Evaluation of Risk Factors,and Development and Validation of Prognostic Prediction Models for Distant Metastasis in Patients With Rectal Cancer: A Study Based on the SEER Database and a Chinese Population

Abstract

Purpose

This study aimed to construct a prediction model regarding risk factors and prognostic factors for distant metastasis of T1-T3 stage rectal cancer. For this purpose, a population-based retrospective cohort study was conducted.

Methods

Data on 7872 patients diagnosed with rectal cancer between 2004 and 2020 were obtained from the Surveillance, Epidemiology, and End Results database, of whom 746 had distant metastases at diagnosis. Independent risk factors for distant metastasis of rectal cancer were determined using univariate and multivariate logistic regression analyses. Cox proportional hazards regression analyses clarified the independent prognostic factors for distant metastases of rectal cancer. A 7:3 randomization process was used to place all patients into the training and internal validation groups. Furthermore, we retrospectively collected clinical data from 226 patients who had both rectal cancer and distant metastases between 2012 and 2024 at the Weifang Hospital of Traditional Chinese Medicine. We used the calibration curve, DCA curve, C-index, and area under the curve (AUC) to assess the discriminatory and pre-precision qualities of the models.

Results

The multivariate logistic regression analysis identified race, tumor grade, T stage, N stage, radiotherapy, chemotherapy, surgery, tumor size, and histological subtype as risk factors for distant metastases in rectal cancer, with AUC values for both training and validation sets exceeding 0.8. Using Cox regression analysis, we determined that the age, sex, tumor size, surgery, chemotherapy, and radiotherapy were independent predictors of distant metastasis of rectal cancer. In the prognostic model, the C-index of the training cohort was 0.687 (95% CI: 0.6615-0.7125), that of the internal validation cohort was 0.692 (95% CI: 0.6508-0.7332), and that of the external validation cohort was 0.704 (0.6785-0.7295).

Conclusion

Our nomogram can predict risk factors and analyze the 1-, 2-, and 3 year prognosis of distant metastases in patients with rectal cancer, providing valuable guidance for future clinical work.

Plain language summary

Purpose: This study aimed to construct a prediction model regarding risk factors and prognostic factors for distant metastasis of rectal cancer with T1-T3 stage. Methods: Data of patients diagnosed with rectal cancer between 2004 and 2020 were obtained from the Surveillance, Epidemiology, and End Results database. Independent risk factors for distant metastasis of rectal cancer were determined using univariate and multivariate logistic regression analyses. Cox proportional risk regression analyses clarified the independent prognostic factors for distant metastases of rectal cancer. A 7:3 randomization process was used to place all patients into the training and internal validation groups. Furthermore, as part of the validation cohort, we retrospectively collected clinical data from 226 patients who had both rectal cancer and distant metastases between 2012 and 2024 at the Weifang Hospital of Traditional Chinese Medicine. We used the calibration curve, DCA curve, C-index, and area under the curve (AUC) to assess the discriminatory and pre-precision qualities of the models. Results: The multivariate logistic regression analysis revealed that the race, tumor grade, T, N, radiotherapy, chemotherapy, surgery, tumor size, and histological subtype were among the risk factors for distant metastases in rectal cancer, and the AUC values for both the training and validation sets in the risk model were greater than 0.8. Using Cox regression analysis, we determined that the age, sex, tumor size, surgery, chemotherapy, and radiotherapy were independent predictors of distant metastasis of rectal cancer. In the prognostic model, the C-index of the training cohort was 0.687 (95% CI: 0.6615-0.7125), that of the internal validation cohort was 0.692 (95% CI: 0.6508-0.7332), and that of the external validation cohort was 0.704 (0.6785-0.7295). Conclusion: Our nomogram can predict risk factors and analyze the 1-, 2-, and 3 year prognosis of patients with metastatic rectal cancer, providing valuable guidance for future clinical work.

Keywords

rectal cancer distant metastasis factual database prognosis logistic models risk factors nomogram

Introduction

The metastatic tumor cell (“seed”) vs its organ microenvironment (“soil”) hypothesis has been widely used to explain the metastatic spread of tumors studying the tumor-stroma interactions is possible at the molecular level.^1,2 Specific tumor cells may have a propensity to metastasize to particular target organs; the most common metastatic target organs are the bone, brain, liver, and lungs.³ The incidence of rectal cancer has been increasing year after year due to the insidious nature of its onset and the lack of clinically specific indicators, among others, making it the second most common cause of cancer deaths.^4,5 In recent years, with the advent of novel targeted agents and immunotherapies, new insights into the treatment of this population have been provided, reducing the incidence of rectal cancer and leading to an improved prognosis. Nonetheless, patients with stage I–II rectal cancer have a 5-year survival probability of 88%-95%, whereas those with metastatic rectal cancer have a survival span of 3-5 years, and approximately 60% of patients with metastatic colorectal cancer die in 1-2 years.⁶ Tumor metastasis is the primary cause of cancer-related mortality, and most current treatments have not yet been able to provide a lasting therapeutic effect.⁷

