Background

Neuroendocrine neoplasms (NENs), originating from peptidergic neurons and neuroendocrine cells, are a heterogeneous group of rare tumours with distinct clinical behaviours [1]. According to the World Health Organization (WHO) classification of 2010, neuroendocrine carcinomas (NECs) are defined as poorly differentiated, high-grade malignant neoplasms composed of small cells or intermediate to large cells. The current grading system for NEC classification is based on the Ki-67 proliferation index or mitotic count. For the diagnosis of NEC, the patient should have a Ki-67 index or mitotic count higher than 20%. If the mitotic count and Ki-67 index are different, the higher of the two is used [2]. The incidence of NEN has been rising in recent decades [3,4,5,6]. Only approximately 5% of all gastrointestinal NENs have a Ki-67 index higher than 20% [7, 8]. This frequency might differ by organ, at approximately 7% in the pancreas [8] and up to 40% in the colon [9]. NEC of the colon and rectum (CRC-NEC) is rare, comprising less than 1% of colon and rectal cancers [10].

NEC has a poor prognosis, and the survival ranges from 38 months for patients with localized disease to 5 months in the metastatic setting [11]. The median survival of metastatic NEC patients could be as short as 1 month for patients who received only best supportive care and up to 12 to 19 months for those treated with best available therapy [6, 12]. It has been reported that only 5% of NEC patients are long-term survivors [8]. The median overall survival of CRC-NEC patients is 13.2 to 14.7 months [13, 14], with 5-year overall survival (OS) rates of 8 to 16.3% [14, 15]. However, previous publications are limited to small retrospective series [10].

Due to the low incidence and the lack of large randomized clinical studies of CRC-NEC, the prognostic factors and survival of CRC-NEC remain unclear. Therefore, the objective of this study was to investigate the clinicopathologic characteristics, prognostic factors, and treatment outcomes of patients with CRC-NEC using data from the Surveillance, Epidemiology, and End Results (SEER) database.

Methods

Study population

The best way to explore the prognostic factors of a rare disease, such as CRC-NEC, is to conduct a large population-based study. In this study, we identified individuals diagnosed with microscopically confirmed adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma (non-NEC) or CRC-NEC of the colon or rectum from the SEER database during the period from 2004 to 2013. The SEER dataset is supported by the US National Cancer Institution and is an authoritative source of information on cancer statistics from the population-based cancer registries, which covers 28% of the US population [16, 17]. It is publicly accessible for studies in cancer epidemiology and health policy, and SEER Stat software (SEER Stat 8.3.6) was used to access the data (reference number: 12296-Nov2018). Patients diagnosed after 2013 were excluded from the study to ensure adequate follow-up. Data regarding clinicopathological variables, such as age and year at diagnosis, marital status, race, sex, tumour grade, tumour type, tumour site, adjuvant chemotherapy, insurance status, carcinoembryonic antigen (CEA), tumour deposit, peritoneum invasion, survival time, and cause of death, were retrieved.

The specific inclusion criteria were as follows: (1) patients were diagnosed from 2004 to 2013; (2) site record ICD-O-3 was limited to the colon and rectum with pathologically confirmed diagnosis (C18.0–18.7, C 18.9, C 19.9, C 20.9); (3) histological type ICD-O-3 was limited to large cell neuroendocrine carcinoma, small cell carcinoma, neuroendocrine carcinoma, adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma (8013/3, 8041/3, 8246/3, 8140/3, 8480/3, 8490/3); (4) TNM staging was clear (staging based on 6th or 7th American Joint Committee on Cancer (AJCC) system according to the version of the day); and (5) patient age was limited to 20 to 74 years old. The exclusion criteria were as follows: (1) patients without information of age or race at diagnosis; (2) patients with multiple primary tumours; (3) patients with survival time less than 1 month; and (4) patients with low-grade (well or moderately differentiated) neuroendocrine carcinoma [18] (Fig. 1).

