Survival impact of surgical resection of primary tumor in patients with stage IV colorectal cancer: Results from a large population-based cohort study
Shahid Ahmed MD,
Saskatchewan Cancer Agency, Regina, Canada
Department of Oncology, University of Saskatchewan, Saskatoon, Canada
Department of Community Health and Epidemiology, University of Saskatchewan, Saskatoon, Canada
Corresponding author: Shahid Ahmed, MD, FRCPC, FACP, Clinical Professor of Medicine, Site Leader of Gastrointestinal Cancer Group, Saskatoon Cancer Center, 20 Campus Drive, University of Saskatchewan, Saskatoon, SK, Canada, S7N4H4; Fax: (306) 655-0633; firstname.lastname@example.org
We thank Dr. Bonnie Janzen, Mrs. Carla Woites, Dr. Tong Zhu, Mrs. Coralee Prodaehl, and Saskatchewan Cancer Agency for their support for this project.
Currently, there is very low-quality evidence available regarding benefit of surgical resection of the primary tumor (SRPT), in patients with stage IV colorectal cancer (CRC). In the absence of randomization, the reported benefit may reflect selection of younger and healthier patients with good performance status. A large population-based cohort study was undertaken to determine the survival benefit of SRPT in advanced CRC by eliminating various biases reported in the literature.
A retrospective cohort study involving patients with stage IV CRC, diagnosed between 1992 and 2005, in the province of Saskatchewan, Canada. Survival was estimated by using the Kaplan-Meier method. Survival distribution was compared by log-rank test. Cox proportional multivariate regression analysis was performed to determine survival benefit of SRPT by controlling other prognostic variables.
A total of 1378 eligible patients were identified. Their median age was 70 years (range, 22-98 years) and male:female ratio was 1.3:1; 944 (68.5%) of them underwent SRPT. Among 1378 patients, 42.3% received chemotherapy and 19.1% received second-generation therapy. Patients who underwent SRPT and received chemotherapy had median overall survival of 18.3 months (95% confidence interval [CI] = 16.6-20 months) compared with 8.4 months (95% CI = 7.1-9.7 months) if they were treated with chemotherapy alone (P < .0001). Cox proportional analysis revealed that use of chemotherapy (hazard ratio [HR] = 0.47, 95% CI = 0.41-0.54), SRPT (HR = 0.49, 95% CI = 0.41-0.58), second-line chemotherapy (HR = 0.47, 95% CI = 0.45-0.64), and metastasectomy (HR = 0.54, 95% CI = 0.45-0.64) were correlated with superior survival.
Whereas surgery is the primary treatment of localized colorectal cancer (CRC), resection of the primary tumor in patients with incurable metastatic disease is usually recommended for palliative purposes to manage obstruction, perforation, or bleeding. Only a subset of patients with limited liver and or lung metastases can be cured with multimodality therapy. Complete resection of the primary tumor with metastatic lesions, in these patients, has been associated with durable remission in approximately 40% cases. The role of surgical resection of the primary tumor in patients with newly diagnosed incurable stage IV CRC remains controversial. Despite uncertain survival benefit, a high rate of surgical resection has been reported in patients with unresectable metastatic disease. Currently, there is very low-quality evidence available regarding survival benefit of resection of primary tumor, in patients with stage IV CRC and otherwise unresectable metastatic lesions. Although some have advocated for surgery,[3, 4] others have failed to demonstrate survival benefit with resection of primary tumor in patients with stage IV CRC.[5, 6] A recent meta-analysis of literature involving 15 observational studies revealed 31% reduction in mortality (hazard ratio [HR] = 0.69, 95% confidence interval [CI] = 0.61-0.79) with surgical resection of the primary tumor, with an absolute difference in median survival of approximately 4 months.[7, 8] Notably, comparable survival benefits were noted in studies using newer generation chemotherapy versus older regimens, and in minimally symptomatic or asymptomatic patients versus symptomatic patients. Nevertheless, in the absence of randomization, the reported benefit may reflect selection of younger and healthier patients with good performance status. Taking selection bias into consideration, recent observational studies do not advocate surgery and suggest that outcomes are not compromised, by leaving the primary colon tumor intact, in patients treated with modern chemotherapy.[9, 10]
To date, no randomized trial has assessed the survival impact of primary tumor resection, in patients with stage IV CRC. In the absence of randomized controlled trials, this study was undertaken to investigate the optimal surgical management of patients diagnosed with stage IV CRC, by controlling for various known confounding factors including age, comorbid illness, performance status, and other important prognostic variables.
