Adverse Histological Features Alone Was Not Sucient for Decision of Adjuvant Chemotherapy in T3N0M0 colonic cancer: A Propensity Score Matching Study Based on Population Analysis

Purpose: Colorectal cancer (CRC) is one of the leading causes of cancer-related deaths worldwide. Approximately 25% of patients with AJCC stage II suffered recurrence after radical surgery, but the role of adjuvant chemotherapy in Stage II CRC is controversial. We aimed to evaluate the benet from chemotherapy for T3N0M0 patients with adverse histological features. Methods: An analysis of the Surveillance, Epidemiology, and End Results (SEER) 18 databases was performed to identify patients underwent radical resection 20-80 years old with colonic carcinoma diagnosed during year 2004 to 2012. Poorly differentiated, undifferentiated, mucinous adenocarcinoma and signet ring cell carcinoma (SRCC) were dened as adverse histological features, and well or moderately differentiated adenocarcinoma as good histological features. Multivariate analyses were used to adjust the effects of other covariates. The survival benet of adjuvant chemotherapy was demonstrated in the propensity score matching cohort. Results: The adjuvant chemotherapy brought no survival benet (good histological features HR, 1.076; 95% CI, 0.957-1.211; P=0.220; adverse histological features HR, 0.995; 95% CI, 0.777-1.274; P=0.967); Neither mucinous adenocarcinoma nor signet ring cell carcinoma brought a signicance on survival (mucinous adenocarcinoma HR, 0.699; 95% CI, 0.488-1.002; P=0.051; signet ring cell carcinoma HR, 0.750; 95% CI, 0.386-1.459; P=0.397) compared to adenocarcinoma. Left-side colon was associated with a poorer survival in good histological features group (HR, 1.136; 95% CI, 1.034-1.248; P=0.008). Conclusions: The adverse histological features don’t inuence on survival for patients with T3N0M0 staging, and adjuvant chemotherapy is invalid to improve the survival for these patients, indicating adverse histological features alone was not sucient for decision of adjuvant chemotherapy in this situation rather than the recommendation of the guidelines.


Introduction
With its incidence currently increasing, colorectal cancer (CRC) is one of the most common cancer types worldwide contributing a major cause of cancer death, and 24-27% have stage II and 27-31% stage III disease [1,2]. Although an bene t of adjuvant treatment with uoropyrimidines is con rmed to improves the outcome of stage II/III cancer patients with radically resected colorectal cancer [3], controversy is common about the current clinical guidelines for stage II disease, especially when it refers to the bene t of the adjuvant chemotherapy [3]. Current clinical algorithms are unable to precisely predict which colorectal cancer patients would bene t from adjuvant chemotherapy. Adjuvant uoropyrimidine-based chemotherapy is commonly administrated for colonic carcinoma in patients with stage III and to highly selective stage II with risks including T4 status, poorly differentiated histology, tumor perforation and insu cient examined lymph nodes [4][5][6], with expectation that high-risk patients could bene t from uorouracil adjuvant chemotherapy, however, the survival bene t has not been con rmed [2]. Adjuvant chemotherapy in stage II colonic cancer is still a matter of debate and remains a challenge due to the limited survival bene t and considerable toxicity.
Identi cation of prognostic and predictive biomarkers would contribute to the improvement of strategies to optimize selection of treatment options. A careful evaluation of precise bene t for the completely resected colonic cancer with stage II disease is of great signi cance, especially for patients staging as T3N0M0 with so called "high risks".
We sought to facilitate a more practical approach based on predictive and prognostic factors, so as to prove the survival bene t of chemotherapy with the aim to minimizing its side effects in a large cohort of patients from the SEER database.

