Poor survival in stage IIB/C (T4N0) compared to stage IIIA (T1-2 N1, T1N2a) colon cancer persists even after adjusting for adequate lymph nodes retrieved and receipt of adjuvant chemotherapy

A survival paradox between Stage IIB/C and Stage IIIA colon cancers exists. It is unclear how adequate lymph nodes dissection (LN) and post-surgery chemotherapy contribute to the survival paradox. We intended to assess the impact of these two factors on the survival paradox. We evaluated 34,999 patients diagnosed with stage IIIA or stage IIB/C colon cancer in 2003–2012 from the National Cancer Data Base. The 5-year overall survival (OS) was 73.5 % for stage IIIA and 51.1 % for stage IIB/C (P < 0.0001). The 5-year OS was 84.1 % for stage IIIA with post-surgery chemotherapy, 70.8 % for stage IIB/C with ≥ 12 LNs retrieved with chemotherapy, 53.9 % for stage IIB/C < 12 LNs with chemotherapy, 49.5 % for stage IIIA without chemotherapy, 43.7 % for stage IIB/C ≥ 12 LNs retrieved without chemotherapy, to 27.7 % for stage IIB/C < 12 LNs without chemotherapy. Even among stage IIB/C who had optimal treatment (≥12 LNs retrieved, received chemotherapy), OS remains lower than stage IIIA with chemotherapy. After adjusting LN dissection and chemotherapy in addition to the adjustment of other clinical factors, the survival paradox was reduced from HR = 1.76 (95 % CI: 1.68–1.85) to HR 1.51 (95 % CI: 1.44–1.59). LN dissection and post-surgery chemotherapy partially explained the survival paradox. More research is warranted to identify other factors that contribute to this paradox. Future iteration of TNM staging system should take this into consideration.


Background
For most solid cancers, the 7 th edition of the American Joint Committee on Cancer (AJCC) TNM staging system accurately prognosticates outcome with lower stage cancers having better prognosis than higher stage cancers [1]. However, colon cancer is one of the few exceptions. For stage IIB/C and stage IIIA, there exists a survival paradox [2][3][4][5]; the 5-year overall survival for patients with stage IIIA is approximately 70 % versus 46-61 % for stage IIB/C [1]. Such a paradox is attributed to several factors according to previous studies, such as stage migration due to inadequate nodal sampling or lack of systemic therapy for stage IIB/C [2,6]. We hypothesize that stage IIB/C is inherently more aggressive than stage IIIA, even after adjusting for receipt of chemotherapy and adequate nodal sampling. We propose to assess the simultaneous contribution of lymph node dissection and receipt of post-surgery chemotherapy to this survival paradox.

Data Source
The nationally recognized National Cancer Data Base (NCDB) is a joint project of the Commission of Cancer (CoC) of the American College of Surgeons and the American Cancer Society. More than 1500 CoCaccredited facilities in the U.S. contribute clinical information to the database. Approximately 70 % of newly diagnosed cancer cases in the U.S and 30 million historical records are captured in the database (https://www.facs.org/quality-programs/cancer/ncdb/puf). The data in the Participant User File (PUF) were de-identified and in compliance with the privacy requirements of the Health Insurance Portability and Accountability Act (HIPAA). The study was exempted from Institutional Review Board (IRB) approval by the Louisiana State University Health Sciences Center-Shreveport.
According to the NCDB's PUF dictionary [7], comorbidity was reported as Charlson/Deyo score: 0, 1 or 2 [8,9]. Age at diagnosis, race, facility type, facility location, urban/rural, insurance status, income and education levels for each patient's area of residence, comorbid conditions, anatomic site, tumor grade, surgical margin status, chemotherapy, and number of lymph nodes retrieved were variables selected for evaluation. NCDB does not have information on cause-specific survival and therefore, overall survival was calculated based on death from all causes.

