Prognostic implications of T stage in different pathological types of colorectal cancer: an observational study using SEER population-based data

Abstract

Objectives

Colorectal cancer (CRC) encompasses a spectrum of pathological types, each exhibiting distinct biological behaviours that challenge the conventional T-staging system’s predictive efficiency. Thus, this study aims to explore the prognostic significance of the T stage across various CRC pathological types, seeking to unravel insights that could enhance prognostic assessment in this complex disease.

Study design

We performed a retrospective analysis using the Surveillance, Epidemiology, and End Results (SEER) database for primary CRC cases from 2010 to 2017.

Setting

The SEER database, comprising data from various US regional and state cancer registries, identified 39 321 patients with CRC. Our analysis focused on the three most common CRC pathological types: adenocarcinoma (AC), mucinous adenocarcinoma (MC) and signet ring cell carcinoma (SR).

Primary outcome measures

The study used Cox regression models to evaluate how different pathological characteristics impact mortality risk in patients with CRC. Time-dependent receiver operating characteristic curves were also applied to assess the prognostic accuracy of various tumour node metastasis (TNM)/non-mucinous (NM) stages.

Results

We observed significant associations between T stage and mortality risk for patients with AC and MC. Notably, in comparison to those at T1 stage, patients with AC in the T4 stage demonstrated a 2.01-fold increase in mortality risk (HR=2.01, 95% CI: 1.89 to 2.15), while patients with MC at T4 stage showed a 1.42-fold increase (HR=1.42, 95% CI: 1.03 to 1.97). However, within the SR group, T stages did not independently impact survival, showing no significant distinction (HR=1.07, 95% CI: 0.59 to 1.95). Intriguingly, the traditional TNM staging systems demonstrated limited discriminatory power in predicting prognosis for patients with SR when compared with the more innovative NM staging systems.

Conclusions

This study uncovers important insights about the prognostic significance of the T stage in different types of CRC, highlighting the need for personalised assessments based on specific histological subtypes.

Strengths and limitations of this study.

• This research reveals key findings on the prognostic value of T stage across various colorectal cancer types, underscoring the importance of tailored evaluations for distinct histological subtypes.

• The lack of key data like tumour recurrence, targeted therapies and patient comorbidities in the Surveillance, Epidemiology, and End Results database limits our analysis depth.

• The limited cohort size of patients with signet ring cell tumour, particularly in early stages, raises statistical challenges in assessing tumour node metastasis staging for this group.

Introduction

Colorectal cancer (CRC) is a significant contributor to global cancer-related mortality, being the second most common cause of cancer deaths worldwide. Annually, there are approximately 1.8 million newly diagnosed cases of CRC, resulting in half a million fatalities.¹ The WHO has classified CRC into various histological types,² among which adenocarcinoma (AC) is predominant, constituting approximately 90% of all CRC cases.^3–5 The less common histological subtypes of CRC, mucinous adenocarcinoma (MC) and signet ring cell carcinoma (SR), represent approximately 10–20% and 1% of CRC cases, respectively.^{6 7}

Interestingly, the prognosis of CRC varies significantly across these histological subtypes. Patients with SR typically face a poorer prognosis compared with AC, while the prognosis for MC remains a subject of ongoing debate.^8–10 The American Joint Committee on Cancer (AJCC) has updated the tumour node metastasis (TNM) staging classification to the eighth edition, superseding the seventh edition as the definitive method for prognosticating CRC. It is widely regarded as the gold standard in this field.^{11 12} This comprehensive system examines three essential elements: the extent of tumour infiltration, the occurrence of lymph node metastasis and the presence of metastasis. Of these, the depth of tumour invasion, or the T stage, is a crucial prognostic factor for CRC. However, the prognostic value of T staging becomes intriguingly complex when considering the diverse pathological types of CRC. For instance, SR has been found to exhibit unique biological behaviour, which challenges the conventional association between T stage and prognosis in CRC.^{13 14}

In this study, we embarked on an in-depth exploration of the rich data obtained from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database, spanning from 2010 to 2017. Our objective was to elucidate the prognostic significance of T stage across different pathological types of CRC, with a view towards enhancing the tumour staging system for these patients.

Materials and methods

Patient and public involvement

Patients and the public were not involved in the development, design or analysis of this study.

