with SCLLM. Eight features, involving race, tumor location, pathological grade, histological type, T stage, N stage, and tumor size as well as CEA, could be used as the independent predictors. As the nomogram shown, the T4 stage contributed the most to SCLLM, followed by the N2 stage, elevated CEA, and rectal cancer. A multivariate regression analysis discriminated 9 independent prognostic factors, including age, race, marital status, pathological grade, T stage, colectomy/proctectomy, chemotherapy, CEA, and TD. The prognostic nomogram illustrated that nonresection/NOS played as the poorest prognostic factor, followed by nonchemotherapy, ≥75-year old and T4 stage. The cumulative survival curves revealed the influence of each prognostic factor on survival after controlling the other variables. Conclusions This study identified independent predictors and prognostic factors for SCLLM based on a large database of the United States. The predictors and prognostic factors can provide supporting evidence for the prevention and treatment of SCLLM.


Introduction
Colorectal cancer (CRC) ranks as the third most common malignancy in males and the second in females [1]. In spite of widespread early detection screening for CRC, approximate 25% of CRC patients are found to have distant metastases at the time of diagnosis [2,3]. Moreover, metastasis is the main cause of high mortality among CRC patients [4].
Currently, there has been a continuous increase in the number of CRC patients diagnosed with pulmonary metastases, accounting for 32.9% of all metastatic CRCs (mCRCs) [5], after the widespread use of chest CT scans in recent years. Meanwhile, some research reported that 4-9% patients with CRC suffered from synchronous lung metastasis [6][7][8].
The retrospective data from China reported that lungs being the first metastatic site reached 24.5% among patients with mCRC [9]. Nevertheless, there is limited information to guide clinical practice in colorectal lung metastasis. It is a mainstream practice that the therapeutic strategy for colorectal liver metastases is applied to lung metastasis [10][11][12]. Undoubtedly, the treatment experience from colorectal liver metastasis is conducive to the rapid development of therapeutic strategy of colorectal lung metastasis. However, some scholars believe that there are differences involving the  [13,14]. Thus, it is important to further investigate the risk factors of colorectal lung metastasis. In addition, in order to exclude the interference from other metastatic sites, this study focused on synchronous colorectal lunglimited metastasis (SCLLM), which was defined as colorectal cancer with lung-limited metastases at the time of diagnosis. SCLLM is considered less frequent due to the different metastatic route. The routine metastatic process of CRC involves discrete steps (CRC cancer cells initially migrate to the liver via the portal system, followed by the lungs and finally other locations) [15,16], while the spread of metastatic CRC to the lungs, either in isolation or as the first of several distant sites, may be attributable to venous drainage which bypasses the portal system and instead enters systemic circulation [17]. Nevertheless, the frequency of synchronous lung metastasis increased significantly by a nearly 3-folds in the past decades [15].
Due to the rareness of SCLLM, a large public database is needed to explore this issue. The Surveillance, Epidemiology, and End Results (SEER) database is a kind of populationbased cancer registration system of the USA taking 34.6% Americans into account, which can provide some necessary clinical data and be used to be an excellent database to explore issues regarding various cancers.
Therefore, this study is aimed at investigating predictive and prognostic factors of SCLLM based on SEER database.

2.1.
Patients. This retrospective analysis used data from the SEER-linked database. The SEER program of the National Cancer Institute is an authoritative source of information on cancer incidence and survival in the United States (U.S.) with updated annually. SEER currently collects and publishes cancer incidence and survival data from population-based cancer registries covering approximately 34.6% of the U.S. population [18]. Data from SEER was used to identify patients with CRC diagnosed between 2010 and 2016, and For each patient, the following data was acquired: age at diagnosis, married status, insurance, gender, race, grade, histological type, T stage, N stage, regional nodes examined (RNE), CEA, surgery for primary tumor, surgery for hepatic metastasis, tumor deposits (TD), perineural invasion (PNI),