Fig. 1
figure 1

The inclusion and exclusion criteria in the SEER database

Full size image

Variable declaration

Age was grouped as less than 60 years old (< 60) and more than 60 years old (≥ 60). Race was divided into black, white, and other. Marital status was grouped as married, single, and divorced. Histology was grouped as colorectal neuroendocrine carcinoma (CRC-NEC) and common colorectal adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma (non-NEC). Differentiation grades were divided into well, moderately, and poorly differentiated and undifferentiated. Tumour location was grouped as right-side colon and left-side colon, whose cut-off was the splenic flexure, as well as rectum. Operation methods were grouped as no surgery, local excision, radical resection, and unknown. Insurance status was grouped as insured, any Medicaid, uninsured, and unknown. Chemotherapy was grouped as yes or no, regardless of the sequence. CEA, tumour deposit, and peritoneum invasion were all grouped as negative, positive, and unknown.

Statistical analyses

Survival curves were generated using the Kaplan-Meier method, followed by the log-rank test for detecting the differences in survival. The duration from the date of diagnosis until death due to cancer was calculated as the cancer-specific survival (CSS), and CSS was the primary endpoint in this study. Death due to other causes was defined as a censored observation. A competing-risks regression model was applied to conduct the multivariate analysis with backward, stepwise elimination of variables. Hazard ratios (HRs) were adjusted for age (< 60 or ≥ 60), sex, race, marital status, tumour location, T classification, N classification, staging, operation, chemotherapy, insurance status, CEA, tumour deposit, and peritoneum invasion. Variables that were believed to be associated with CSS or those with P < 0.20 in the univariate analysis were enrolled in multivariate analysis. The demographic and clinical characteristics of the two cohorts are summarized in Table 1.

Table 1 Descriptive characteristics of 67,484 patients of study population within the SEER Medicare-linked database and 688 propensity score-matched patients
Full size table

Moreover, propensity score matching (PSM) analysis was performed to balance the differences for known potential baseline confounding variables in the populations with CRC-NEC and signet ring cell cancer (SRCC), which is considered a rare type of colorectal cancer with an extremely poor prognosis. The MatchIt R package was used to perform bipartite-weighted propensity score matching with stratification for stage, operation method, and chemotherapy [19]. Characteristics such as grade, CEA, tumour deposit, and peritoneum invasion, which did not have a counterpart between the two groups, were excluded from the PSM procedure. Thereafter, baseline variables of the two groups were compared by a stratified logistic regression including covariates of sex, race, age, marital status, staging, grade of differentiation, T classification, N classification, tumour site, operation method, chemotherapy, and insurance status. Finally, survival curves were generated using the Kaplan-Meier method in the two PSM cohorts. The adjusted cohort was used to validate that CRC-NEC has a poorer prognosis than SRCC. All statistical analyses were conducted using R3.6.2 software (http://www.r-project.org/) or STATA/SE 15.0 software (Stata Corp LP, College Station, TX, USA), and a two-sided value of P < 0.05 was considered statistically significant.

Results

A total of 67,484 patients were identified from the SEER database based on the aforementioned inclusion criteria. Among all the identified patients, 344 (0.51%) were CRC-NEC patients, and 61,057 (99.49%) were non-NEC patients. The detailed clinicopathological characteristics of the two groups are presented in Table 1. The median ages of the patients in the CRC-NEC and non-NEC cohorts were 58 (range, 24 to 74 years old) and 60 (range, 20 to 74 years old), respectively. The proportion of female patients was 43.60% (n = 150) in the CRC-NEC group and 44.16% (n = 29,651) in the non-NEC group (P = 0.835). Advanced stage, increased depth of invasion, and greater risk of lymph node metastasis were more common in CRC-NEC patients than in non-NEC patients (P < 0.001). The percentage of lymph node metastasis (LNM) was 75.29% (n = 259) in CRC-NEC and higher than 51.53% (n = 34,600) in non-NEC (P < 0.001). In the CRC-NEC cohort, N0 was found in 24.71% (n = 85), while N0 accounted for 48.47% (n = 32,540) of patients in the non-NEC cohort. Patients with stage III/IV disease accounted for a significantly larger proportion of CRC-NEC patients than non-NEC patients (86.63% vs. 55.94%, P < 0.001). The majority of CRC-NECs (56.40%) were located on the right side of the colon, which was different from the distribution of non-NECs (P < 0.001). Chemotherapy is the main treatment for CRC-NEC (66.18%) and non-NEC (55.50%). The median CSS of CRC-NEC patients with chemotherapy was 10.0 months (interquartile range [IQR], 4.0 to 30.0 months), while the median CSS of patients without chemotherapy was 8.0 months (interquartile range [IQR], 2.0 months to not reached [NR]). There were still several significant differences in the characteristics of the 688 propensity score-matched patients despite the effort (Table 1).