We focused on cancer metastasis because 90% of cancer-related deaths are caused by metastasis rather than the primary tumor.⁸ Metastatic disease poses significant therapeutic challenges, especially as peritoneal spread reduces the quality of life and discourages outcomes. Tumor metastasis is a complex process. Through gene regulation, tumor cells spread from their original location and progressively develop into new tumors, producing intractable consequences and diminishing therapeutic efficacy.⁹ The primary colorectal cancer metastatic pathway is lymphatic metastasis, and Fan et al found that FAP-α in tumor cells induces extracellular matrix remodeling and establishes an immunosuppressive environment through the recruitment of regulatory T cells, which promotes lymph node metastasis of rectal cancer.¹⁰ Furthermore, the liver is the most frequently metastasized organ in patients who develop metastases following primary resection of rectal cancer.^11,12 If occult metastases are present at the time of resection, intrahepatic recurrence of future liver remnant (FLR) cannot be detected using standard imaging. All liver tumors must be removed with enough FLR for liver regeneration.^13,14 Early diagnosis is necessary for patients at high risk of distant metastasis from rectal cancer, which helps improve the prognosis of rectal cancer patients.¹⁵ Therefore, practical tools are urgently required to guide in effective clinical diagnosis and treatment of such patients.

A nomogram can provide predictive models based on numerical estimates for cancer prognosis and recurrence.^16-18 A concise and clear presentation can help clinicians make decisions and provide guidance.¹⁹ Several cancers are caused by multiple clinical factors, and TNM staging systems cannot provide personalized guidance. Therefore, nomograms have greater advantages over TNM staging systems and are gradually becoming the new standard.^20,21 To the best of our knowledge, no model has predicted the overall survival (OS) of patients with rectal cancer with distant metastases. Therefore, we collected and screened a large amount of clinical data on rectal cancer combined with distant metastases from the Surveillance, Epidemiology, and End Results (SEER) database and used the clinical data of patients from Weifang Hospital of Traditional Chinese Medicine as external validation. By retrospectively analyzing these data, we constructed and validated the risk factors and 1-, 2-, and 3 year prognostic prediction models for such patients.

Materials and Methods

Patients

The study was a population-based retrospective cohort study. Data on 7872 patients diagnosed with rectal cancer between 2004 and 2020 were obtained from the Surveillance, Epidemiology, and End Results database, of whom 746 had distant metastases at diagnosis. Additionally, we retrospectively collected data on 226 patients with rectal cancer combined with distant metastases, diagnosed at the Weifang Hospital of Traditional Chinese Medicine from 2012 to 2024, as an external validation cohort for our study. The inclusion criteria were as follows: (1) patients diagnosed with a primary tumor of rectal cancer with a definite site of metastasis between 2004 and 2020; (2) Rectal cancer patients with T-stage of T1-3 were selected; (3) a reporting source that was neither an autopsy nor a death certificate; (4) patients with full survival and follow-up data; and (5) patients with complete demographic and clinicopathological information such as TNM stage and tumor size. The screening flowchart is shown in Figure 1, which depicts 7872 patients with rectal cancer, including 746 patients who experienced distant metastases.

Figure 1.

A flowchart showing patient screening details.

Our study methods were strictly based on the research criteria published in the SEER database. This study obeyed the Declaration of Helsinki and was approved by the Medical Research Ethics Committee of Weifang Traditional Chinese Medicine Hospital (approval number: 2024YX152). In accordance with national legislation and institutional requirements, we have obtained verbal informed consent from the patients.‘Informed consent’ is a core principle that ensures that the rights and interests of the subject (patient) are respected and protected. With the help of a combination of an electronic medical records system and telephone follow-up visits, we meticulously documented the patients’ OS to apply it in further data analysis and evaluation. Our exclusion criteria were: patients based only on autopsy registration or death certificates, zero or unknown number of months of survival, and unclear or lack of relevant clinical data (e.g., TNM stage, pathological staging, tumour size, ethnicity and degree of differentiation). The reporting of this study conforms to TRIPOD guidelines.²²

Data Collection

The following variables were collected to clarify the risk factors for developing distant metastases in rectal cancer: age at diagnosis, race, sex, histological subtype, tumor grade, primary site, TN stage, surgery, radiotherapy, chemotherapy, tumor size, and occurrence of metastases. Age was analyzed as a continuous variable with original data. We selected 65 years as the age cut-off point because it is an important criterion for delineating the elderly population and is widely used in policy making and healthcare planning. In this study, the number of patients <65 and ≥65 years of age was approximately equal, which ensured a comparable and balanced study and also helped us to effectively identify those who require more frequent medical monitoring and intervention. Race was divided into (American Indian/Alaska Native、Asian or Pacific Islander、Black、White)、Measurements regarding tumour size were based on imaging findings (e.g. CT, MRI, etc.) and pathological assessment. Patients (55684 cases) who were based only on autopsy registrations or death certificates (205 cases), with zero or unknown number of months of survival (1980 cases), and with ambiguous or lack of relevant clinical data (e.g., TNM staging, pathological staging, tumour size, ethnicity, and degree of differentiation) were also excluded according to the exclusion criteria. Patients in the diagnostic group with distant metastases comprised the training and validation groups of patients in the prognostic cohorts. A nomogram was created for each cohort using the training set patients and validated with the matching validation set patients to identify the risk factors for patients with rectal cancer with distant metastases. Fundamental clinical characteristics were chosen for this investigation. In addition, survival analysis was performed to explore prognostic factors in patients with rectal cancer combined with distant metastases.