Study participants

Within the SEER database spanning the years 2010 to 2017, a total of 65 739 patients with CRC were identified. A meticulous curation process was employed to exclude patients based on specific criteria, such as those diagnosed with T0 or Tis local disease, individuals with concurrent malignant tumours, cases with multiple primary CRC tumours, subjects with unknown survival time, pathological types outside the scope of AC, MC and SR, as well as individuals with an indeterminate T stage. This rigorous selection process resulted in a final cohort of 39 321 patients, comprising 36 184 AC, 2689 MC and 448 SR cases, thereby providing a robust and discerning dataset for analysis (figure 1).

Figure 1.

Flow chart of study participants. SEER, Surveillance, Epidemiology, and End Results

The cohort encapsulates a comprehensive clinical profile for each patient, integrating key variables such as age, sex, race, marital status, histological grade, tumour location, T, N and M stages, tumour TNM stage, carcinoembryonic antigen (CEA) levels, perineural invasion, liver, lung, bone, brain metastases, surgery, chemotherapy and radiotherapy. Tumour location was categorised into colon and rectum. The narrative also delves into the profound influence of surgical intervention and the strategic deployment of chemotherapy and radiotherapy, with each patient’s therapeutic trajectory meticulously documented to meet the stringent standards of scientific rigour. Surgery was considered performed if primary tumour destruction or excision was documented. Receipt of chemotherapy or radiotherapy was defined as having received these treatments during any phase of care. In cases where chemotherapy or radiotherapy were not administered, or if such information was unavailable, patients were classified as not having received these treatments, following the SEER database’s practice of combining 'No' and 'Unknown' categories.

The ‘NM’ staging system

In the context of patients with SR, the relationship between T stages and survival is shrouded in uncertainty, raising questions about the aptness of the conventional TNM staging system for predicting mortality risk in this subgroup. To refine our classification of participants with CRC, we used the non-mucinous ('NM') staging system. This system categorises Stage I as N0M0, Stage II as N1M0, Stage III as N2M0 and Stage IV as any N (N0, N1, or N2) with M1 (NxM1), where 'Nx' indicates any lymph node involvement status, but 'M1' consistently represents the presence of distant metastasis. To assess overall survival (OS), Kaplan-Meier analysis was employed to compare the outcomes between the TNM and NM staging systems. Cox regression models were also used to assess the relationship between the TNM or NM staging systems and mortality risk across different CRC pathological types. Additionally, the predictive power of the TNM/NM stages in forecasting survival was gauged using time-dependent receiver operating characteristic curves (time-ROC).

Statistical analysis

Age was represented as mean±SD and compared using one-way analysis of variance. The distribution and comparison of clinicopathological characteristics across the three pathological types of CRC were evaluated using frequency counts (n) and percentages (%), with the χ² test used to determine the significance. Cox regression models were used to examine the correlation between pathological types and the risk of mortality. Multivariate models included age, sex, race, marital status, tumour location, histological grade, T, N and M stage, CEA levels, perineural invasion, surgery, chemotherapy and radiotherapy. The primary measure of interest was the duration of survival, defined as the time from the diagnosis of cancer to either the occurrence of death or the study’s predefined endpoint. The SEER database provided vital survival time data, recorded in months, along with vital status indicators. Subgroup analyses were performed, stratifying patients by age, tumour site and sex, while multiplicative models were employed to investigate potential interactions between the T stage and these covariates.

Statistical analyses were executed using established software tools, including SAS software (V.9.4) and R software (V.4.0.4, http://www.R-project.org/). The threshold for statistical significance was set at a two-sided p value <0.05. These stringent statistical methodologies and criteria were adopted to ensure the robustness and reliability of our findings.

Results

In our comprehensive cohort of 39 321 patients with CRC, the distribution of histological subtypes was as follows: AC in 36 184 (92.0%) patients, MC in 2689 (6.8%) patients and SR in 488 (1.2%) patients. A detailed evaluation of demographic and tumour characteristics revealed significant differences among these groups (online supplemental table 1). Over a median follow-up of 42 months (range: 26–65), 14 265 patients succumbed to CRC. Our exploration of the association between different pathological types and overall mortality risk revealed that patients with SR faced a 1.34-fold increased risk of death compared with patients with AC (HR=1.34, 95% CI: 1.20 to 1.50). However, no significant difference in prognosis was observed between patients with MC and AC (HR=1.06, 95% CI: 0.99 to 1.14) (online supplemental table 2).