Table 2 summarizes the results of the univariate analysis and multivariate competing-risks regression analysis of the predictors of OS in CRC-NEC patients. In the univariate analysis, age, sex, race, marital status, tumour site, chemotherapy, insurance status, CEA, tumour deposit, and peritoneum invasion were not significantly associated with a poor prognosis in CRC-NEC. However, advanced N classification (N1: HR, 2.08, 95% CI, 1.44–3.02, P < 0.001; N2: HR, 2.11, 95% CI, 1.48–3.02, P < 0.001) and stage IV (HR, 3.91, 95% CI, 1.56–9.81, P = 0.004) were associated with a poor prognosis in CRC-NEC (Table 2). The results of the multivariate analysis indicated that male sex (HR, 1.30, 95% CI, 1.01–1.68, P = 0.042), left-side colon location (HR, 1.44, 95% CI, 1.01–2.06, P = 0.043), and stage IV (HR, 4.67, 95% CI, 1.65–13.22, P = 0.004) had a significant negative effect on the survival of CRC-NEC patients. However, only radical surgery (HR 0.58, 95% CI 0.39–0.87, P = 0.007) and chemotherapy (HR 0.59, 95% CI 0.40–0.87, P = 0.008) had a positive significant influence on the survival of CRC-NEC, as illustrated by the multivariate analysis.

Table 2 Univariate and multivariate analyses of predictors of CRC-NEC in the study population
Full size table

The CSS is illustrated in Figs. 2, 3. There was a significant difference between CRC-NEC and non-NEC patients in terms of cancer-specific survival (P < 0.001) (Fig. 2a). Moreover, there was a significant difference in CSS between CRC-NEC, adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma patients (P < 0.001) (Fig. 2b). Figure 3 shows that there was a significant difference between CRC-NEC and SRCC patients in terms of CSS (P < 0.001) in both the raw cohort and the PSM cohort. Kaplan-Meier survival curves illustrate how overall mortality changed with histology. CRC-NEC had the poorest CSS among the assessed tumour types (median CSS, 9.0 months), even poorer than SRCC (median CSS, 24.0 months). Interestingly, neither T classification nor N classification, which were grouped according to the TNM staging system of the colorectal carcinoma of the day when the disease was diagnosed, could well predict the CSS (Table 2, Fig. 4). Figure 5 demonstrates how the operation method and chemotherapy influence the CSS of CRC-NEC (Fig. 6).

Fig. 2
figure 2

Survival analysis of CRC-NEC and non-NEC patients in the entire cohort. a CSS between CRC-NEC and non-NEC patients. b The CSS between CRC-NEC, adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma patients

Full size image
Fig. 3
figure 3

Survival analysis of CRC-NEC and SRCC patients. a CSS between CRC-NEC and SRCC patients in the raw cohort. b CSS between CRC-NEC and SRCC patients in the PSM cohort. c CSS between CRC-NEC and SRCC patients of stage III disease in the raw cohort. d CSS between CRC-NEC and SRCC patients of stage IV disease in the raw cohort

Full size image
Fig. 4
figure 4

Predictive value of T and N classification for prognosis of NEC. a Predictive value of T classification for prognosis of NEC. b Predictive value of N classification for prognosis of NEC