Statistical Analysis
We calculated the propensity scores by stage IIB/C and stage IIIA using a multivariable logistic regression model. Only cases with matched scores based on potential confounders (i.e., age, race, distance from cancer reporting facility, facility type, facility location, rural/urban, insurance, income, education, comorbidity, primary site, grade, and surgical margins) were included in the multivariable analysis. The purpose of this approach was to ensure that stage IIB/C and stage IIIA cases were comparable to reduce potential confounder effect. Descriptive statistics for the different variable were presented. Univariable analysis of each variable was performed using chisquare test for categorical data and ANOVA for numerical data. The Kaplan-Meier method was used for survival analysis. Univariable Cox proportional hazard regression was used to identify factors significantly associated with the risk of death for all causes. Multivariable Cox proportional hazards regression analysis was used to determine independent significant factors associated with the risk of death for all causes, and hazard ratios (HR) and confidence intervals (CI) were calculated. Insurance status, income and education levels for each patient's area of residence were also adjusted in the multivariable analysis. Results are based on adjusted variables. A p-value ≤ 0.05 was considered statistically significant. All statistical analyses were performed using SAS Version 9.4 statistical software, (SAS Institute Inc., Cary, NC, U.S.A., 2013).

Results
The median follow-up was 39 months. Figure 1 demonstrates the Kaplan-Meier OS curve for stage IIB, stage IIC, and stage IIIA. Note that there is a significant survival difference between stage IIB/C and stage IIIA (P < 0.0001), although there was no significant survival difference between stage IIB and stage IIC (P = 0.46). Figure 2 demonstrates the Kaplan-Meier OS curve for the 6 subgroups which were defined by the number lymph nodes retrieved (<12 LNs vs ≥12 LNs) and whether or not systemic chemotherapy was given. For the entire cohort, the 5-year OS rate was 73.5 % for stage IIIA and 51.1 % for stage IIB/C. For the subgroups, the 5-year OS are 84.1 % for stage IIIA plus chemotherapy, 70.8 % for stage Fig. 1 Overall Survival for Stage IIB/C and Stage IIIA: Note that there is a statistically significant survival difference between stage IIIA and stage IIB/C (P < 0.0001). However, there is no significant difference between stage IIB and stage IIC (P = 0.46) IIB/C with ≥ 12 LNs plus chemotherapy, 53.9 % for stage IIB/C < 12 LNs plus chemotherapy, 49.5 % for stage IIIA without chemotherapy, 43.7 % for stage IIB/C ≥ 12 LNs without chemotherapy, and 27.7 % for stage IIB/C < 12 LNs without chemotherapy (P < 0.0001). The median survival has not been reached by the end of follow up (132 months) for stage IIIA with chemotherapy; it was 122.6 months for stage IIB/C, ≥ 12 LNs with chemotherapy, 72.5 months for stage IIB/C, < 12 LNs with chemotherapy, 58.9 months for stage IIIA without chemotherapy, 46.5 months for stage IIB/C, ≥ 12 LNs without chemotherapy, and 23.0 months for stage IIB/C, < 12 LNs without chemotherapy.
The poorest survival subgroup was stage IIB/C with < 12 LNs retrieved and without receipt of adjuvant chemotherapy. Note that even when patients with stage IIB/C received optimal treatment (≥12 LNs retrieved and receipt of chemotherapy), their OS remains significantly lower than those with stage IIIA who had chemotherapy (Fig. 2). Table 1 compares the demographic and therapeutic characteristics of stage IIB/C and stage III A before and after matching. Note that the two groups were fairly balanced after matching. Table 2 is a univariable analysis of factors associated with survival and Table 3 is the multivariable analysis based on the adjusted matched analysis. Note that stage IIB/C is an independent predictor of worse outcome compared to stage IIIA. T-stage and N-stage were not included because they were co-linear with stage of disease. Other independent factors associated with high hazard ratio include less than 12 lymph nodes retrieved, lack of receipt of chemotherapy, and positive margins. Before including LN dissection and chemotherapy in the multivariable model, the HR of stage IIB/C versus stage IIIA was 1.76 (95 % CI: 1.68-1.85). After adjusting for LN dissection and chemotherapy in addition to the adjustment of clinical variables included in Table 3, the HR was reduced to 1.51 (95 % CI: 1.44-1.59). Additional adjustment of demographic variables in Table 3 did not change the HR much (HR = 1.52; 95 % CI: 1.44-1.60).