The primary objective of this study was to compare survival of patients diagnosed with stage IV CRC, who underwent surgical resection of primary tumor, with that of patients who did not have surgery. The secondary objectives were as follows: 1) to compare survival of patients with stage IV CRC who were treated with chemotherapy and underwent resection of the primary tumor, with survival of patients who did not have surgery; 2) to compare survival of asymptomatic or minimally symptomatic patients with stage IV CRC, who underwent surgical resection of primary tumor, with the patients who did not have surgery; and 3) to determine survival benefit of primary tumor resection by controlling for various clinicopathological variables that affect the outcome of patients with metastatic CRC.
MATERIALS AND METHODS
The study protocol was approved by the Research Ethics Board of the University of Saskatchewan, Canada. The study population was composed of a cohort of patients 18 years or older, with histologically documented adenocarcinoma of colon and rectum, intact primary tumor, and evidence of metastases (stage IV CRC), who were diagnosed between the period of 1992 and 2005, in the province of Saskatchewan. Patients with other histological diagnosis or with other active secondary malignancy or with fixed inoperable primary tumor at the time of diagnosis were excluded. The Saskatchewan Cancer Registry database is prospectively collected and updated. Eligible patients were identified from the Saskatchewan Cancer Registry. All medical records were individually reviewed, and data was abstracted by a trained research associate, using a prespecified and validated abstraction sheet.
Because a high rate of primary tumor resection has been reported in the literature, assuming two-thirds of patients in Saskatchewan underwent resection of primary tumor, a ratio of 1:2 was used for the nonresection group versus the resection group. Using a power of 90%, type 1 error of 0.05 with 2-sided P value, and survival difference of 20% in the 2 groups (resection versus nonresection) with a follow-up period of 60 months, a total sample size of 959 patients was estimated (326 in the nonresection group and 633 patients in the resection group).
Analysis of Primary and Secondary Endpoints
Overall survival was defined as “time from the diagnosis of stage IV CRC to death from any cause.” Patients who did not die at the end of the follow-up period, or were lost to follow-up during the study, were censored at the last date they were known to be alive. Survival of the entire cohort and subgroups was estimated by using the Kaplan-Meier method. Survival distribution of different groups was compared by the log-rank test. The overall significance level was set at 0.05. A multivariate analysis was performed to determine the prognostic significance of primary tumor resection in patients with stage IV CRC. The Cox proportional hazard model was used and the HR and its 95% CI were estimated. Following variables were examined with respect to their prognostic significance: Resection of primary tumor, age (< 65 years versus ≥ 65 years), sex, major comorbid illness, Eastern Cooperative Oncology Group (ECOG) performance status (< 2 versus ≥ 2), smoking, sodium level (≤ 135 mEq/L versus > 135 mEq/L), serum creatinine (≥ 120 versus < 120 umol/L), blood urea nitrogen (BUN) (≥ 8 versus < 8 mmol/L), albumin (≥ 36 versus < 36 g/L), bilirubin (≥ 26 versus < 26 umol/L), alkaline phosphatase (≥ 120 versus < 120 U/L), hemoglobin (≥ 120 versus < 120 g/L, white blood cell (WBC) (≥ 11 × 109/L versus < 11 × 109/L), platelet (≥ 450 × 109/L versus < 450 × 109/L), carcinoembryonic antigen (CEA) (≥ 6 versus < 6 μg/L), site (colon versus rectum), grade (3 versus < 3), symptomatic disease, extrahepatic metastases, disease burden (patients with metastasis confined to one organ or site versus metastases in more than one organ or site), metastesectomy, use of chemotherapy, second-line chemotherapy, and second-generation chemotherapy. For the variables examined in the final mathematical model, the proportional hazards assumption was assessed using log-log survival curves.
The asymptomatic or minimal symptomatic disease was defined as absence of obstruction, perforation, or bleeding. Major comorbid illnesses were defined as presence of coronary artery disease, diabetes mellitus, chronic renal insufficiency, chronic obstructive lung disease, and others (uncontrolled hypertension, peripheral vascular disease, stroke or transient ischemic attack, interstitial lung disease, congestive heart failure, cardiac arrhythmia, among others). Postoperative mortality was defined as death that occurred within 30 days of surgery. Second-generation therapy was defined as use of bevacizumab or anti–epidermal growth factor receptor (anti-EGFR) antibodies cetuximab or panitumumab and or oxaliplatin- or irinotecan-based therapy (FOLFOX or FOLFIRI).
All variables with P < .05, on univariate analysis, were examined in a multivariate model to assess their correlation with survival. A best-fitted but a most parsimonious model was built by identifying the important predictors for survival. The likelihood ratio test and t test were used to determine if the addition of independent variables of interest add significantly to the prediction of survival in the model. Test for interaction was performed for resection of primary tumor and all the variables that were significantly correlated with survival. An imputation technique was used for missing data. A 2-sided P value of < .05 was considered to be statistically significant. SPSS software, versions 20.0 and 21.0, were used for statistical analysis (SPSS Inc, Chicago, Ill).
A total of 1378 eligible patients were identified (Fig. 1). Their median age was 70 years (range, 22-98 years) and male-to-female ratio was 1.3:1. Seven-hundred eighty one (56.7%) patients were male, 1005 patients (72.9%) had ECOG performance status of < 2, 856 patients (62.1%) had a comorbid illness, and 204 patients (14.8%) had a history of secondary malignancy. Patient characteristics are described in Table 1. A significant difference was noted between the 2 groups with respect to age, performance status, extrahepatic disease, mean serum albumin, alkaline phosphatase, bilirubin, BUN, CEA, and WBC count. As expected, patients in the surgical group tended to be younger with better performance status and organ function and less likely to have extrahepatic disease.
Table 1. Baseline Characteristics of Patients in the 2 Groups: Patients Who Underwent Surgical Resection of Primary Tumor and the Control Group Who Did Not Have Surgery
Of 1378 patients, 29.5% had rectal or rectosigmoid tumor (rectal, 20.1%, rectosigmoid, 9.4%). One hundred ninety-seven (14.3%) patients had mucinous tumor. A total of 544 patients (39.5%) were symptomatic, among whom 454 (83%) had obstructive symptoms, 87 (16%) developed bowel perforation, and 56 (10%) had major bleeding (drop in hemoglobin of ≥ 20 g/L or required blood transfusion). Symptoms were not mutually exclusive. One thousand thirty-eight (75.3%) patients had liver metastases and 698 (50.7%) had extrahepatic metastases. Among 698 patients with extrahepatic disease, 217 (31%) had lung metastases, 205 (29.3%) had peritoneal involvement, 31 (4.4%) had bony metastases, and 9 (1.3%) patients had documented brain metastases.
Of 1378 patients, 944 (68.5%) underwent resection of primary tumor and 434 (31.5%) did not have surgery. Five hundred and eighty-three (42.3%) patients received 5-fluorouracil–based chemotherapy (Table 1). Of 583 patients, 222 (38%) patients, on progression, were treated with second-line chemotherapy mostly oxaliplatin- and/or irinotecan-based therapy with or without biologics. A significantly higher number of patients in the surgical group were treated with chemotherapy, or received second-line or newer generation chemotherapy. Likewise, a significantly higher number of patients in the surgical group underwent metastasectomy.
Follow-Up and Survival
Median follow-up time was 7.1 months (interquartile range, 2.5-17.5 months) for the whole cohort. There was no loss of follow-up in both groups. Median survival of whole cohort of patients who underwent surgical resection of primary tumor was 10.6 months (95% CI = 9.5-11.7 months) compared with 3.0 months (95% CI = 2.5-3.5 months) of patients who did not have surgery (P < .0001). Overall 30-day mortality rate in the surgical group was 6.6%. Median survival of cohort of patients who received chemotherapy was 15.9 months (95% CI = 14.5-17.3 months). Patients who underwent surgical resection of primary tumor and received chemotherapy had median overall survival of 18.3 months (95% CI = 16.6-20 months) compared with 8.4 months (95% CI = 7.1-9.7 months) if they were treated with chemotherapy alone (P < .0001) (Fig. 2). In a subgroup of patients who were treated with second-generation chemotherapy, median survival of patients who underwent surgical resection of primary tumor was 24.6 months (95% CI = 20.2-29.0 months) versus 11.0 months (95% CI = 7.8-14.3 months) if they did not have surgery (P < .0001).
The median survival of a subcohort of 834 asymptomatic or minimally symptomatic patients who were treated with chemotherapy and underwent surgical resection of primary tumor was 18.3 months (95% CI = 16.6-20.0 months) compared with 8.4 months (95% CI = 7.1-9.7 months) if they did not have surgery (P < .0001) (Fig. 3). After excluding the patients who underwent metastasectomy, median survival of patients who underwent surgical resection of primary tumor was 15.2 months (95% CI = 13.5-16.9 months) compared with 8.3 months (95% CI = 7.1-9.6 months) if they did not have surgery (P < .0001).
On univariate analysis, various clinicopathological factors were indentified that were correlated with survival (Table 2). Among them, ECOG performance status of > 1 was most strongly correlated with a poor survival, whereas use of chemotherapy was most strongly correlated with better survival. Tests for interaction between surgical resection of primary tumor and age, performance status, CEA level, second-line therapy, or more than one metastatic site were significant. The surgical resection of primary tumor was associated with better survival in younger patients, patients with good performance status, normal CEA level, patients treated with second-line therapy, and patients with one metastatic site. For instance, unadjusted HR for mortality in patients 65 years or younger was 0.28 (95% CI = 0.22-0.36) compared with 0.48 (95% CI = 0.42-0.55) in older patients in relationship with surgical resection of primary tumor. Likewise, with respect to surgery, unadjusted HR for mortality in patients with metastasis confined to one organ or site was 0.36 (95% CI = 0.31-0.42) compared with HR of 0.56 (95% CI = 0.45-0.68) in patients with metastases involving more than one organ or site. Cox proportional multivariate regression analysis using interaction terms revealed that use of chemotherapy (HR = 0.47, 95% CI = 0.41-0.54), surgical resection of the primary tumor (HR = 0.49, 95% CI = 0.41-0.58), second-line chemotherapy (HR = 0.47, 95% CI = 0.45-0.64), and metastasectomy (HR = 0.54, 95% CI = 0.45-0.64) were correlated with a favorable survival in patients with advanced CRC, whereas older age, poor performance status, low albumin, elevated bilirubin, elevated alkaline phosphatase, anemia, leukocytosis, colonic primary, and grade 3 tumor were correlated with inferior survival (Table 3, Fig. 4). After controlling for the other significant or clinically important variables, only the interactions between the surgical resection of primary tumor and second-line therapy, or more than 2 metastatic sites, were significant. Because approximately 86% of patients who were treated with second-line therapy received second-generation chemotherapy, in order to avoid multicollinearity, second-generation therapy was not included in the final model. In a secondary analysis, after excluding 198 patients who underwent metastasectomy, surgical resection of primary tumor significantly correlated with better survival HR of 0.43 (95% CI = 0.41-0.52). After adjusting for other important prognostic variables in a Cox proportional multivariate model, the HR for survival with surgical resection of primary tumor was 0.54 (95% CI = 0.48-0.62).
Table 2. Univariate Correlation Between Various Clinicopathological Variables and Overall Survival in Patients with Stage IV Colorectal Cancer
This is the first large population-based cohort study that demonstrated survival advantage of surgical resection of primary tumor, in patients with stage IV CRC, independent of other important prognostic factors including age, performance status, comorbid illness, and chemotherapy. Several studies previously have shown survival benefit of resection of the primary tumor. However, many studies did not provide baseline characteristics of patients in each group, whereas others showed significant differences between the 2 groups and failed to control for the differences.[7, 8] Hence, validity of survival benefit observed in these studies has been questioned and is believed to be biased by the selection of healthier and younger patients for surgery. Despite significant differences in the baseline characteristics between the 2 groups in the study cohort, when these variables were included in a multivariate model, resection of the primary tumor remained an important prognostic variable and was associated with 51% relative reduction in mortality when adjusted for age, performance status, comorbid illnesses, chemotherapy, metastasectomy, second-generation chemotherapy, disease burden, and various other important prognostic factors. Notably, survival differences between the 2 groups (resection versus nonresection), within the study cohort and various subgroups, were consistently more than 6 months (range, 7.6-13.6 months).