Patient
Characteristics. This study enrolled 168,007 CRC patients without metastatic diseases and 1,298 cases with SCLLM. The entire cohort was predominantly elderly (≥65, 58.07%) and white people (75.27%). The rectum was the main site occurring lung-limited metastases in CRC. Besides, SCLLM was related to marital status, race, pathological grade, and histological type. Meanwhile, there were significant differences regarding the depth of tumor invasion and regional lymph node status between the two cohorts. Moreover, a lower rate of surgery but a significantly higher rate of chemotherapy and radiotherapy can be observed in the patients with SCLLM. Furthermore, SCLLM patients suffered a larger tumor size and a higher positive ratio of CEA, TD, and PNI, as well as a shorter median survival (Table 1).

Predictive Factors of Synchronous Colorectal Lung-Limited
Metastasis. This section of the study excluded therapeutic variables and postoperative variables, including colectomy, pulmonary surgery, radiotherapy, chemotherapy, TD, and PNI. All variables with p values less than 0.05 in the univariate logistic regression model were brought into the multivariate regression analysis, which displayed that 8 features, involving race, tumor location, pathological grade, histological type, T stage, N stage, and tumor size as well as CEA, could be used as the independent predictors (Table 2). Furthermore, a nomogram was constructed to clearly show the weight of each independent predictor. As the nomogram shown, the T4 stage contributed the most to SCLLM, followed by the N2 stage, elevated CEA, and rectal cancer (Figure 1). Various methods, including ROC curves, calibration curves and decision curve analysis (DCA), were utilized to evaluate the discriminating superiority of the

Prognostic Factors of Synchronous Colorectal
Lung-Limited Metastasis. The qualified variables, that identified by a univariate Cox regression model, were further analyzed by a multivariate regression analysis, which discriminated 9 independent prognostic factors, including age, race, marital status, pathological grade, T stage, colectomy/proctectomy, chemotherapy, CEA, and TD (Table 3). In order to visually demonstrate the impact of each prognostic factor on survival, the cumulative survival curves and nomogram were utilized in accordance with the result of the multivariate Cox regression model. The prognostic nomogram illustrated that nonresection/NOS played as the poorest prognostic factor, followed by nonchemotherapy, ≥75-year-old and T4 stage ( Figure 2). Meanwhile, the AUC values of ROC were 79.67%, 79.67%, and 76.97% regarding nomograms predicting 1-, 2-, and 3-year OS. The calibration curves demonstrated optimal agreement between model prediction and actual observations for 1-, 2-, and 3-year OS. The DCA indicated net benefits of the nomogram and each prognostic factor. Moreover, the cumulative survival curves revealed the influence of each prognostic factor on survival after controlling the other variables (Figure 3).

