Construction of conditional survival nomograms for metastatic early onset colon cancer patients

Introduction

Colorectal cancer (CRC) is the third most commonly diagnosed cancer worldwide and the second most common cause of cancer-related death (1). Following therapeutic advances, the overall incidence and mortality of CRC patients have declined (2). However, the increasing incidence of CRC in young patients is of concern.

Early onset colon cancer (EO-CC), which is defined as CC in patients younger than 50 years old, exhibits specific features in terms of its biological behaviors, genomic landscape, and prognosis (3-6). Metastatic early onset colon cancer (mEO-CC) is defined as the presence of distant metastasis along with the primary tumor in such patients. Despite improved treatment adherence and higher treatment intensity administration, the progression rate is notably higher in younger patients than elderly patients (7,8). Thus, more robust prognostic factors need to be identified and predictive models need to be developed that can make precise survival predictions and aid in the enhancement of clinical decision making.

As an increasing number of prognostic factors, including the lymph node ratio (LNR), primary tumor location, differentiation, carcinoembryonic antigen (CEA) level, and perineural invasion (PI) status, have been identified that enhance the precision of survival prediction in mEO-CC, prognostic nomograms have been developed. These nomograms include independent prognostic factors and aim to streamline the process of survival prediction (9-12). In addition to the aforementioned clinicopathological prognosticators, the postoperative survival duration is a frequently underappreciated pivotal factor that exerts a dynamic effect on prognosis, with the risk of mortality evolving over time post-surgery.

A traditional survival analysis conventionally commences from the date of diagnosis or surgical intervention. Conversely, a conditional survival (CS) analysis examines the likelihood of surviving additional years after having already achieved a specific period of survival. CS evaluations provide invaluable quantitative insights into the evolving probability of survival across temporal dimensions and merit recognition as more refined instruments for prognostic analysis (13-15). In routine settings where molecular or regimen-level data are unavailable, CS provides pragmatic, time-updated probabilities for patient counselling and surveillance decisions.

To date, no studies appear to have formulated nomograms for the prediction of CS in mEO-CC. Within its purview, this study sought to delineate CS patterns in mEO-CC patient cohorts and to establish nomograms to prognosticate the probability of CS. These nomograms were developed using data sourced from the SEER database. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-582/rc).

Methods

Patient selection

This study used data from the SEER program, a publicly accessible database curated by the Unites States (U.S.) National Cancer Institute. The SEER program compiles detailed demographic, clinical, treatment, and outcome data from 18 regional cancer registries, representing roughly one-third of the U.S. population. We retrospectively identified patients diagnosed with mEO-CC between 2010 and 2019. Patients were included in the study if they met the following inclusion criteria: (I) were aged below 50 years at the time of diagnosis; and (II) had pathological confirmation of metastatic CRC. Patients lacking essential clinical data were excluded from the analysis. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Since all information in the SEER database is de-identified and publicly available, informed consent was not required.

Statistical analysis

The continuous variables are presented as the median and interquartile range (IQR), while the categorical variables are described as the frequency and percentage. To identify the independent prognostic indicators, multivariate Cox proportional hazards models were employed. Variables with univariable P<0.10 or strong prior evidence were eligible for multivariable entry. Proportional hazards were assessed using Schoenfeld residuals. In addition, a CS analysis was conducted to assess dynamic survival probabilities based on the time already survived. CS was defined as the probability of surviving an additional “b” years given survival up to “a” years from the date of diagnosis, and was calculated using the following formula: CS(b|a) = S(a + b)/S(a) (16). Prognostic nomograms were constructed using variables retained in the multivariate models to estimate CS probabilities. SEER captures systemic therapy only as a high-level indicator (chemotherapy yes/no/unknown) without regimen, dose, or cycle details; because of missingness and limited granularity, this indicator was not included in the primary model. To avoid reduced events-per-parameter and potential calibration drift, no train-test split was performed; internal performance was reported on the full cohort. The predictive performance of the nomograms was evaluated through time-dependent receiver operating characteristic (ROC) curve analysis, with the AUC serving as a measure of discriminative ability. All the statistical analyses were performed using R software (version 3.4.1; R Foundation for Statistical Computing, Vienna, Austria). All P values were two-sided with α=0.05.