Table 1 presents the association between T stage and mortality risk across different pathological types. In patients with AC, T4 stage was associated with a 2.01-fold higher risk of death (HR=2.01, 95% CI: 1.89 to 2.15), while in patients with MC, it was associated with a 1.42-fold increased risk (HR=1.42, 95% CI: 1.03 to 1.97). However, among patients with SR, T stage did not independently impact survival, as indicated by an HR of 1.07 (95% CI: 0.59 to 1.95). Subgroup analyses further supported these findings, as detailed in figure 2.

Table 1.

The association of T stages with OS among three pathological types

	Adenocarcinoma		Mucinous		Signet ring cell
	Cases/person-years	Adjusted HR (95% CI)	Cases/person-years	Adjusted HR (95% CI)	Cases/person-years	Adjusted HR (95% CI)
T1	1853/30579	Ref.	44/447	Ref.	15/68	Ref.
T2	911/20202	0.93 (0.86 to 1.01)	46/1051	0.58 (0.38 to 0.89)	5/100	0.34 (0.12 to 0.98)
T3	6058/68602	1.29 (1.22 to 1.37)	472/5330	0.89 (0.65 to 1.24)	81/499	0.65 (0.35 to 1.18)
T4	3501/15513	2.01 (1.89 to 2.15)	519/2685	1.42 (1.03 to 1.97)	195/449	1.07 (0.59 to 1.95)
P for trend		<0.001		<0.001		0.003

Note: models were adjusted for age, sex, race, marital status, tumour location, grades, N, M surgery, carcinoembryonic antigen, perineural invasion, chemotherapy and radiotherapy.

OS, overall survival.

Figure 2.

Subgroup analyses of the association of T stage with the risk of death stratified by colorectal cancer histological subtypes.

Figure 3 illustrates the impact of advanced TNM stages on OS in patients with AC and MC. Early TNM stages were associated with a significantly more favourable OS, while advanced stages were associated with a poorer prognosis. However, among patients with SR, the distinction between stage I and II did not yield a discernible difference in OS. In contrast, the NM staging system demonstrated a progressive decline in OS with advancing stages, effectively capturing the dynamic nature of the SR subgroup.

Figure 3.

The Kaplan-Meier methods were used to compute the OS for patients with CRC of AC (A), MC (B) and SR (C) stratified by the TMN staging system and NM staging system. AC, adenocarcinoma; MC, mucinous adenocarcinoma; NM, non-mucinous; SR, signet ring cell carcinoma; TNM, tumour node metastasis.

Figure 4 presents an analysis of the time-dependent ROC curves, offering insights into the predictive capacities of different TNM/NM stages across the diverse pathological types of CRC. The traditional TNM staging system demonstrated good discriminatory power, particularly in AC. However, in the SR cohort, the area under the curve (AUC) of the TNM staging system was lower, suggesting that the NM staging system may be a superior prognostic indicator in patients with SR.

Figure 4.

The time-dependent ROC of TNM/NM staging system for AC, MC and SR. AC, adenocarcinoma; MC, mucinous adenocarcinoma; NM, non-mucinous; ROC, receiver operating characteristic; SR, signet ring cell carcinoma; TNM, tumour node metastasis.

Table 2 elucidates the relationship between the TNM/NM staging system and the risk of death in various pathological types. Patients with AC and MC faced an escalating risk of death with the progression of TNM/NM stages. Within the SR group, NM stage II was associated with a substantial 2.5-fold increase in the risk of death (HR=2.50, 95% CI: 1.45 to 4.30) compared with NM stage I. However, TNM stage II did not confer a heightened risk of death compared with TNM stage I in patients with SR, highlighting their unique characteristics. These findings underscore the need for tailored assessments and a more refined staging system for this rare and enigmatic subgroup.

Table 2.