Discussion
To the best of our knowledge, this analysis was the first to look into the predictive and prognostic factors regarding OS for CRC with synchronous lung-limited metastasis. Colorectal oncologists have mainly focused on CRC with liver metastasis. Nevertheless, there is limited research on CRC with lung metastasis. The treatment of SCLLM commonly learns from the clinical experiences and strategies of treatment of colorectal hepatic metastasis [19]. In order to further improve treatment, it is essential to identify the specialized predictive and prognostic factors of SCLLM. CRC patients with high risk factors of lung metastasis should receive the particular treatments against prognostic factors and increase the frequency of follow-up. Previous studies reported that the pattern of colorectal lung metastasis was the direct invasion of cancer cells into the systemic circulation through the veins [13], which was different from the method of colorectal liver metastasis, that was thought to result from the lymphatic drainage of the colon and rectum [14]. It may be the reason why the T stage can be used as both predictor and prognostic factor but the N stage can only play as a predictor of SCLLM. Moreover, numerous researches reported that TD was associated with reductions in survival [20,21]. In fact, most of TD were thought to arise from lymphovascular invasion [22] and significantly related to T staging [22,23]. Therefore, TD may be a manifestation of the ability and depth of tumor invasion affecting the survival of SCLLM patients. 7 Gastroenterology Research and Practice RNE were considered as the priority for the assessment of the quality of surgery, which was mentioned in previous study [24], especially for the lack of the data concerning total mesorectal excision (TME) and complete mesocolic excision (CME) in the SEER database. The prognostic nomogram and survival curve manifested that standard colectomy/proctectomy with RNE ≥ 12 owned the clearest survival benefit comparing with noncolectomy and simplified resection. It is a consensus that high-quality colectomy/proctectomy means sufficient circumferential resection margin (CRM), which can be used as a specific therapeutic indicator against the depth of tumor invasion. Considering the critical role of T staging in patients with SCLLM, eligible TME/CME may be the most effective way to treat and prevent colorectal lung metastasis.
It is feasible to remove the primary tumor and liver metastasis in a simultaneous or staged approach for patients present with synchronous colorectal liver metastasis [25,26]. Although existing some controversy concerning the order of resection of the liver metastasis and the primary tumor [19], none of synchronous, sequential liver-first, or bowel-first surgery appeared inferior to the others [25,26]. Can the experience from colorectal liver metastasis be completely applied to SCLLM? The result of this study confirmed that independent pulmonary surgery, as a nonindependent prognostic factor in Cox regression analysis, did not improve the survival for SCLLM patients. Therefore, we believe that the approach of lung resection before resection of the primary tumor may be unreasonable for patients with SCLLM. Besides, more studies are needed to confirm whether the pulmonary surgery following by the colectomy/proctectomy cutting off the source of cancer cells and chemotherapy eliminating micrometastases can provide a survival benefit. In addition, CRC patients with metastatic diseases should receive radiation therapy cautiously [19]. This study believed that radiotherapy cannot improve survival for SCLLM patients as a whole. Nevertheless, it is meaningful to identify CRC patients who are sensitive to radiotherapy, as some other studies did [27,28].
A growing body of data has shown that the location of the primary tumor can be both prognostic and predictive of response to EGFR inhibitors in metastatic colorectal cancer [29][30][31]. This study demonstrated inconsistent risk of lunglimited metastasis among right colon, left colon, and rectum. Several studies also proposed that rectal cancer is prone to metastasize to the lungs [15,32]. Interestingly, there was no correlation between the primary site and the prognosis of patients with SCLLM. The mainstream opinions presently considered that targeted chemotherapy drugs, like cetuximab and panitumumab, improve survival for left-side colon patients but confer little benefit to right-side colon patients with metastatic diseases [29][30][31]. Does the consistent prognostic coefficient mean that the existing targeted drugs may not significantly prolong survival in all patients with SCLLM,  including left colon and rectal cancer? It is uncertain and requires prospective research to verify. A recent study involved the prognostic factors regarding cancer-specific survival for CRC with synchronous lunglimited metastasis [33]. However, study only focusing on cancer-specific survival inevitably misses some cases, such as those being not first tumor. Meanwhile, it is more reasonable to choose OS as the research endpoint since SCLLM, as a systemic disease, is able to affect the whole-body function. Limitations of this study include the following: (1) the use of retrospective data; (2) detailed treatment information for included patients were not recorded in the SEER cohort, and we could not investigate specific options, including chemotherapy regimen and specific surgical method, in the survival of SCLLM patients; and (3) the lack of some important genetic indicators, such as KRAS, NRAS, and BRAF. Future studies can focus on the molecular mechanisms of CRC with lung-limited metastasis.

Conclusion
This study identified independent predictors and prognostic factors for SCLLM based on a large database of the United States. The predictors and prognostic factors can provide supporting evidence for the prevention and treatment of SCLLM.

Consent
Patients' informed consent was waived because of the retrospective nature of the study design.

Conflicts of Interest
The authors declare that they have no competing interests and consent for publication.