Results

Demographic and clinicopathological characteristics

Ultimately, the data of 1,988 patients with mEO-CCwere included in this study. The basic clinicopathological features of the patients are set out in Table 1. The median age of all patients was 42.22 (IQR: 39–47) years. Of the patients, 52.4% (n=1,041) were male, and most of the patients were white (n=1,430, 71.9%). A large proportion of the patients had tumors located in the left colon (n=1,396, 70.2%), while others had tumors located in the right colon (n=461, 23.2%) and transverse colon (n=131, 6.6%). In terms of histological grade, 29.1% (n=579) of the tumors were categorized as grades III and IV, indicating poor differentiation and undifferentiation. The histological type of most tumors (n=1,811, 91.1%) was adenocarcinoma. Signet ring cell carcinoma was observed in 2.1% (n=42). In terms of tumor (T) stage, the majority of patients had either T3 (n=1,075; 54.1%) or T4 tumors (n=743; 37.4%). Additionally, a high proportion of mEO-CC patients presented with lymph node metastases, with 36.7% (n=729) classified as N1 and 46.2% (n=918) as N2. Additionally, 75.5% (n=1,501) of the patients had elevated CEA levels. Further, the majority of patients had liver metastasis (n=1,416, 71.2%).

Conditional overall survival and cancer-specific survival estimates

The actuarial OS rates and conditional OS probabilities are shown in Table 2 and Figure 1. The actuarial OS declined over time, such that the OS rate was 83% in the first year and 26% in the fifth year. Conditional OS improved progressively as patients survived longer, reflecting a dynamic increase in survival probability over time. Moreover, a compelling trend emerged wherein patients who had survived a greater number of years post-diagnosis had notably improved prospects of attaining a high survival rate. For example, the probability of achieving 5-year survival increased from 26% directly to 31%, 42%, 60% and 79% in patients who had already survived for 1, 2, 3, and 4 years, respectively. Notably, the CSS rates closely paralleled the OS rates (the 5-year CSS for patients who had already survived for 1, 2, 3, and 4 years were 32%, 43%, 61% and 80% respectively), and the temporal trajectory of conditional CSS exhibited a consistent pattern. This observation suggests that non-cancer-related mortality contributed to only a relatively minor proportion of events in the context of the mEO-CC patients.

Figure 1 Kaplan-Meier estimates of survival after surgery (0 year) and conditional survival according to the years already survived after surgery (1–6 years). (A) Overall survival; (B) cancer-specific survival.

Predictors of overall survival and cancer-specific survival

In the univariate analysis, age and gender were not found to be predictors of either OS or CSS; however, race, tumor location, differentiation, histology, T stage, N stage, the LNR, the CEA level, and organ metastasis were found to be risk factors of both OS and CSS. After balancing all these variables in the multivariable analysis, a tumor in the right colon, poor differentiation, mucinous adenocarcinoma or signet-ring cell carcinoma, late T stage, a high LNR, an elevated CEA level, PI, and specific organ (bone, liver, or lung) metastasis were found to be significantly independent risk factors of OS and CSS (Table 3).