The association of tumour stages defined by TNM and NM with OS among three pathological types

	Adjusted HRs (95% CI)
	Adenocarcinoma	Mucinous	Signet ring cell
TNM stages
I	Ref.	Ref.	Ref.
II	1.53 (1.43 to 1.64)	1.53 (1.14 to 2.06)	2.07 (0.79 to 5.45)
III	2.88 (2.69 to 3.08)	2.87 (2.13 to 3.88)	6.65 (2.65 to 16.69)
IV	9.99 (9.30 to 10.74)	8.80 (6.45 to 12.02)	21.05 (8.24 to 53.8)
NM stages
I	Ref.	Ref.	Ref.
II	1.90 (1.79 to 2.01)	1.6 (1.31 to 1.96)	2.50 (1.45 to 4.30)
III	3.26 (3.04 to 3.49)	2.93 (2.37 to 3.62)	4.52 (2.88 to 7.08)
IV	7.86 (7.42 to 8.33)	6.44 (5.32 to 7.79)	11.76 (7.33 to 18.85)

Note: models were adjusted for age, sex, race, marital status, tumour location, grades, surgery, carcinoembryonic antigen, perineural invasion, chemotherapy and radiotherapy.

NM, non-mucinous; OS, overall survival; TNM, tumour node metastasis.

Discussion

In this prospective cohort study, we have gained valuable insights into patients diagnosed with SR by using the SEER database. Our exploration has unveiled a marked propensity among patients with SR towards advanced tumour staging, coupled with significantly compromised survival outcomes. Intriguingly, the relationship between the T stage and prognostic outcomes in patients with AC and MC emerged as a pivotal determinant, while its autonomous influence on the survival of patients with SR remained ambiguous. Moreover, we have discerned constraints in the efficacy of the traditional TNM staging system in proficiently categorising patients with SR, suggesting the need for alternative or modified staging approaches in this unique subgroup.

In the present investigation, a robust analysis has demonstrated that patients with SR endure a substantially inferior OS relative to their MC and AC counterparts. These observations resonate seamlessly with a plethora of preceding studies, thereby fortifying the existing corpus of evidence.^{2 15–17} Notably, Lee et al, esteemed trailblazers in the realm of medical research, have posited that patients with SR suffer a markedly dismal survival prognosis when contrasted against patients with non-mucinous adenocarcinoma. Their pioneering efforts have illuminated a disconcerting reality, with patients with SR confronting a mere 46% cumulative survival rate over a modest 5-year period.¹⁶ Furthermore, an exhaustive study encompassing an extensive cohort of 244 794 individuals has highlighted the severe consequences of SR histology, showcasing a staggering 42%–57% elevated risk of mortality compared with non-mucinous, non-signet ring ACs.²

The TNM staging system by the AJCC is internationally recognised and serves as a crucial prognostic tool for predicting the survival outcomes of patients with CRC. Within this comprehensive schema, the T stage assumes a pivotal role as a crucial determinant.^18–21 Notably, the formidable T4 stage emerges as a critical factor that profoundly influences the prognosis of patients with CRC. A comprehensive study from the Netherlands involving 889 patients with colon cancer unveiled a disconcerting reality: those in the T4 stage confronted a fivefold elevated risk of mortality (HR=5.05, 95% CI: 2.73 to 9.37) compared with individuals in the T1–T2 stages.²² Similarly, a Chinese investigation underscored that patients with CRC with T4 tumours endured inferior survival outcomes when juxtaposed against those with stages T1–T3 tumours.¹⁹ Consequently, the T stage has evolved into a vital indicator, reflecting disease severity, guiding clinical decision-making, and even facilitating the prediction of relapse.

In harmony with existing evidence, our study elucidated the independent prognostic value of the T4 stage in patients with CRC with AC and MC. However, within the enigmatic realm of SR, the T stage did not demonstrate an association with the risk of mortality. This intriguing inverse relationship between the T stage and survival outcomes in patients with SR finds support in previous investigations. A Korean study concluded that survival in patients with SR did not correlate with T stages (all p values >0.05).¹³ Similarly, an analysis conducted using the renowned SEER database highlighted that the T stage did not independently predict adverse outcomes in patients with SR, regardless of whether the endpoint was OS or cancer-specific survival.¹⁴ This unexpected observation may be attributed to the unique cellular characteristics of SR. These cells typically manifest as single entities or loosely clustered formations, disrupting cell–cell adhesion and facilitating diffuse spread.¹³ Moreover, the growth pattern of colorectal SR involves infiltrative growth over a prolonged period without penetrating the mucosa.^23–25 The aggressive biological characteristics commonly observed in colorectal SR often result in advanced tumour stages being identified at the time of diagnosis.^{7 26 27} Furthermore, the distinctive metastatic pattern of SR is characterised by rapid dissemination from the serosa to the peritoneum.²⁸ Consequently, the prognostic significance of the T stage in patients with SR may not manifest overtly.