There is a general agreement that patients with evidence of perforation, significant obstruction, or uncontrolled bleeding should undergo resection of the symptomatic tumor. However, surgical resection of primary tumor, in patients with asymptomatic or minimally symptomatic disease and otherwise unresectable metastases, is not recommended. In the study cohort, approximately 60% of patients were asymptomatic or had minimal symptoms. These patients had comparable survival benefit with surgical intervention. Moreover, the surgical intervention was independently associated with a survival benefit, after controlling for other important prognostic variables (analysis is not shown).
Modern chemotherapy regimens incorporating novel cytotoxic and biological agents have been associated with high response rates of 40% to 50%. Although only approximately 19% of the study cohort was treated with modern chemotherapy, the magnitude of benefit of surgical intervention was substantially higher in a subgroup of patients who were treated with second-generation and/or second-line chemotherapy. In the study cohort, less than 2% of patients received biologic therapeutics including bevacizumab or anti–epidermal growth factor receptor (anti-EGFR) monoclonal antibodies such as cetuximab or panitumumab, hence, a subgroup analysis was not feasible. Nevertheless, a retrospective analysis of a Dutch trial that primarily assessed the efficacy of combination of chemotherapy and biologics demonstrated a superior survival with surgical resection of primary tumor. In a subgroup of patients who underwent surgical resection of primary tumor, a significantly better median overall survival of 20 months was observed compared with 13.4 months for the patients in the nonresection group, (P < .0001; HR = 0.65, 95% CI = 0.52-0.80). Likewise, a pooled analysis of 4 French trials including study involving targeted therapy demonstrated 37% reduction in mortality (HR = 0.63, 95% CI = 0.53-0.75) with surgical resection of primary tumor. Therefore, evidence supports comparable benefit of resection of primary tumor in patients treated with modern regimens. Notably, the tests for interactions between the surgical resection of primary tumor and second-line therapy or disease burden were significant after adjusting for other variables, suggesting that patients who received second-line therapy or who had metastasis confined to one organ or site tend to get most benefit from surgical removal of primary tumor.
The underlying mechanism of potential survival benefit with removal of the primary tumor is not known. It is well known that surgical resection of primary tumor, in some cancers, such as advanced ovarian and renal cell cancer, has been associated with significant survival benefit.[13, 14] The primary tumor may secrete cytokines that promote tumor growth and reduce response to cytotoxic agents. Moreover, noncurative resection of the primary tumor in patients with advanced cancer may prevent local tumor complications, improve disease control by reducing the tumor bulk and tolerance to systemic therapy.
A variety of clinical parameters such as age, performance status, WBC count, hemoglobin, serum albumin, alkaline phosphatase, CEA, pathological grading, or localization of the primary tumor have been identified as prognostic markers in patients with stage IV CRC.[16-19] In the study cohort, in addition to resection of primary tumor, use of 5-fluorouracil–based chemotherapy; second-line chemotherapy; metastasectomy; older age; performance status; colonic primary; high-grade tumor; and baseline elevated alkaline phosphatase, bilirubin, anemia and leukocytosis were independently correlated with survival. As expected, significantly more patients were diagnosed with a rectal or rectosigmoid tumor, in the control group. Despite that, for reasons that are not clear, patients with rectal or rectosigmoid tumors compared with those who had more proximal tumor had better survival. Ferrand and colleagues reported a similar observation in their retrospective analysis of the FFCD 9601 trial. These differences in patient survival were maintained after exclusion of patients with rectal primary tumors.
To our knowledge, this is the only large population-based cohort study that examined age, comorbid illness, performance status, serum albumin, and other important clinical variables in a multivariate model and demonstrated survival benefit of primary tumor resection independent of other variables. One of the limitations of the current study is that it did not assess BRAF mutation, which is an important prognostic marker in stage IV CRC. In addition, disease burden was not measured in patients with single versus multiple site metastases. Hence, better survival secondary to selection of patients with low disease burden, who have better prognosis, cannot be eliminated.
In summary, this well-designed population-based cohort study, with minimal selection and information biases, supports surgical resection of primary tumor in patients with stage IV CRC, independent of other important prognostic factors. In addition to systemic therapy and metastasectomy, primary tumor removal was associated with better survival. A Dutch randomized controlled trial (CAIRO 4) assessing survival benefit of surgical resection of primary tumor in patients with advanced CRC, with estimated improvement in median survival from 13 months to 19 months, once completed, will be valuable to confirm our findings.
The study is supported by a research grant provided by the Saskatchewan Cancer Agency, Saskatchewan, Canada.