Full size image
Fig. 5
figure 5

CSS benefit of radical surgery and chemotherapy for CRC-NEC. a CSS benefit of radical surgery for CRC-NEC. b CSS benefit of chemotherapy for CRC-NEC

Full size image
Fig. 6
figure 6

Survival analysis of histological types in the NEC cohort

Full size image

Discussion

CRC-NECs are rare and extremely aggressive tumours, accounting for approximately 0.6% of all colorectal cancers [10], which is in agreement with our results. According to histogenesis, NEC is a neuroendocrine tumour (NET). It lacks specific clinical manifestations but shows a higher degree of malignancy and a poorer prognosis than non-NEC CRC. Poorly differentiated NEC usually leads to a poor prognosis [20]. NEC is defined both by proliferative activity and morphology, but it is reported tumour’s morphology was more important to predict sensitivity to chemotherapy [21]. Due to limitations in the current grading system for neuroendocrine neoplasms, high-grade neuroendocrine carcinoma, which comprise a subgroup of differentiated neuroendocrine tumours (NET) with high proliferation rates (G3-NET), as well as undifferentiated G3-NEC (large or small cell carcinomas), is difficult to define in case of the lack of information on Ki-67 or proliferation rate in seer database. Moreover, G3-NET is a different subtype with a much lower response rate to chemotherapy compared to NEC [21]. So, we enrolled only NEC in the current study to avoid the bias brought by G3-NET. Meanwhile, the ICD-10 code C 18.1 (malignant Tumour of the Appendix) was also included into the search algorithm, raising a doubt that a better prognosis of appendix NECs may probably explain the better prognosis of right-sided NECs as a whole. However, only 816 patients were actually diagnosed as malignant tumours of the appendix from the SEER database initially, and the majority were non-NEC, with only 2 patients diagnosed as large cell neuroendocrine carcinoma and 1 patient diagnosed as small cell neuroendocrine carcinoma. After the data arrangement, no appendix NEC was finally included in the current study, left the hypothesis that the appendix may not be a frequent localization for NEC, despite of the frequent neuroendocrine tumours of all grades.

In our study, we found that the median CSS of CRC-NEC was only 9.0 (interquartile range [IQR], 4.0–30.0) months, similar to that reported by previous studies [6, 12, 18]. The CSS of CRC-NEC was significantly poorer than that of adenocarcinoma and signet ring cell carcinoma based on the Kaplan-Meier survival curves (P < 0.001), and this was confirmed in PSM cohort. Unfortunately, after propensity score matching, which was performed by stratification for staging, operation method, and chemotherapy, there was still a statistically significant difference in characteristics as shown in Table 1, mainly in staging and operation methods. Any effort to balance the characteristics seemed to be a vain attempt, mainly due to the inherent difference between the two types of tumours. As shown in Table 1, patients with stage IV disease accounted for a significantly larger proportion of CRC-NEC patients than non-NEC patients (60.47% vs. 20.40%, P < 0.001), indicating CRC-NEC as a more aggressive tumour with a poor prognosis. Furthermore, CRC-NEC patients had less opportunity to accept radical operation than non-NEC patients (72.10% vs. 93.03%, P < 0.001). In our study, 86.63% of CRC-NEC patients were diagnosed in stage III or IV, and surgical resection was the primary treatment for CRC-NEC patients. However, for patients with metastatic CRC-NEC, a previous study reported that patients would not benefit from surgery and that surgery was not a prognostic factor for CRC-NEC [22]. Moreover, a retrospective study of 126 high-grade CRC-NEC patients reported that primary tumour resection did not improve the survival outcome [13]. In addition, a recent population-based study found that operation is not a prognostic factor in early-stage small cell CRC-NEC [15]. This result again proved that local surgery brought no benefit in terms of survival (HR, 0.89; 95% CI, 0.23–3.47; P = 0.867). However, radical surgery seemed to improve the prognosis, indicating that radical surgery is a necessary treatment, which was in agreement with the results of Fields et al. [18].