Prediction nomogram for conditional overall survival and cancer-specific survival

A prognostic nomogram was developed incorporating tumor site, histological grade, histological type, T stage, LNR, preoperative CEA level, PI status, and the presence of bone, liver, and lung metastases to estimate the probability of 5-year OS and 5-year conditional OS, conditional on having already survived 1, 2, or 3 years (Figure 2A). Similarly, Figure 2B presents a nomogram predicting 5-year CSS and 5-year conditional CSS at the same survival landmarks. For example, a patient with a left-colon tumor, grade II, mucinous adenocarcinoma, T3 stage, a LNR of 0.5, an elevated CEA level, and liver metastasis had a total nomogram score of 224, and a 5-year OS probability of 14%, which increased to 42% if the patient had already survived for 3 years. The two nomograms were examined separately, and the survival ROC curves were used to assess the discrimination ability of both nomograms; The AUC 0.765 for 5-year OS (Figure 3A) and 0.761 for 5-year CSS (Figure 3B). The developed nomograms performed better than any other clinicopathological features alone in predicting 5-year conditional OS and 5-year conditional CSS.

Figure 2 Conditional nomograms for the prediction of the conditional 5-year probability of overall survival (A) and cancer-specific survival (B). CEA, carcinoembryonic antigen; CSS, cancer-specific survival; LNR, lymph node ratio; OS, overall survival; PI, perineural invasion; T, tumor.

Figure 3 Survival ROC curves of the nomograms and other risk factors for OS (A) and CSS (B) prediction. AUC, area under the curve; CEA, carcinoembryonic antigen; CSS, cancer-specific survival; OS, overall survival; PI, perineural invasion; ROC, receiver operating characteristic; T, tumor.

Discussion

CRC represents a significant malady that poses a formidable threat to human wellbeing and carries a substantial societal burden due to its elevated incidence and mortality rates. More than half of the general population afflicted with CRC present with distant metastasis (17). Notably, the incidence of EO-CC in individuals under the age of 50 has shown a consistent annual increase of 2% (18). Compared to late-onset CC, EO-CC is characterized by superior performance status and a higher likelihood of treatment plan completion. However, EO-CC has a significantly worse cancer-specific outcome, indicative of a more aggressive disease biology (7). Further research needs to be conducted on EO-CC due to its specific features in biological behaviors, genomic landscape, and prognosis (3,4). Patients with EO-CC who receive more aggressive treatment are generally associated with improved life expectancy (19,20).

In the realm of conventional clinical research, the analysis of survival revolves around examining the temporal distribution of patients’ survival from the date of diagnosis or the initiation of treatment. Survival estimates grounded in fixed temporal landmarks may conceivably exert an influence on mortality rates, but could also have misleading implications for medical practitioners and patients alike. Thus, it is important to examine how prognosis changes over time in relation to the length of patient survival. As the survival duration extends, there ensues a corresponding potential for the enhancement of long-term patient prognosis. CS estimates provide a dynamic evaluation of protracted oncological outcomes for patients, contingent on the temporal intervals they have already endured (21,22). CS has been used to more precisely evaluate the survival of different kinds of tumors, such as hepatocellular carcinoma, esophageal cancer, and pancreatic cancer (16,23-25).

In this study, our primary focus was on mEO-CC patients. Our study revealed a noteworthy trend wherein the likelihood of conditional OS steadily increased with each successive year of survival. Notably, we found that the more years a patient had already traversed on their survival journey, the greater their prospects of attaining an increased survival rate. In a striking example, the probability of achieving the 5-year survival milestone increased significantly, leaping directly from a modest 26% to an impressive 79%, provided that the patient had already survived for four years.

Of particular significance is the observation that the conditional CSS rates closely mirrored the trends in the OS rates, and the temporal trajectory of conditional CSS exhibited a consistent pattern. This phenomenon can be attributed to the limited incidence of non-cancer-related mortality in the advanced stages of the disease and the relatively lower prevalence of comorbidities among mEO-CC patients.

The univariate and multivariate Cox regression analyses revealed that a tumor in the right colon, poor differentiation, non-adenocarcinoma, late T stage, a high LNR, an elevated CEA level, PI and specific organ metastasis were significant independent risk factors of OS and CSS in the mEO-CC patients. Research has shown that the outcomes of patients with right-sided tumors are worse than those of patients with left-sided tumors (26,27). Poor differentiation implies a high malignant degree and invasiveness. Consistent with our results, previous studies have shown that an advanced T/N stage and an elevated CEA level are associated with poor survival (28,29).