Statistical Analyses

All variables were analyzed using univariate logistic analysis, and variables significantly associated with rectal cancer were identified using multivariate logistic analysis to construct a prediction model for the occurrence of distant metastasis in patients with rectal cancer. We analyzed the survival of 746 patients who developed metastases to determine the prognostic factors. A training group (n = 522) and an internal validation group (n = 224) were randomly assigned to all patients in a 7:3 ratio. The grouping method was simple random grouping. Randomization was computer-generated by using the random () function in R language. To identify independent predictive factors for rectal cancer distant metastases, we ran one-way Cox proportional hazards regression analysis for all included variables and included variables with P < 0.05 in “glm” in R for multifactorial Cox proportional hazards regression analyses. Furthermore, two nomograms were developed to estimate the risk and OS of rectal cancer distant metastases, based on risk factors and independent prognostic indicators. For comparing the continuous data, we used independent t test or Mann-Whitney U-test to ensure the accuracy and reliability of the data, while for comparing the categorical data, we chose the chi-square test or the Fisher’s exact test in order to obtain more accurate and scientific analysis results. In addition, time-dependent ROC curves of nomogram and all independent prognostic variables at 12, 24, and 36 months were generated, and the corresponding time-dependent AUCs were applied to show the discrimination. Calibration curves and DCA of 12, 24, and 36 months were plotted to evaluate the nomogram. According to the median risk score, all rectal cancer patients with distant metastasis were divided into high- and low-risk groups. Kaplan–Meier (K–M) survival curves with the log-rank test were performed to show the difference OS status between the two groups. The graphs’ correctness was assessed using the C-index and ROC, and their discriminatory power was confirmed by calibration curves. Decision Curve Analysis (DCA) is a method that facilitates decision making regarding test selection and use,²³ which allows comparison of nomogram with other models. Compared to strictly quantitative performance metrics like the area under the receiver operating characteristic curve, decision curves are superior.²⁴ The R software (version 4.3.2; https://www.r-project.org/) was used for all statistical analyses.

Results

Baseline Clinical Characteristics of the Patients

Based on our rigorous screening, 7872 rectal cancer from the SEER database were included in the study. According to the status of distant metastasis, we categorized all patients into two groups: rectal cancer without distant metastasis (n = 7126) and rectal cancer with distant metastasis (n = 746). Table 1 details the baseline clinical characteristics of all patients. The chi-square test showed no significant differences between the training and validation groups for all variables except the primary site (P < 0.05). In the overall population, patients <65 and ≥65 years of age were approximately equally divided. Total 60.1% of patients were male, and 77.4% were white. Adenocarcinoma was the most prevalent pathological kind (74.6%). The most prevalent degree of distinction was Grade II (80.6 %). Stages T3 (58.2%) and N0 (62.4%) were most commonly observed. A tumor size <50 mm was found in 67.3% of patients.

Table 1.

Baseline Clinical Characteristics of Patients Diagnosed With Rectal Cancer.

	Training (N = 5510)	Validation (N = 2362)	Overall (N = 7872)	X²	P
Sex				0.0067896	0.9343
Female	2195 (39.8%)	944 (40.0%)	3139 (39.9%)
Male	3315 (60.2%)	1418 (60.0%)	4733 (60.1%)
Age				0.94743	0.3304
<65	3143 (57.0%)	1376 (58.3%)	4519 (57.4%)
>=65	2367 (43.0%)	986 (41.7%)	3353 (42.6%)
Race				1.9291	0.3812
Black	359 (6.5%)	151 (6.4%)	510 (6.5%)
Other	909 (16.5%)	361 (15.3%)	1270 (16.1%)
White	4242 (77.0%)	1850 (78.3%)	6092 (77.4%)
Primary_site				5.4145	0.01997
Rectosigmoid junction	1422 (25.8%)	670 (28.4%)	2092 (26.6%)
Rectum	4088 (74.2%)	1692 (71.6%)	5780 (73.4%)
Histologic_type				0.011538	0.9145
Adenocarcinoma	4107 (74.5%)	1764 (74.7%)	5871 (74.6%)
Other	1403 (25.5%)	598 (25.3%)	2001 (25.4%)
Grade				1.1252	0.771
I	412 (7.5%)	177 (7.5%)	589 (7.5%)
II	4432 (80.4%)	1916 (81.1%)	6348 (80.6%)
III	584 (10.6%)	232 (9.8%)	816 (10.4%)
IV	82 (1.5%)	37 (1.6%)	119 (1.5%)
T				0.18656	0.9109
T1	1189 (21.6%)	508 (21.5%)	1697 (21.6%)
T2	1108 (20.1%)	485 (20.5%)	1593 (20.2%)
T3	3213 (58.3%)	1369 (58.0%)	4582 (58.2%)
N				1.1753	0.2783
N0	3417 (62.0%)	1496 (63.3%)	4913 (62.4%)
N1	2093 (38.0%)	866 (36.7%)	2959 (37.6%)
M				0.086833	0.7682
M0	4974 (90.3%)	2138 (90.5%)	7112 (90.3%)
M1	536 (9.7%)	224 (9.5%)	760 (9.7%)
Tumor-size				1.0501	0.5915
<50	3708 (67.3%)	1589 (67.3%)	5297 (67.3%)
>100	111 (2.0%)	56 (2.4%)	167 (2.1%)
50-100	1691 (30.7%)	717 (30.4%)	2408 (30.6%)
Surgery				0.016243	0.8986
NO	553 (10.0%)	240 (10.2%)	793 (10.1%)
Yes	4957 (90.0%)	2122 (89.8%)	7079 (89.9%)
Radiation				0.065611	0.7978
None/unknown	2837 (51.5%)	1208 (51.1%)	4045 (51.4%)
Yes	2673 (48.5%)	1154 (48.9%)	3827 (48.6%)
Chemotherapy				0.042843	0.836
No/unknown	2245 (40.7%)	969 (41.0%)	3214 (40.8%)
Yes	3265 (59.3%)	1393 (59.0%)	4658 (59.2%)

Independent Risk Factors of Rectal Cancer With Distant Metastases

The independent risk factors for distant metastasis of rectal cancer are shown in Table 2. We performed univariate logistic analysis of 12 potential factors, resulting in the identification of 11 relevant variables. Multivariate logistic regression analysis (P < 0.05) showed that the independent predictors of distant metastasis in rectal cancer were race, tumor stage, T, N, radiotherapy, chemotherapy, surgery, tumor size, and histological subtype. These significant variables were then incorporated into the final risk factor prediction model. Univariate analysis showed that race (OR: 0.6389, CI: 0.4939-0.8373, P = 0.000863), histological type (OR: 0.3952, CI: 0.3153-0.4901, P < 0.001), later tumour stage (III/IV: both P-values <0.05), T2 stage (OR: 0.4126, 95% CI: 0.2962-0.5670, P < 0.001),T3 stage (OR: 1.6808,95% CI: 1.3840-2.0549, P < 0.001),N1 stage (OR: 3.3484, 95% CI: 2.8672-3.9177, P < 0.001),was associated with a high risk was significantly associated. Men have a higher risk of distant metastases compared to women. Patients who had undergone surgery and radiotherapy had a reduced risk of distant metastases, however, patients who had completed chemotherapy had an increased risk of distant metastases. The results of multifactorial logistic regression analysis showed that patients with adenocarcinoma, no surgery and no radiotherapy, completed chemotherapy, tumour size of 50-100 mm, T3 stage, N1 stage, tumour stage IV and rectal cancer of black ethnicity were still associated with a higher risk of distant metastasis.

Table 2.

Univariate and Multivariate Logistic Analyses of Rectal Cancer With Distant Metastasis.

	Univariate Analysis			Multivariate Analysis
	OR	95% CI	P	OR	95% CI	P
Age
<64	Reference
>=65	0.8514	0.7300-0.9916	0.0394	0.8865	0.7397-1.0609	0.1904
Sex
Female	Reference
Male	1.1965	1.0251-1.3992	0.0237	1.1586	0.9687-1.3883	0.1086
Race
Black	Reference
Other	0.5711	0.4167-0.7868	0.00054	0.7474	0.5172-1.0861	0.1235
White	0.6389	0.4939-0.8373	0.000863	0.6908	0.5077-0.9511	0.0208
Primary_Site
Rectosigmoid junction	Reference
Rectum	0.9089	0.7707-1.0753	0.26
Histologic.Type
adenocarcinoma	Reference
other	0.3952	0.3153-0.4901	<0.001	0.4898	0.3770-0.6304	<0.001
Grade
I	Reference
II	1.4127	1.0233-2.0067	0.0438	1.2093	0.8296-1.809	0.3383
III	2.6237	1.8193-3.8636	<0.001	1.7675	1.1517-2.7631	0.0106
IV	2.0340	1.0527-3.7526	0.0275	2.4468	1.1644-4.9275	0.0145
T
T1	Reference
T2	0.4126	0.2962-0.5670	<0.001	0.4389	0.3003-0.6351	<0.001
T3	1.6808	1.3840-2.0549	<0.001	1.0676	0.8133-1.4074	0.6401
N
N0	Reference
N1	3.3484	2.8672-3.9177	<0.001	2.1094	1.7471-2.5521	<0.001
Tumor-size
<50	Reference
50-100	2.6728	2.2902-3.1205	<0.001	1.6289	1.3565-1.9561	<0.001
>100	3.7941	2.5389-5.5319	<0.001	1.7166	1.0406-2.7686	0.0302
Surgery
No	Reference
Yes	0.0987	0.08315-0.1172	<0.001	0.0963	0.07839-0.1182	<0.001
Radiation
No	Reference
Yes	0.7980	0.6861-0.9276	0.00335	0.1797	0.1474-0.2188	<0.001
Chemotherapy
No	Reference
Yes	4.4266	3.6206-5.4604	<0.001	6.0867	4.6934-7.9394	<0.001

Diagnostic Nomogram Modeling and Validation

Using independent predictors from the multifactorial logistic regression, we created a nomogram model for predicting rectal cancer risk in combination with distant metastases (Figure 2A). The nomogram AUC values for both training and validation sets exceeded 0.8, indicating good discriminative properties (Figures 2B, E). Strong agreement was observed in the calibration curves between the predictions and observations (Figures 2C, F). In addition, the DCA suggested that our nomogram could be used as a tool for diagnosing distant metastases in patients with rectal cancer (Figures 2D, G). Meanwhile, we also generated ROC curves of nomogram vs TN staging (Supplemental Figure 1A–(B)): the results show that our prediction model has a good diagnostic effect. To further validate the applicability of the model, we retrospectively collected data on 226 patients with rectal cancer combined with distant metastases in the Weifang Hospital of Traditional Chinese Medicine from 2012-2024 to form an expanded test set for our study. The nomogram 1-, 2-, and 3 year AUCs of the expanded test set were 0.612, 0.590, and 0.610, respectively (Supplemental Figure 5), and the calibration and DCA curves for each independent factor showed good predictability.

Figure 2.