Even with the latest iteration of AJCC, the OS for stage IIB/C remains lower than those with stage IIIA; the OS for stage IIB was 60.6 % versus 45.7 % for stage IIC and 67.2 to 73.7 % for stage IIIA [1]. Several groups attributed the survival paradox to either stage migration or lack of receipt of systemic therapy [2][3][4][5][6]. The implication is that if those with stage IIB/C received adequate lymph node dissection and adjuvant chemotherapy, their survival would have been better than those with stage III disease. No studies have demonstrated whether such is the case. Because the assumption has not been challenged, there is little impetus to revise the TNM staging system to account for the survival paradox. However, 1 % for stage IIIA plus chemotherapy, 70.8 % for stage IIB/C with ≥ 12 LNs plus chemotherapy, 53.9 % for stage IIB/C < 12 LNs plus chemotherapy, 49.5 % for stage IIIA without chemotherapy, 43.7 % for stage IIB/C ≥ 12 LNs without chemotherapy, and 27.7 % for stage IIB/C < 12 LNs without chemotherapy (P < 0.0001). The median survival has not been reached by the end of follow up (132 months) for stage IIIA with chemotherapy, 122.6 months for stage IIB/C, ≥ 12 LNs with chemotherapy, 72.5 months for stage IIB/C, < 12 LNs with chemotherapy, 58.9 months for stage IIIA without chemotherapy, 46.5 months for stage IIB/C, ≥ 12 LNs without chemotherapy, and 23.0 months for stage IIB/C, < 12 LNs without chemotherapy given our large robust database, we were able to analyze our cohorts based not only on receipt of chemotherapy, but also on the number of lymph nodes retrieved. To our knowledge, ours is the first and the largest dataset that assesses how these two factors, when combined, contribute to the paradox. Surprisingly, we found that even when adequate lymph nodes were retrieved and chemotherapy was administered, the survival paradox persists. What this suggests is that T4N0 lesions (i.e. stage IIB/C) may be inherently more aggressive than stage IIIA, although further research is necessary to delineate the root cause of the poorer survival among them. Positive margins portend a poor outcome [15]. Inadequate resection of locally advanced lesions (i.e. T4) can lead to positive surgical margins, especially if an en-bloc resection was not performed [16]. In our study, positive margins impart a 56 % increased risk of death compared to negative margins. What impact surgical margins have on the paradox for optimally treated patients with stage IIB/C colon cancer will be the subject of our next investigation.
The role of adjuvant chemotherapy for patients with resected stage II colon cancer remains controversial. The 5-year OS following definitive colectomy for stage II colon cancer is in the range of 85-89 % [17]. Whether adjuvant systemic therapy can further improve this rate remains an area of intense investigation. Besides the QUASAR (Quick and Simple and Reliable) trial [18], which was a phase III trial that found OS benefit for patients with Stage II colon cancer, multiple other randomized trials as well as meta-analysis found no such significant advantage with chemotherapy [17,[19][20][21][22][23][24][25]. Although the commonly cited QUASAR trial (Quick and Simple and Reliable) showed an absolute improvement of 3.6 % for stage II colon cancer that received adjuvant therapy, the trial had several limitations [18]. Approximately 8.5 % of 3,239 patients who were thought to have stage II colon cancer actually had stage I or III  disease and almost 30 % had rectal cancer (many of these patients received radiation therapy) [18]. To our knowledge, our analysis is the first to demonstrate a survival advantage of using adjuvant chemotherapy for stage II disease, but only for those with stage IIB/C colon cancer since we have 138,572 stage IIA patients in NCDB. Our data lend support to the American Society of Clinical Oncology's recommendation of postoperative chemotherapy for patients with T4 tumors [26]. Of interest is that even when optimal treatment was rendered to those with stage IIB/C disease, their OS remains significantly lower than those with stage IIIA who had chemotherapy. Other factors that contribute to the survival paradox will need to be further investigated. Given our provocative data, we believe that randomized control studies are needed to determine whether there is a survival advantage of using adjuvant chemotherapy for stage II colon cancer.
One of the limitations of our study is the lack of recurrence data since this is not recorded in the NCDB. Additionally, we do not have specific causes of death and therefore it is plausible that deaths may not be related to cancer. We also do not have data on other factors such as microsatellite instability (MSI), preoperative CEA level, whether or not patients presented with obstruction or perforation, whether there was evidence of venous or perineural invasion, and whether blood transfusion had taken place, all of which are important prognosticators. However, our results are hypothesisgenerating and can serve as a platform to further evaluate and explain the survival paradox.

Conclusions
Based on our analysis of nearly 35,000 patients, we confirmed that a survival paradox between Stage IIIA and Stage IIB/C colon cancer patients exists for cases diagnosed in 2003-2012 in the ACoS' hospitals. Although inadequate lymph nodes retrieved and lack of receipt of adjuvant chemotherapy contributes to stage IIB/C poor survival compared to stage IIIA, they themselves do not entirely explain the paradox. Further studies are necessary