Our study has unveiled compelling insights, illuminating the predictive prowess of survival discrimination among distinct NM stage subgroups in patients with SR. Intriguingly, the NM staging system, when juxtaposed with the conventional TNM staging system and excluding the T stage, demonstrated superior prognostic efficacy in forecasting the outcomes of patients with SR. This suggests that the NM staging system may be a more fitting classification tool for this unique patient subgroup. Although SR is relatively rare, patients with this condition have typically been managed using the conventional TNM staging system, originally designed for the broader population of CRC cases. However, our findings question this approach, revealing that an advanced T stage does not unequivocally correlate with a heightened risk of mortality in patients with SR. As a result, the traditional TNM staging system falls short in accurately assessing the prognosis of patients with SR. Our study emphasises the need to acknowledge SR as a distinct disease entity, meriting separate consideration in clinical management. Moreover, we champion further research endeavours aimed at delving into the biological characteristics of SR and devising a more suitable staging system tailored to this rare pathological type.

The SEER database dynamically updates its TNM staging information to align with the latest editions of the AJCC staging guidelines. With the implementation of the eighth edition of the AJCC staging system in 2018, SEER adjusted its classification and reporting protocols for cancer cases accordingly. However, analyses of data spanning from 2010 to 2017 were conducted under the purview of the seventh edition, potentially bypassing the enhancements introduced in the subsequent edition. To address this, we extended our investigation to include SEER data from 2018 to 2021, aiming to elucidate the prognostic utility of NM staging across various pathological subtypes of CRC. Our findings corroborate earlier observations, revealing a diminished AUC for the TNM staging system within the SR cohort. This suggests the NM staging system’s superior prognostic value for patients with specific rectal cancers, underscoring its potential as a more effective tool in prognostic stratification (online supplemental figure 1).

The current investigation stands on the robust foundation of a comprehensive dataset, encompassing a significant cohort of patients with CRC. This wealth of information not only furnishes a more precise and holistic understanding of the clinicopathological features and prognosis among varied histological subtypes but also marks a pioneering stride in probing the prognostic value of the T stage and the conventional TNM staging system across these distinct CRC subgroups. However, despite the robustness of our findings, it is crucial to recognise the inherent limitations that moderate our conclusions. First, the modest cohort of 448 patients with SR tumour in our study, particularly those with early-stage tumours, introduces the risk of a type II error and statistical limitations, especially in evaluating the traditional TNM staging for patients with SR. This under-representation and potential underpowering necessitate caution in generalising our findings for the SR subgroup. The lack of statistical significance in T staging’s prognostic impact for patients with SR warrants careful interpretation. Future research with a more extensive and diverse patient population with SR is essential to thoroughly understand and validate these prognostic implications. Second, while the SEER database is a valuable resource, it unfortunately lacks key variables such as tumour recurrence data, targeted molecular therapies, microsatellite instability status, patient comorbidities and family history. The omission of these pivotal factors curtails the depth of our analysis and underscores the necessity for comprehensive datasets in future research endeavours. Third, the decentralised collection of pathological data, conducted by a myriad of pathologists and diverse healthcare institutions, introduces inherent variations and potential biases that may sway the interpretation of our findings. Lastly, our study focuses on AC, MC and SR in CRC, selected for their prevalence and clinical importance. However, by not including rarer pathological types, our findings have limited generalisability and may miss aspects of disease heterogeneity. Future research should explore a broader array of pathologies to enhance understanding and applicability to diverse patient populations.

Conclusions

This prospective cohort study underscores the unique prognostic implications of T stage across different histological subtypes of CRC. It is evident that patients with SR endure the most unfavourable survival outcomes, while the prognostic relevance of T stage exhibits variation across AC, MC and SR subtypes. These observations accentuate the necessity for individualised assessment and refinement of the tumour staging system, with the aim to enhance prognostic precision for patients with CRC.

Data availability statement

Data are available in a public, open access repository. The data were abstracted from the Surveillance, Epidemiology, and End Results (SEER) database. This is an open database. (https://seer.cancer.gov).

Ethics statements

Patient consent for publication

Consent obtained from parent(s)/guardian(s)

Ethics approval

This study was approved by Ethics Board of the Aerospace Center Hospital with the reference number of 200605. Participants gave informed consent to participate in the study before taking part.