Construction and validation of a diagnostic nomogram (A) Development and validation of a diagnostic risk model for assessing the distant metastasis of rectal cancer. Rectal cancer distant metastasis risk model. (B) Area under the ROC curve for the training set. (C) The training set nomogram correction plot. (D) Decision curve analysis of training set. (E) Area under the ROC curve for the validation set. (F) Validation of nomogram correction plot. (G) Decision curve analysis of the validation set.

Analysis of Prognostic Factors in Patients with Rectal Cancer and Distant Metastases

The clinical features of the 746 patients with rectal cancer in our study who developed distant metastases, including 437 patients (58.6%) who underwent surgery, 329 patients (44.1%) who received radiotherapy, and 644 patients (86.3%) who received chemotherapy, are shown in Supplemental Table 1. Suitable prognostic factors were screened using Cox proportional hazards regression analyses, in which the age, sex, tumor size, surgery, radiotherapy, and chemotherapy were independent prognostic factors for patients with rectal cancer and distant metastases (Table 3). The function “surv_cutpoint” in the R package “survminer” was applied to determine the optimal cut-off value of the risk threshold according the total score. Specifically, when the hazard ratio (HR) of a variable is less than 1, we consider it as a protective factor or low-risk variable, i.e., the higher the expression level of the variable, the lower the risk the patient is exposed to; conversely, it is considered as a risk factor or high-risk variable. To construct the risk score model, we used the R package ‘glmnet’ for regression analysis. We calculated the regression coefficients for each variable based on the prognostic variables previously screened through the regression model. The value at risk is equal to the sum of the product of the expression level of each variable and its corresponding regression coefficient. We used the median risk value as a dividing criterion to classify patients into high- and low-risk groups. That is, patients with risk values above the median were classified as high risk group, while patients with risk values below the median were classified as low risk group. In conclusion, the criteria for low and high risk were based on the magnitude of HR values, the results of the risk scoring model and the median risk value. As indicated by the K-M survival curves, the OS status of the high-risk group was comparatively worse than that of the low-risk group (P < 0.0001).

Table 3.

Unifactorial and Multifactorial Cox Regression Analyses of Rectal Cancer With Distant Metastasis.

	Univariate Analysis			Multivariate Analysis
	OR	95%CI	P	OR	95% CI	P
Age
<64
≥65	1.4991	1.274-1.764	<0.001	1.3623	1.1542-1.608	0.000257
Sex
Female
Male	1.2124	1.022-1.438	0.0268	1.1902	1.0002-1.416	0.04973
Race
Black	1.1763	0.8985-1.540	0.237
Other	1.0025	0.7933-1.267	0.983
White
Primary_Site
Rectosigmoid	0.8577	0.7165-1.027	0.0945	0.8870	0.7292-1.079	0.230643
Junction
Rectum
Histologic_type
adenocarcinoma
other	1.0301	0.8086-1.312	0.811
Grade
Ⅰ
II	1.0185	0.6952-1.492	0.925
III	1.3375	0.8820-2.028	0.171
IV	0.8115	0.3940-1.671	0.571
T
T1
T2	0.849	0.5998-1.2018	0.355942	1.2867	0.8931-1.854	0.175983
T3	0.6903	0.5596-0.8516	0.000541	1.1475	0.9135-1.442	0.236962
N
N0
N1	0.9527	0.8052-1.127	0.573
Tumor_size
<50
50-100	1.2213	1.035-1.441	0.0177	1.1856	0.9996-1.406	0.050487
>100	1.7599	1.213-2.554	0.00293	1.5798	1.0796-2.312	0.01855
Surgery
No	2.5350	2.151-2.988	<0.001	2.4432	2.0367-2.931	<0.001
Yes
Radiation
No	1.4313	1.215-1.686	<0.001	1.2722	1.0628-1.523	0.008709
Yes
Chemotherapy
No	2.5193	2.024-3.136	<0.001	2.1986	1.7514-2.760	<0.001
Yes

Development and Validation of Prognostic Nomogram

We established a nomogram for predicting survival in patients with rectal cancer and distant metastases (Figure 3) based on six factors: age, sex, tumor size, surgery, chemotherapy, and radiotherapy. A significant association was noted between the nomogram calibration curves and 1-, 2-, and 3 year survival probability. A consistency was noted between the actual outcomes in the training (Supplemental Figure 2(A)–(C)) and validation sets (Supplemental Figure 2(D)–(F)), and the nomogram-predicted OS. The DCA curves showed that the nomogram could be applied in clinical practice (Supplemental Figure 3(A)–(F)). Furthermore, the results of the ROC analysis indicated that the nomogram AUC was also discriminative for predicting the OS in patients with rectal cancer combined with DM, reaching 0.743, 0.750, and 0.736 at 12, 24, and 36 months, respectively, in the training set (Supplemental Figure 4(A)) and 0.828, 0.774, and 0.744 in the validation set (Supplemental Figure 4(B)). The C index (Supplemental Table 2) was 0.687 (95% CI: 0.6615-0.7125) in the training cohort and the C-index was 0.692 (95% CI: 0.6508-0.7332) in the internal validation cohort. The external validation cohort was categorized as 0.704 (0.6785-0.7295). The K-M curves showed that patients in the high-risk group had significantly worse OS than those in the low-risk group (Supplemental Figure 4(C) and (D)).

Figure 3.

Nomogram for predicting 1, 2 and 3 year overall survival in patients with metastatic rectal cancer.

Discussion

Studies on the development of risk factors and prognostic models for rectal cancer with distant metastases are lacking. Therefore, using a large number of clinical samples from the SEER database, we developed risk factors and prognostic prediction models. Although previous studies have achieved some results in the construction of prognostic models, there are still some limitations, such as the lack of external validation, insufficient sample size, and incomplete selection of risk factors. In response to these limitations, we have improved and added accordingly. We combined traditional clinicopathological factors with treatment options for comprehensive analysis. Among these, patients with adenocarcinoma, no surgery and no radiotherapy, completed chemotherapy, tumour size of 50-100 mm, T3, stage N1, tumour stage IV, and rectal cancer of black ethnicity were at greater risk of distant metastasis. Patients of older age, male, larger tumours, no surgery, no chemotherapy and no radiotherapy had poorer OS. However, the association between metastasis and age is still unclear. There are two potential explanations: one pertaining to the immune system and the other to the mechanical characteristics of the tissue, have been proposed by certain scientists.²⁵ As we age, immune cells become less functional, making them unable to effectively recognize and remove tumor cells and providing an opportunity for tumor cells to spread. The tumor microenvironment (TME) and pre-metastatic ecological niche environment play key roles in driving the reprogramming of cancer cells towards a metastatic progression state. The extracellular matrix (ECM) is a highly dynamic microenvironmental component of the metastatic cascade that changes and evolves during aging.⁷ Interestingly, although race and T and N stages are independent risk factors for distant metastases in patients with rectal cancer, their impact on prognosis is not significant. Surgery, radiotherapy, and chemotherapy are included as important factors that may affect patient survival and disease progression as predictor variables. These treatments do not directly cause distant metastases per se, but they can influence the ability of a tumour to grow, invade and metastasise, thus indirectly affecting a patient’s prognosis. For example, genetic mutations and adaptive changes in tumour cells after radiotherapy may also lead to treatment failure and tumour recurrence. Population studies and evidence suggest that chemotherapy may provide a survival advantage and improve quality of life for distant metastases in patients with rectal cancer.²⁶ Based on our analysis, chemotherapy is essential to improve the prognosis of distant metastases in patients with rectal cancer. Of the samples in the SEER database, 3214 (40.8%) did not receive chemotherapy and 4658 (59.2%) did. However, chemotherapeutic agents may produce additional toxicities (e.g., neurotoxicity) from which not all patients will benefit. Zhang et al found that radiotherapy leads to an increased risk of metastasis.²⁷ Therefore, it is important to consider each patient’s specific clinical situation when deciding whether to use adjuvant therapy for rectal cancer.²⁸ In addition, Cox regression analysis showed that radiotherapy as an adjuvant treatment had the same benefit as chemotherapy. Notably, shorter courses of radiotherapy have been found to reduce treatment duration and toxicity, thereby increasing treatment completion rates in all patients.²⁹ Although some previous retrospective studies have suggested that patients with distant metastases do not benefit from palliative surgery and may even facilitate the progression of systemic metastases due to surgical resection, our study yielded the opposite result.^30,31 This difference may be explained by the fact that our study included a broader group of patients, including those who were traditionally considered unsuitable for surgery, but who may benefit from surgery thanks to new treatment strategies. Enhanced postoperative management (e.g., pain control, nutritional support, and rehabilitation exercises) similarly improved patients’ postoperative survival. This difference in AUC values may be due to the different characteristics of survival data at different time periods. For example, short-term prediction models (e.g., 1-year survival prediction models) may be more focused on reflecting rapid disease progression, and factors such as surgical recovery period, psychological changes, and complications may have a significant impact on survival time in the short term. Long-term prediction models (e.g., 3-year survival prediction models) may be more focused on assessing factors such as the overall health of the patient, the effectiveness of the treatment regimen, and the potential risk of recurrence.

Currently, most studies have assessed only one aspect of risk factors and prognosis prediction. As far as we know, we are the first to establish a comprehensive retrospective study on predicting the risk of distant metastases and prognosis in patients with rectal cancer. This visualized predictive model may help physicians determine targeted treatment strategies.

A study by Han et al found that adenocarcinoma, gender, tumor size, poor grading, T1 stage, and N1/N2 stage were significantly associated with the development of metastasis,^32,33which is similar to our study. Therefore, clinicians should be aware of these risk factors in patients with rectal cancer, while physicians should recommend timely treatment to these potentially at-risk patients. There are persistent racial and ethnic disparities in the causes of death from rectal cancer. In addition, nomogram modeling suggests that the tumor size is an independent risk factor for distant metastases in patients with rectal cancer. De et al showed that in the rodent tumor model KHT-C, tumor oxygenation decreased with increasing tumor size, and hypoxic tumors exhibited a higher risk of metastasis. The hypoxic environment triggered by changes in tumor size may play a pivotal role in the metastatic ability of human tumors.³⁴ In a study on breast cancer, Yang et al similarly found that the larger the tumour the worse the prognosis.³⁵ Further, our research clinically validated that surgery, chemotherapy, and radiotherapy could improve the survival prognosis of patients with metastasis.

Our study has the following strengths: First, we comprehensively analyzed the risk factors and prognostic predictions of distant metastases in patients with rectal cancer, and included external validation data from Asians, which expanded the diversity and generalizability of the validation and helped to assess the performance and reliability of the model more comprehensively. Finally, the independence of the validation dataset ensures the objectivity and accuracy of the assessment results, making the model more convincing.

This study had certain limitations. First, specific data on chemotherapy drugs and doses were not available in the SEER database, so we were unable to explore the impact of chemotherapy further. Meanwhile, we excluded a large amount of data with missing or ambiguous information, which exacerbated the risk of bias in the results to some extent. Second, we had to exclude biomarkers such as CEA and CA199 due to the lack of information on their specific parameters in the pre-2010 SEER database. Purely clinical and therapeutic means may not be able to satisfy the evaluation of tumor prognosis, and comprehensive consideration of combining a variety of biomarkers, laboratory indices, and clinical characteristics is expected to provide a more accurate and valuable basis for the determination of tumor prognosis.³² With the progress of medicine, the application of immune and targeted therapies has gradually become widespread and has changed the prognosis of patients to a certain extent. Looking ahead, clinical practice urgently needs more outstanding prediction models to guide and support medical decision-making, in order to drive continuous progress in the field of healthcare.

Conclusion

In summary, using univariate and multivariate logistic regression analyses, We identified independent risk factors for the development of distant metastases in patients with rectal cancer. Subsequently, through univariate and multivariate Cox regression analyses, we determined the factors related to disease prognosis. This can help doctors accurately determine the prognosis of patients and provide strong support for the development of personalised treatment plans. Our column line graphs can also provide an important reference for clinical trial screening and patient stratification.

Supplemental Material

Supplemental Material - Evaluation of Risk Factors, and Development and Validation of Prognostic Prediction Models for Distant Metastasis in Patients With Rectal Cancer: A Study Based on the SEER Database and a Chinese Population

Supplemental Material for Evaluation of Risk Factors, and Development and Validation of Prognostic Prediction Models for Distant Metastasis in Patients With Rectal Cancer: A Study Based on the SEER Database and a Chinese Population by Huiru Zhang, Haojun Wang, Yan Yao, Lijuan Liu, Fubin Feng, Huayao Li, and Changgang Sun in Cancer Control.

Footnotes

Acknowledgments

Thanks to for the drawing material.This work was funded by the National Natural Science Foundation of China (82174222) and Shandong Province Natural Science Foundation (ZR2021LZY015), whose support is greatly appreciated.

Author Contributions

HRZ, HJW conceived of and designed the study. HJW, YY and LJL performed literature search. HRZ, YY and HYL generated the figures and tables. HRZ,FBF and HYL analyzed the data. HRZ, HJW wrote the manuscript and HRZ critically reviewed the manuscript. CGS supervised the research. All authors have read and approved the manuscript.

Declaration of Conflicting Interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the National Natural Science Foundation of China (82174222) and Shandong Province Natural Science Foundation (ZR2021LZY015).

Ethical Statement

ORCID iDs

Huiru Zhang

Haojun Wang

Yan Yao

Lijuan Liu

Fubin Feng

Huayao Li

Changgang Sun

Supplemental Material

Supplemental material for this article is available online.

References

Aparicio

Caldas

. The implications of clonal genome evolution for cancer medicine. N Engl J Med. 2013;368(9):842-851. doi:10.1056/NEJMra1204892

Langley

Fidler

. The seed and soil hypothesis revisited—the role of tumor‐stroma interactions in metastasis to different organs. Int J Cancer. 2011;128(11):2527-2535. doi:10.1002/ijc.26031

Sleeman

Nazarenko

Thiele

. Do all roads lead to Rome? Routes to metastasis development. Int J Cancer. 2011;128(11):2511-2526. doi:10.1002/ijc.26027

Boatman

Nalluri

Gaertner

. Colon and rectal cancer management in low-resource settings. Clin Colon Rectal Surg. 2022;35(05):402-409. doi:10.1055/s-0042-1746189

Ciardiello

Martini

Napolitano

Tabernero

Cervantes

. Clinical management of metastatic colorectal cancer in the era of precision medicine. CA Cancer J Clin. 2022;72(4):372-401. doi:10.3322/caac.21728

Fan

Wang

, et al. Immunotherapy in colorectal cancer: current achievements and future perspective. Int J Biol Sci. 2021;17(14):3837-3849. doi:10.7150/ijbs.64077

Fane

Weeraratna

. Normal aging and its role in cancer metastasis. Cold Spring Harb Perspect Med. 2020;10(9):a037341. doi:10.1101/cshperspect.a037341

Siegel

Wagle

Cercek

Smith

Jemal

. Colorectal cancer statistics, 2023. CA Cancer J Clin. 2023;73(3):233-254. doi:10.3322/caac.21772

Gómez-Cuadrado

Tracey

Qian

Brunton

. Mouse models of metastasis: progress and prospects. Dis Model Mech. 2017;10(9):1061-1074. doi:10.1242/dmm.030403

10.

Fan

, et al. Targeting FAPα-positive lymph node metastatic tumor cells suppresses colorectal cancer metastasis. Acta Pharm Sin B. 2024;14(2):682-697. doi:10.1016/j.apsb.2023.11.002

11.

Amri

Bordeianou

Sylla

Berger

. Variations in metastasis site by primary location in colon cancer. J Gastrointest Surg. 2015;19(8):1522-1527. doi:10.1007/s11605-015-2837-9

12.

Pugh

Shinkins

Fuller

Mellor

Mant

Primrose

. Site and stage of colorectal cancer influence the likelihood and distribution of disease recurrence and postrecurrence survival: data from the FACS randomized controlled trial. Ann Surg. 2016;263(6):1143-1147. doi:10.1097/SLA.0000000000001351

13.