Using graphical representation to illustrate the influence of various predictors on the final outcome, a nomogram provides more tangible and interpretable insights into the effects of these predictors on the outcome. This tool serves as a practical model for healthcare professionals, enabling them to estimate survival rates based on crucial clinicopathologic predictors. Nomograms can incorporate independent predictors identified by Cox analyses, including variables such as the tumor site, histological grade, histological type, T stage, LNR, CEA level, PI, bone metastasis, liver metastasis, and lung metastasis. Nomograms also have increased predictive capacity to estimate the probability of achieving a 5-year OS and 5-year conditional OS if a patient has already survived 1, 2, and 3 years.

This study has several limitations. First, due to its retrospective nature and the inherent constraints of the SEER database, some important variables were unavailable, including detailed treatment information and molecular profiles, both of which may significantly impact survival outcomes. Additionally, SEER does not reliably capture liver or lung metastasectomy for the 2010–2019 period, which may leave residual confounding. Second, the SEER database primarily includes patients from the United States and lacks detailed data on race-specific biological differences or socioeconomic status, which may limit the generalizability of the findings to more diverse populations. Finally, external validation using independent multicenter cohorts with more comprehensive clinical and molecular data is warranted. Future external, multi-center validation is warranted.

Conclusions

This study developed prognostic nomograms for patients with mEO-CRC. CS improved over time, offering a more realistic outlook for long-term survivors. The nomograms showed moderate discriminative ability for predicting 5-year survival and CS. Time-adjusted models incorporating survival duration enhanced prognostic accuracy and may support individualized clinical decisions. Further validation in more diverse populations with molecular and treatment data is needed.

Acknowledgments

We would like to thank the SEER database for providing the valuable data.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-582/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-582/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-582/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Siegel RL, Miller KD, Goding Sauer A, et al. Colorectal cancer statistics, 2020. CA Cancer J Clin 2020;70:145-64. [Crossref] [PubMed]
3. Jácome AA, Vreeland TJ, Johnson B, et al. The prognostic impact of RAS on overall survival following liver resection in early versus late-onset colorectal cancer patients. Br J Cancer 2021;124:797-804. [Crossref] [PubMed]
4. Lieu CH, Golemis EA, Serebriiskii IG, et al. Comprehensive Genomic Landscapes in Early and Later Onset Colorectal Cancer. Clin Cancer Res 2019;25:5852-8. [Crossref] [PubMed]
5. Samuel SM, Varghese E, Büsselberg D. Complexity of insulin resistance in early-onset colorectal cancer. Cancer Cell 2025;43:797-802. [Crossref] [PubMed]
6. Huang T, Qiu J, Wang C, et al. Trends and Projections of Early-Onset Colorectal Cancer Burden in China, 1990-2036: Findings From the Global Burden of Disease 2021 Study. Cancer Control 2025;32:10732748251341524. [Crossref] [PubMed]
7. Fontana E, Meyers J, Sobrero A, et al. Early-Onset Colorectal Adenocarcinoma in the IDEA Database: Treatment Adherence, Toxicities, and Outcomes With 3 and 6 Months of Adjuvant Fluoropyrimidine and Oxaliplatin. J Clin Oncol 2021;39:4009-19. [Crossref] [PubMed]
8. Carbone F, Spinelli A, Ciardiello D, et al. Prognosis of early-onset versus late-onset sporadic colorectal cancer: Systematic review and meta-analysis. Eur J Cancer 2025;215:115172. [Crossref] [PubMed]
9. Tang M, Wang H, Cao Y, et al. Nomogram for predicting occurrence and prognosis of liver metastasis in colorectal cancer: a population-based study. Int J Colorectal Dis 2021;36:271-82. [Crossref] [PubMed]
10. Huang YQ, Liang CH, He L, et al. