Background
Colorectal cancer represents the most commonly diagnosed gastrointestinal cancer and the third most common cause of cancer-related death in the United States [
1]. The current TNM staging system for colorectal cancer is based on three elements: the penetration of tumor into the intestinal wall (T), the number of positive lymph nodes present (N), and the presence of metastasis (M). For patients without metastatic disease, surgery offers the only curative option. Chemotherapy is largely reserved for patients with positive lymph nodes (stage III disease) [
2], because it can reduce the risk of disease recurrence by 40 to 50%.
Clinicians do not currently question the benefit of chemotherapy for stage III colon cancer patients, despite the fact that 50% of these patients will eventually develop metastatic disease. Results of the Quick and Simple and Reliable (QUASAR) study implied that certain patients with stage II colon cancer (T3, T4/N0) may have more favorable outcomes with adjuvant therapy [
3]. Despite being controversial, chemotherapy for stage II disease is advised for patients with poor prognostic factors including T4 stage, less than 12 lymph nodes sampled at the time of resection, clinical bowel obstruction and perforation, and poor histologic grade with lymphovascular and perineural invasion [
3,
4]. The predictive accuracy of those clinico-pathologic characteristics has not been evaluated independently for stages II and III colon cancer. In this study, we aimed to examine the performance of those clinical predictors of recurrence-free survival for stage II and III colon cancer patients who were treated in our institution.
Methods
Patient selection
Eight hundred seventy-one patients with stage II colon cancer treated surgically between 1995 and 2007 and 265 patients with stage III colon cancer treated surgically between 1996 and 2001 were available for study. All patients had signed consent to be included in the study, and the appropriate approval from the Mayo Clinic Institutional Review Board (IRB) had been obtained.
Clinical and pathologic features
The clinical and pathologic features studied for patients with stage II colon cancer included year of surgery, age at surgery, sex, adjuvant chemotherapy, tumor location, primary tumor size, primary tumor classification, the total number of lymph nodes examined, and tumor differentiation. The clinical and pathologic features studied for patients with stage III colon cancer included year of surgery, age at surgery, sex, adjuvant chemotherapy, tumor location, primary tumor size, primary tumor classification, regional lymph node involvement, the numbers of positive, negative, and total lymph nodes; tumor differentiation, lymphovascular invasion, and perineural invasion.
Follow-up and recurrence of disease
We included all recurrences in this patient population, both local (anastomotic and regional) and distant (hepatic and lung metastases). The follow-up included a colonoscopy one year after surgery, with a yearly CT of the abdomen and pelvis every year for stage II and III disease, for up to five years. The frequency of repeat colonoscopies depended on the findings of the first surveillance colonoscopy. Patients with normal exam had a repeat colonoscopy three years later. Chest examination consisted of a chest x-ray, though current NCCN guidelines call for a chest CT. The patients were primarily followed by either their medical oncologists or a colorectal surgeon within our institution. As our institution is a large tertiary referral center, a high number of patients elected to be surveyed by local physicians. Those patients have been excluded due to lack of data for follow-up. In our study, we have included only patients who had recurrences either reported at their six-month surveillance visit or at a later date. The follow-up period for this cohort of patients by our institutions’ oncology team ranged up to ten years. There is a comprehensive multidisciplinary approach for all cancer patients and surveillance after colon cancer surgery is primarily carried by our medical oncologists. The recurrences we reported are not second primaries; these patients are followed closely, as our institution is part of the National Cancer Database sites.
Statistical methods
Continuous features were summarized with means, standard deviations (SD), medians, and ranges. Categorical features were summarized with frequency counts and percentages. Changes in features by year of surgery were evaluated, using Spearman rank correlation coefficients, Kruskal-Wallis and Wilcoxon rank sum tests, and chi-square tests. Recurrence-free survival rates were estimated, using the Kaplan-Meier method. Associations of the features studied with time to recurrence were evaluated, using Cox proportional hazards regression models and summarized with hazard ratios and 95% confidence intervals (CIs). Multivariable models were developed, using stepwise selection with a significance level for a feature to enter or leave the model of 0.05. The predictive ability of the features in a model was evaluated, using the
c (for concordance) index proposed by Harrell et al. [
5]. The interpretation of the c-index is identical to the interpretation of the area under a receiver operating characteristic curve. A
c-index of 1.0 indicates that the features in the model perfectly separate patients with different outcomes, while a value of 0.5 indicates that the features contain prognostic information equal to that obtained by chance alone. Statistical analyses were performed, using the SAS software package (SAS Institute, Cary, NC). All tests were two-sided and p-values <0.05 were considered statistically significant.
Discussion
The results of this study, analyzing the data of a total of 871 patients with stage II colon cancer, demonstrated a five-year recurrence rate of 10%. Most recurrences occurred in the first two years after surgery. The prognostic factor identified was the T stage. The population of the 265 patients with stage III colon cancer had, as expected, a much higher five-year recurrence rate of 30%, with most recurrences occurring within the first two years after surgery. The clinical prognostic factors for stage III colon cancer included the number of positive lymph nodes and the use of adjuvant chemotherapy.
Compared with other studies, the findings for risk of recurrence for stage II and III colon cancer are similar to those in our findings [
6]. The 7
th edition of the American Joint Commission on Cancer (AJCC) [
7] further classified T4 stage II tumors into the sub categories of T4a and T4b. This change was the result of observed differences in outcomes within the T4 classification, based on the tumor spread through the bowel wall either to just serosa (T4a) or to adjacent organs (T4b). The study that supported this finding examined 119,363 colon cancer patients from the Surveillance, Epidemiology, and End Results (SEER) database and showed that the survival rate of patients with stage IIB was lower than those with stage IIIA. The authors attributed this finding to the following factors: first, that patients with stage III received adjuvant treatment and therefore fared better than those with stage II disease that did not receive chemotherapy, and second, that patients with stage T4 N1 tumors might have been understaged as stage T4 N0 tumors. The first argument has been challenged by another study that had not shown statistically significant differences in survival among patients with stage IIB and IIIA disease [
8]. To support the argument that patients with stage IIB disease fare worse than patients with stage IIIA, a Dutch study that examined 2,282 patients with all stages of colorectal cancer [
9] demonstrated that patients with stage IIB tumors had a higher risk of developing locoregional recurrence when compared to patients with stage IIIA.
None of the other factors, including total number of lymph nodes, lymphovascular and perineural invasion, and tumor differentiation or clinical obstruction at the time of diagnosis were significantly associated with the risk of recurrence for the patients with stage II colon cancer. These findings differ from other studies demonstrating that certain pathologic characteristics, such as histologic grade, carry prognostic value. In particular, in a study of 1,031 patients who underwent a curative resection for colon adenocarcinoma, tumor differentiation was related to local recurrence with no events for patients with well-differentiated tumors. In comparison, patients with poorly differentiated tumors experienced a 6.8% risk of local, regional, or distant recurrence [
6].
The 7
th edition of AJCC [
7] emphasizes that at least 10–14 nodes should be retrieved in colon specimens for adequate staging. In our study, the mean number of lymph nodes was 16.5. Notably, the total number of lymph nodes examined was positively correlated with year of surgery (Spearman rank correlation coefficient 0.37; p < 0.001), an increase from a mean of 14.1, between years 1995 and 2001, to a mean of 20.5 for patients treated between 2002 and 2007 (p < 0.001). This increase of total number retrieval, however, did not improve disease-free survival rates for this 871 patient cohort. Studies focusing on stage II disease suggest that patients with fewer total lymph nodes retrieved at surgery fare worse than those who had a high number of total nodes recovered and examined [
10,
11]. This argument is founded on the potential of stage migration and the encounter of micrometastases, a finding that was not observed in our study. Furthermore, a c-index of 0.55 indicates that we cannot adequately predict recurrence for our stage II colon cancer patients, using current clinico-pathological features. It is difficult at this juncture to determine why the number of lymph nodes retrieved per specimen has increased, but acknowledge that the pathologic techniques of lymph node retrieval have improved over the years.
In stage III colon cancer, the increasing number of positive lymph nodes present was a stronger indicator of risk, as expected. It has been shown that after adjusting for T stage, patients with N0 disease (0 positive lymph nodes) have an expected 5-year survival rate of 86%, compared to those with N2 disease (patients with >3 positive lymph nodes) with expected 5-year-survival rate of 69% [
12]. Our results show that, after adjusting for adjuvant chemotherapy, each one-node increase in the number of positive lymph nodes was associated with a 24% increased risk of recurrence (HR 1.24; (118 – 1.31 95% CI) p < 0.001), verifying that an increasing number of positive lymph nodes is the most significant predictor of recurrence.
As expected, adjuvant chemotherapy for the stage III patients improved five-year disease-free survival rates, a finding consistent with those from the randomized clinical trials [
2]. In a prognostic nomogram of all stages of colon cancer [
13], adjuvant chemotherapy negated the negative prognostic factors of advanced T and N stage, and the c-index was 0.77 in predicting relapse for all stages of colon cancer. Although their reported c-index is promising, the model is driven by a larger proportion of stage I and IIA patients in the cohort and not by the stage III patients. Further, the published nomogram has not been validated by other institutions. The c-index of our multivariate model of the stage III colon cancer patients in our study was 0.70, much higher than the one found for stage II (c-index 0.56), however, not adequate. These findings illustrate the need to augment the TNM system for identifying individuals at high risk of recurrence.
A limitation of our study was the lack of follow-up of carcineoembryonic antigen (CEA). The role of CEA after surgical resection for colon cancer has been broadly assessed and, in spite of its widespread use, its utility has been controversial [
14]. The argument in support of CEA in follow-up is based on the fact that early detection of asymptomatic recurrences is possible in patients with an elevated CEA. Opponents of CEA testing argue that approximately 40% of all colorectal recurrences do not demonstrate increased CEA levels [
15], and no studies have demonstrated improved quality of life with frequent measurements. For these reasons, CEA measurements were not part of the surveillance for our patient cohort. Another limitation of our analysis was that all procedures took place in a specialized tertiary center, and the results may not be generalizable. In our institution, however, the fact that a group of specialized surgeons, medical oncologists and pathologists treated this patient cohort reduces the effects of treatment heterogeneity that exist in cohort studies of this nature. We are therefore better able to evaluate the independent predictiveness of current clinic-pathologic factors separately for stage II and III disease.
Further, in our study, we did not include any molecular markers of these tumors, including microsatellite instability (MSI) status. Our primary aim was to examine clinical and pathologic characteristics of stage II and III colon cancers, characteristics that are routinely obtained in community and specialty practice settings. Molecular profiling of colorectal tumors in the clinical setting carry great promise, but are not yet routinely performed as part of the current standard of care in the management of early stage colorectal cancers. For example, despite convincing evidence that MSI is a promising molecular marker with both prognostic and predictive value for chemosensitivity [
16,
17], it is not routinely obtained.
Competing interests
None of the authors listed have any competing interests, neither financial, nor non financial. There was no funding for this study.
Authors’ contributions
VT participated in the drafting of manuscript and developed study concept and design. DL performed critical revision of manuscript for intellectual content and database design. MH was the biostatistician and provided critical revision of manuscript for intellectual content. CL performed statistical analysis. PT participated in the drafting of the manuscript and development of study concept & design. All authors read and approve the final manuscript.