Subdivision of metastatic colorectal cancer for identifying patients who benet more from primary tumor resection

Stage IV colorectal cancer (CRC) patients are heterogeneous with distinctive clinicopathologic features and prognosis. Radical resection of primary tumor and distant metastases is associated with improved survival outcomes in metastatic CRC. The value of palliative primary tumor resection is controversial. The present study explored which subgroups beneted more from primary tumor resection in metastatic CRC. Methods Between 2004 and 2015, patients with metastatic CRC were identied using the surveillance, epidemiology, and end results (SEER) database. Uni- and multivariable Cox regression analysis were performed to identify factors associated with decreased cancer-specic mortality. The subgroups were divided based on the independent prognostic factors.


Introduction
Colorectal cancer (CRC) is one of the most common malignant diseases globally [1,2]. Approximately half of CRC patients developed distant metastases and over 80% of these patients have unresectable metastases [3][4][5]. The most common site of involvement is liver that accounted for more than one third of the total population [6].
Patients with a resectable primary and distant metastases could be managed with a staged or simultaneous resection and systemic chemotherapy. Cure is possible for these patients. However, for patients with unresectable metastases, palliative primary tumor resection is only required to manage bleeding, obstruction or perforation. The value of palliative primary tumor resection in patients with asymptomatic primary tumors and unresectable metastatic disease is still under debate. A previous study demonstrated that patients with asymptomatic unresectable stage IV colorectal cancer did not bene t from resection of the primary tumor using propensity score matching [7]. Similarly, JCOG1007 study found that primary tumor resection followed by chemotherapy had no bene t over chemotherapy alone for CRC patients with asymptomatic primary tumor and synchronous unresectable metastases [8]. On the contrary, Shida et al revealed that palliative resection of the primary tumor may be associated with improved overall survival. Two large real-world studies also found that palliative primary tumor resection resulted in improved overall and cancer-speci c survival [9,10]. Besides that, the bene ts of palliative resection of the primary tumor appeared to be consistent whether patients received systemic chemotherapy or not [11]. In short, the results seem to be not conclusive enough resulted from inconsistency, and it remains unclear which subgroup of these patients would bene t more from primary tumor resection. In the present study, we performed a large population-based study and subdivided patients with stage IV CRC into different subgroups to identify patients who responded well to primary tumor resection.

Statistical analysis
Uni-and multivariable Cox regression analysis were performed to recognize the prognostic factors. Groups with survival difference were divided into different subgroups. The Cox proportional hazards model was built to estimate the hazard ratios (HR) for cancer-speci c mortality after receiving primary tumor resection. All statistical analyses were performed with SPSS 25.0.

Baseline Characteristics
A total of 60358 stage IV CRC patients were identi ed. 34177 patients received primary tumor resection while 26181 patients did not receive primary tumor resection. If the patients with surgery of metastases were excluded, 25126 patients received palliative primary tumor resection while 24750 patients did not receive palliative primary tumor resection. The data about site of metastases were only available between 2010-2015. There were 302, 107, 16501, and 1918 patients with isolated bone, brain, liver and lung metastases, respectively. There were 18828, 5720, 707, and 41 patients with single, double, triple, and quadruple metastatic sites, respectively.

Prognostic factors of cancer-speci c survival
Based on univariate analysis, age at diagnosis, gender, marital status, race, serum CEA, histologic type, differentiation tumor location, surgery of primary lesion, surgery of metastases, site of metastases, chemotherapy and radiotherapy affected signi cantly cancer-speci c survival. When these factors were entered a multivariate analysis, it showed that age at diagnosis, marital status, race, serum CEA, histologic type, differentiation tumor location, surgery of primary lesion, surgery of metastases, site of metastases, chemotherapy and radiotherapy were signi cantly associated with cancer-speci c survival (  (Table 2). To evaluate the value of palliative resection of primary tumor, patients who underwent surgery of metastases were excluded from the analysis. Intriguingly, subgroups that bene ted more from palliative resection of primary tumor were consistent with those in entire cohort (Table 3).