Conzelmann

Linnemann

Berger

. Detection of disseminated tumour cells in the liver of colorectal cancer patients. Eur J Surg Oncol. 2005;31(1):38-44. doi:10.1016/j.ejso.2004.09.005

14.

Leen

. The detection of occult liver metastases of colorectal carcinoma. J Hepatobiliary Pancreat Surg. 1999;6(1):7-15. doi:10.1007/s005340050078

15.

Keller

Berho

Perez

Wexner

Chand

. The multidisciplinary management of rectal cancer. Nat Rev Gastroenterol Hepatol. 2020;17(7):414-429. doi:10.1038/s41575-020-0275-y

16.

Balachandran

Gonen

Smith

DeMatteo

. Nomograms in oncology: more than meets the eye. Lancet Oncol. 2015;16(4):e173-e180. doi:10.1016/S1470-2045(14)71116-7

17.

Liang

Zhang

Jiang

, et al. Development and validation of a nomogram for predicting survival in patients with resected non–small-cell lung cancer. J Clin Orthod. 2015;33(8):861-869. doi:10.1200/JCO.2014.56.6661

18.

Weiser

Landmann

Kattan

, et al. Individualized prediction of colon cancer recurrence using a nomogram. J Clin Orthod. 2008;26(3):380-385. doi:10.1200/JCO.2007.14.1291

19.

Iasonos

Schrag

Raj

Panageas

. How to build and interpret a nomogram for cancer prognosis. J Clin Orthod. 2008;26(8):1364-1370. doi:10.1200/JCO.2007.12.9791

20.

Boehm

Larcher

Beyer

, et al. Identifying the most informative prediction tool for cancer-specific mortality after radical prostatectomy: comparative analysis of three commonly used preoperative prediction models. Eur Urol. 2016;69(6):1038-1043. doi:10.1016/j.eururo.2015.07.051

21.

Boffa

Rosen

Mallin

, et al. Using the national cancer database for outcomes research: a review. JAMA Oncol. 2017;3(12):1722-1728. doi:10.1001/jamaoncol.2016.6905

22.

Collins

Reitsma

Altman

Moons

KGM

. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015;350(jan07 4):g7594. doi:10.1136/bmj.g7594

23.

Fitzgerald

Saville

Lewis

. Decision curve analysis. JAMA. 2015;313(4):409-410. doi:10.1001/jama.2015.37

24.

Kerr

Brown

Zhu

Janes

. Assessing the clinical impact of risk prediction models with decision curves: guidance for correct interpretation and appropriate use. J Clin Orthod. 2016;34(21):2534-2540. doi:10.1200/JCO.2015.65.5654

25.

Purushotham

Shamil

Cariati

, et al. Age at diagnosis and distant metastasis in breast cancer – a surprising inverse relationship. Eur J Cancer. 2014;50(10):1697-1705. doi:10.1016/j.ejca.2014.04.002

26.

Benson

Venook

Al-Hawary

, et al. NCCN guidelines insights: colon cancer, version 2.2018. J Natl Compr Cancer Netw. 2018;16(4):359-369. doi:10.6004/jnccn.2018.0021

27.

Zhang

Wang

, et al. Nomogram predicts risk and prognostic factors for bone metastasis of pancreatic cancer: a population-based analysis. Front Endocrinol. 2022;12:752176. doi:10.3389/fendo.2021.752176

28.

André

Boni

Navarro

, et al. Improved overall survival with oxaliplatin, fluorouracil, and leucovorin as adjuvant treatment in stage II or III colon cancer in the MOSAIC trial. J Clin Orthod. 2009;27(19):3109-3116. doi:10.1200/JCO.2008.20.6771

29.

Erlandsson

Holm

Pettersson

, et al. Optimal fractionation of preoperative radiotherapy and timing to surgery for rectal cancer (Stockholm III): a multicentre, randomised, non-blinded, phase 3, non-inferiority trial. Lancet Oncol. 2017;18(3):336-346. doi:10.1016/S1470-2045(17)30086-4

30.

Badwe

Hawaldar

Nair

, et al. Locoregional treatment versus no treatment of the primary tumour in metastatic breast cancer: an open-label randomised controlled trial. Lancet Oncol. 2015;16(13):1380-1388. doi:10.1016/S1470-2045(15)00135-7

31.

Demicheli

Retsky

Swartzendruber

Bonadonna

. Proposal for a new model of breast cancer metastatic development. Ann Oncol. 1997;8(11):1075-1080. doi:10.1023/A:1008263116022

32.

Guan

Quan

, et al. A clinical model to predict the risk of synchronous bone metastasis in newly diagnosed colorectal cancer: a population-based study. BMC Cancer. 2019;19(1):704. doi:10.1186/s12885-019-5912-x

33.

Han

Dai

, et al. Nomogram to predict the risk and survival of synchronous bone metastasis in colorectal cancer: a population-based real-world analysis. Int J Colorectal Dis. 2020;35(8):1575-1585. doi:10.1007/s00384-020-03612-z

34.

De Jaeger

Merlo

Kavanagh

Fyles

Hedley

Hill

. Heterogeneity of tumor oxygenation: relationship to tumor necrosis, tumor size, and metastasis. Int J Radiat Oncol Biol Phys. 1998;42(4):717-721. doi:10.1016/S0360-3016(98)00323-X

35.

Yang

Zhang

Tian

, et al. Prognosis prediction and risk factors for triple‐negative breast cancer patients with brain metastasis: a population‐based study. Cancer Med. 2023;12(7):7951-7961. doi:10.1002/cam4.5575

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.99 MB