Study population
It is feasible to investigate the prognostic factors of target disease using the data from the SEER database, which is sponsored by the National Cancer Institution. In this study, we identi ed individuals diagnosed with microscopically con rmed adenocarcinoma in detail of the colon excluding rectum and rectosigmoid junction from 2004-2012.
The current SEER database, which is an authoritative source of information in the United States, currently collects and publishes cancer data from population-based cancer registries across the United States, approximately covering 28% of the United States population [7,8] The inclusion criteria were as follows: (1) patients were diagnosed form 2004 to 2012; (2) the site code (ICD-O-3) was limited to colon whose diagnosis were pathologically con rmed excluding rectum and rectosigmoid junction (C18.0, C18.2-C18.7); (3) histology code (ICD-O-3) was restricted to adenocarcinoma, mucinous adenocarcinoma and signet ring cell carcinoma (8140/3, 8480/3, 8490/3); (4) patients whose TNM staging were limited to T3N0M0 (staging based on 6th or 7th AJCC system according to the version of the day); and (5) patients who underwent radical resection (D2 or D3 lymphadenectomy with negative resection margins (R0) ) with exact number of resected regional lymph nodes for evaluation; (6) age of patients was limited to 20 to 80 (less than) years old; (7) patients with one primary tumors; (8) information on CSS and survival months was available. (Fig. 1).

Variable Declaration
Age was grouped as less than 70 years old (< 70) and more than 70 years old (≥ 70). Race status was divided into black, white and others. Marital status was divided to married, single or divorced. There were three histology groups, adenocarcinoma, mucinous adenocarcinoma or signet ring cell carcinoma (SRCC). Poorly differentiated or undifferentiated adenocarcinoma, mucinous adenocarcinoma and signet ring cell carcinoma (SRCC) were all de ned as adverse histological features, while well or moderately differentiated adenocarcinoma were de ned as good histological features. Tumor site was grouped as right-side colon and left-side colon whose cut-off was splenic exure. Regardless of the sequence, adjuvant chemotherapy was grouped as no or yes.

Statistical Analyses
Cox proportional hazards regression model was used to conduct the multivariate analysis with backward, stepwise elimination of variables. HRs were adjusted for lymph nodes examined, age (< 70 or not), race, sex, marital status, tumor histology, grade and tumor location. The demographic and clinical characteristics of the cohorts are demonstrated in Table 1. Besides the factors mentioned above, patients were well balanced for lymphovascular invasion and perineural invasion. Kaplan-Meier method was used to calculate the survival curves and to plot the survival curves. Differences of survival was compared by the log rank test. The 5-year cancer speci c survival (CSS) was the primary endpoint in this study, and it was de ned from the diagnosis to the date of cancer death. Death attributed to other causes was de ned as a censored observation. Univariate and multivariate analyses were conducted respectively using Cox proportional hazards regression models to identify the factors which would in uence the CSS independently either in population with good histological features or in population with adverse histological features. A method of propensity score matching (PSM) was utilized to balance the baseline characteristics of the population with adverse histological features between groups separated by whether administrated by chemotherapy or not. The propensity score was calculated by logistic regression including covariates of age (< 70 or not), race, sex, marital status, tumor histology, grade, tumor location, lymphovascular invasion and perineural invasion. The PSM cohort was prepared to validate the impact of chemotherapy on CSS in this population with adverse histological features. All statistical analyses were performed using R3.6.2 software (http://www.r-project.org/) or STATA/SE 15.0 software (StataCorp LP, College Station, TX, USA), and a two-sided value of P < 0.05 was considered statistically signi cant.

Results
Totally  Table 1 presented the detailed clinicopathologic characteristics between the two groups.
Patients with adverse histological features were more likely to appear in women rather than men (53.74% in women compared with 46.26% in men, P < 0.01). Right side colon was a more frequent location lying the adverse histological features patients than left side colon (66.86% compared with 33.14%, P < 0.01). In reality, 22.89% of patients staging as T3N0M0 (AJCC stage II) were sequential deal with adjuvant chemotherapy, which was more likely to be administrated in patients of adverse histological features histology (27.65% in adverse histological features compared to 21.97% in good histological features, P < 0.01). No signi cant difference was showed in the characteristics of 1774 PSM patients (Table 1).

Discussion
In the present study, we performed a propensity score matching study based on population analysis to explore the survival bene t of chemotherapy on stage II CRC and described three major ndings: First, the adverse histological features of colonic cancer did not in uence the survival at all for patients with T3N0M0 staging. Second, adjuvant chemotherapy seemed to be invalid to improve the survival of patients of adverse histological features with T3N0M0 staging, so it may be reasonable to subtract chemotherapy due to the considerable toxicity; Third, there has been a contradictory behavior of the tumor location as left side colonic cancer appeared to be associated with a poorer survival in population with good histological features whereas it's well accepted that right side colonic cancer represents a poorer survival, and the in ection point may lie in patients of adverse histological features with T3N0M0 staging just as illustrated in Fig. 2. Females remain at a lower risk for advanced stage CRC across all age groups, and probably bene t from protective effect against colon cancer of estrogen pathway [9] and microsatellite instability arising in females [10].
The therapy of rectal cancer is routinely referred to that of colonic cancer, however, these two diseases show their respective situation frequently, such as different lymphatic drainage pathway, molecular subtypes and application of radiotherapy, so only colonic cancer were included in the current study to minimize the bias. In this study, we also adopted the PSM to eliminate the difference of baseline between patients of adverse histological features with and without chemotherapy. PSM is a technique to improve the comparability between groups, as shown by the well-balanced demographics and clinicopathological characteristics in Table 1. Both the lymphovascular invasion and perineural invasion are recognized as adverse factors responsible for poor prognosis, so these two histological features were well balanced by PSM method to invalidate the confusion, despite of the actuality that the speci c data was only on access from SEER database since year 2010.
The common adenocarcinoma (AC) accounts for the majority of colonic cancer cases, left 10-15% patients with mucinous adenocarcinoma (MC) and approximately 1% of CRC patients with SRCC [11]. Although SRCC represents a rare condition, it is associated with a poor prognosis compared with AC [11,12]. Five-year relative survival rate of SRCC in stage II colonic cancer patients was 67.9% compared with 77.0% of AC; while in stage II rectal cancer patients ve-year relative survival rate of SRCC was 27.4% compared with 67.1% of AC [11], indicating a signi cant in uence of tumor location on survival, which was similar with our ndings that the left side colonic cancer appeared to be associated with a poorer survival in population with good histological features opposite to the well accepted fact that right side colonic cancer represents a poorer survival in metastatic colorectal cancer. In concordance with SRCC, mucinous colorectal adenocarcinomas usually have poorer responses to chemotherapies [13], resulting in a poor prognosis in advanced disease [14]. Either MC or SRCC patients more often develop metastatic disease, and are more likely to develop peritoneal metastases and metastasizes via the lymphatic route, whereas AC metastasizes primarily to the liver [12], which may partially explains the poor prognosis of MC and SRCC. The standard treatments for colorectal adenocarcinoma are recommended to MC and SRCC patients since no clinical guidelines have been developed speci cally for these groups of patients [13]. Due to the low frequency, SRCC and MC has not been well-studied in stage II colonic cancer patients and reports in the literature are uncommon. In the present study, poorly differentiated histology, which was routinely considered as an adverse prognostic factor, was integrated as adverse histological features with MC and SRCC together to illustrate the bene t of adjuvant chemotherapy in stage II colonic cancer excluding T4. In our study population, 22.89% of patients staging as T3N0M0 were sequential deal with adjuvant chemotherapy, which was more likely to be administrated in patients of adverse histological features histology (27.65% in adverse histological features compared to 21.97% in good histological features, P < 0.01). So, this kind of adverse histological features was accepted for clinicians to evaluate the necessity of chemotherapy. However, it is denied to administrate adjuvant chemotherapy in the current study for patients of adverse histological features, which is regarded currently as one of the high risks, implying the adverse histological features alone is not su cient for the decision of adjuvant chemotherapy in T3N0M0.
Despite of the fact that inspiring therapeutic improvements have improved the survival rate in the last decade by an average of 3% every year [15,16], CC serves as the 3rd cause of cancer deaths both in the world, no matter men or women [9], with almost 25% of patients with stage II disease suffering recurrence after radical surgery [17,18]. Surgery is the main therapeutic modality for CRC and adjuvant chemotherapy is the widely accepted pattern of administration with promising e cacy for stage III CC, on the consensus that the pathologic stage is the most helpful prognostic factor for the colorectal cancer [3,4]. Regardless of the recommendation of the NCCN guideline, the adjuvant chemotherapy bene t for stage II CRC is still controversial [19], and it's believed that only a few CRC patients with stage II would bene t from adjuvant therapy [17,20]. Several stage-speci c factors in stage II CRCs, including T4 lesions, positive margins, poorly differentiated histology, inadequately sampled lymph nodes and et. al [21], which are considered as high-risk stage II CRCs, provide certain prognostic value [22]. The 5-year disease-free survival could be more than 90% in patients without any risk factor, compared to 84.8% in patients with one or more risk factors [23]. So far, adjuvant chemotherapy is administrated for high-risk stage II CRC or later stage CRC since 2004 [6], but there are insu cient data based on scienti c evidence to con rm the e cacy of the biomarkers to select the appropriate candidate patients [3,5] except that pathologic T4 stage was a signi cant factor predicting poor DFS [24], and only a trend for bene t from chemotherapy in the stage II was seen, failing to reach a signi cance [17]. Neither in the IMPACT trial nor the NSABP trial, improvement from adjuvant chemotherapy in survival of stage II CRC was seen [1], so the precise decision of whether a CC patient with Stage II disease should receive adjuvant chemotherapy is essential to improve the prognosis for every individual [19]. We came to the conclusion that adjuvant chemotherapy may be invalid to improve the survival even in patients of adverse histological features with T3N0M0 staging, so it may be reasonable to subtract chemotherapy for patients of this situation in case of the presence of adverse histological features alone, and more precise indicators are essential for the selection of advantage cohort.
Nowadays, several researches have been carried out to identify the sensitive CRC patients with AJCC stage II for chemotherapy. Tumoral heterogeneity including KRAS [17,25], BRAF, PTEN or microsatellite instability status [26], as well as tumor metabolism, such as NKX6.1 methylation [20], MMP7 [27] and AQP1 [28], has been established on cancer response to targeted therapy or chemotherapy in metastatic CRCs [9], while it is of no uncertain use for stage II CRCs. Her-2 status should be a predictive factor, especially for patients with pMMR status, because Her-2 positive patients can bene t from 5-uorouracial based adjuvant [29], while low expression of the PD-L1 gene was associated with bene t from adjuvant chemotherapy [30]. Despite of the efforts, there are now insu cient or reliable indicators to identify the proper patients suitable for adjuvant chemotherapy in stage II colonic cancer unfortunately, and the personalized intertumoral and genomic biomarkers would be promising in the future instead of clinicopathological indicators.
The current study had some limitations: First, as a retrospective analysis, it always carried the risk of various biases. Second, due to the carefully chosen criteria, only 887 colonic cancer patients of adverse histological features with T3N0M0 staging who were administrated with adjuvant chemotherapy were enrolled in this study, which would in uence the degree of power more or less. Third, the current SEER database is de cient in providing the entire picture of the registered patients such as genetic markers and disease-free survival (DFS), making it impossible to survey the in uence of adverse histological features and other genomic biomarkers on DFS or CSS. So, more chemotherapeutic regimens are necessary to increase survival in this stage II CRC subset of patients and more valid genomic biomarkers are essential to come up with appropriate selection criteria.
The strengths of our study include: The current study minimized potential biases and had a higher degree of power with the use of a large-scale sample and PSM method. The segmentation of the high risks of the stage II CRCs, and then the only concern on adverse histological features was another merit. The SEER database allows researchers to accurately identify patients with good or adverse histological features based on histology and grade clearly, thus avoiding confusion of differentiation, and we colligate histology and grade of tumors in a new way intend to make the results more convincing. Still, the potential biases of surgery were avoided by super-selective cohort with the limit of radical resection.

Conclusion
According to the ndings of our analysis of the SEER database, the adverse histological features does not in uence on survival for patients with T3N0M0 staging, and adjuvant chemotherapy is invalid to improve the survival of these patients, so adverse histological features alone was not su cient for decision of adjuvant chemotherapy in this situation rather than the recommendation of the guidelines. Figure 1 Page 18/19

Figures
The inclusion and exclusion criteria in seer.  Survival analysis of adjuvant chemotherapy in colonic cancer patients with T3N0M0 staging. A The cancerspeci c survival (CSS) between patients of good histological features with T3N0M0 staging with and without chemotherapy. B The CSS between patients of adverse histological features with T3N0M0 staging with and without chemotherapy. C The CSS between patients of adverse histological features with T3N0M0 staging with chemotherapy and without chemotherapy in PSM cohort. P value from the log rank test.