Discussion
To date, the value of palliative resection of primary tumor in stage IV CRC patients with unresectable metastases is controversial. It is unreasonable to perform resection of primary tumor for all stage IV CRC patients. Previous studies did explore the subgroups who would gain more survival bene t from palliative resection of primary tumor. Our study focused on which subgroups bene ted more from primary tumor resection in metastatic CRC to identify patients who required aggressive therapy. The independent prognostic factors were identi ed using multivariable Cox regression analysis. Based on these factors, we divided stage IV CRC patients into different subgroups. Our ndings indicated that selected patients would gain more survival bene t from primary tumor resection.
Although young CRC patients are apt to have more aggressive characteristics, they have better survival outcomes as compared to elderly patients [12,13]. Our study also found that young stage IV CRC patients had superior prognosis as compared to elderly patients. In the univariable Cox regression analysis, primary tumor resection resulted in 54.6% and 50% decreased mortality risk in patients under 60 years old and those over 60 years old. However, the advantage of primary tumor resection in young patients decreased obviously using multivariable Cox regression analysis. It is reasonable to assume that young patients tend to have better performance status and they can tolerate and bene t from more intensive therapy. After adjusting for chemotherapy and radiotherapy, the predictive value of age for primary tumor resection is no longer signi cant.
Although black patients had poor prognosis as compared to white patients, black patients had 55.8% decreased mortality risk while white patients had 52% decreased mortality risk after receiving primary tumor resection. Serum carcinoembryonic antigen (CEA) testing is performed routinely for diagnosis, surveillance and treatment monitoring among CRC patients. Besides that, preoperative serum CEA was an independent prognostic factor in CRC [14] and CEA levels as an indicator could predict pathological complete response after neoadjuvant therapy for rectal cancer [15]. Our study found that serum elevated CEA levels were associated with worse prognosis in metastatic CRC. Primary tumor resection resulted in approximately 55.9% and 51.9% reduction in mortality risk for patients with normal and elevated CEA levels. Primary tumor resection is preferred in patients with normal CEA levels. Our previous study also evaluated guiding role of CEA levels in clinical decision-making in stage IIA colon cancer. We found that adjuvant chemotherapy did not acquire substantial survival bene t in patients with elevated pretreatment serum CEA levels [16].
Primary colorectal signet-ring cell carcinoma has distinct molecular features [17], aggressive behavior and worse prognosis [18] as compared to other histological subtypes. Our study found that primary tumor resection only decreased 37.7% mortality risk for patients with signet ring cell carcinoma. Shi et al found that metastatic CRC patients with signet ring cell carcinoma receiving chemotherapy had better survival than patients treated with surgery [19]. The value of primary tumor resection for stage IV CRC patients with signet ring cell type is limited and chemotherapy seems to be more useful for these patients. Similarly, for patients with poor or undifferentiated differentiation, survival bene ts conferred by primary tumor resection were less than those with well or moderate differentiation.
Our previous study found that distant metastasis site and the number of metastatic sites were associated with survival of stage IV CRC. Patients with isolated lung metastases presented better prognosis as compared to those with metastases at any other sites [20]. Different metastasis site had distinctive clinicopathological and molecular characteristics. Robinson et al suggested that patients with rectal cancer are more likely than patients with colon cancer to present with synchronous lung metastases [21].
Tie et al. reported that prevalence of KRAS mutations is higher in lung metastases than in liver metastases. It is in question whether different metastasis site affects the therapeutic effect of primary tumor resection [22]. As expected, we found that patients with isolated liver metastases bene ted more from primary tumor resection. Although patients with isolated lung metastases had better prognosis than those with isolated liver metastases, patients with isolated lung metastases gained less survival bene t from primary tumor resection than those with isolated liver metastases. In addition, patients with isolated brain metastases did not bene t from primary tumor resection, which may attribute to limited sample size. The number of metastatic sites also had an effect on therapeutic effect of primary tumor resection. Patients with single metastatic site bene ted more from primary tumor resection while patients with quadruple metastatic sites did not get bene t from primary tumor resection.
For patients receiving primary tumor resection combined with other therapy, such as surgery of metastases, chemotherapy and radiotherapy, the survival bene ts conferred by primary tumor resection were enlarged. Surgery of metastases, chemotherapy and radiotherapy had a synergistic effect on primary tumor resection. For patients who can tolerate the intensive therapy, combination therapy will bene t selected patients. Similarly, for patients with unresectable metastases, palliative primary tumor resection was associated with more survival bene t in patients who received chemotherapy or radiotherapy.
Inevitably, our study has several limitations. Firstly, some prognostic variables were missing in SEER database, such as perineural invasion, vascular invasion, chemotherapy regimens, response to chemotherapy and other metastatic sites, making it impossible to adjust for other potential confounders. Secondly, the sample size in some subgroups is insu cient. Thirdly, it is hard to acquire progression-free survival, which is critical for evaluation of therapeutic effect in stage IV disease.
In conclusion, the subdivision of metastatic CRC contributes to maximization of the value of primary tumor resection and facilitates individualized treatment options.

Declarations Ethics approval and consent to participate
This study was carried out based on SEER database which is publicly accessible. We have acquired permit from o cial mission for the research purpose. No personal identifying information was re ected.
The informed consent was not required for our study.

Consent for publication
Yes.

Competing interests
We declare no con icts of interest.