Efficient chemotherapy treatment for CRC-NEC has yet to be achieved [20, 23]. Our results showed that chemotherapy improved the 5-year cancer-specific survival of CRC-NEC in the univariate analysis (HR, 0.59; 95% CI, 0.40–0.87; P = 0.008); however, this finding was not significant in the multivariate analysis (HR, 1.22, 95% CI, 0.87–1.70; P = 0.252), which may be attributed to the insufficient number of samples and confounding factors. This finding agrees with results reported by a previous study that also indicated that improvement of the 5-year cancer-specific survival of CRC-NEC by chemotherapy treatment was not satisfactory [22]. The role of adjuvant chemotherapy needs further evaluation so that we will be able to better understand the overall effectiveness of chemotherapy treatment for patients with CRC-NEC, as shown in Fig. 5c that adjuvant chemotherapy brought a tendency of survival benefit without a statistic significance. Therefore, further prospective research is needed to shed light on the exact role of adjuvant chemotherapy in a large population. Recently, studies have shown that immunotherapy is a promising treatment for different types of cancers. Moreover, the relationship between NEC of the digestive system and the PD-L1 protein has been accounted for, indicating that immunotherapy may become a new treatment direction for CRC-NEC [24].

In our study, we found that the proportion of LNM and patients with stage III/IV disease was greater in CRC-NEC than in non-NEC CRC (P < 0.001). The LNM and stage of cancer are closely related to prognosis [4]. Therefore, it would be better to diagnose NEC earlier for an improved prognosis. Studies revealed that NEC did not benefit from advances in the prevention and treatment of colorectal adenocarcinoma over the past decade [15], which could be attributed to the fact that we did not find the tumour at early stages. Currently, the tumour, lymph node, metastases (TNM) staging system in colorectal neuroendocrine carcinoma is referred to that of colorectal adenocarcinoma; there have been a few studies assessing the value of current TNM classification. Actually, it is incredible indeed that patients with T3 CRC-NEC have the best survival demonstrated in Fig. 4, and one probable hypothesis is that patients with T3 CRC-NEC were the cohort who were more likely to accept radical surgery. The extraordinary finding was still probably or partially due to the small number of patients with T1 and T2. We stratify by T and N classification in Fig. 4 to call for the a modification of the staging system of CRC-NEC, as the current staging system may be impropriate that both the T and N classification cannot predict prognosis precisely despite of limited sample, although the conclusion may not be unshakable.

Our study showed that left-side colon NEC was an independent adverse prognostic factor for CRC-NEC (P = 0.043) in the multivariate analysis; however, left-side colon location in adenocarcinoma is a favourable prognostic factor accepted in routine work. There is still a lack of research about the difference in NEC in the left- and right-side colon. A previous study reported differences in the molecular biological characteristics, pathological features, clinical manifestations, and survival prognosis of left- and right-side colon adenocarcinoma [25]. Therefore, we think that the different sensitivity to chemotherapy drugs in the left- or right-side colon could be one of the reasons why chemotherapy is an important part of treatment. Therefore, we suggest that the location of NEC could be a predictive marker for the prognosis of patients with NEC.

There are some limitations to our study. First, chemotherapy or other antitumour drugs affect the prognosis of patients. However, the SEER database lacks specific information on patient medications. Second, the SEER database lacks mitotic images and Ki-67 index-relevant information. Finally, this study is only a retrospective analysis of clinical data, and the number of NECs was small. Further study with a larger sample is desired for adequate evaluation.

Conclusions

NEC is a rare and extremely aggressive tumour with a poor prognosis. Patients with NEC on the right side of the colon have a better prognosis than those with NEC on the left side of the colon. Early diagnosis, radical surgery, and chemotherapy are imperative. Further investigations are necessary to understand the differences in behaviour based on tumour morphology and to develop novel targeted therapies to improve the outcomes of rare and aggressive NEC.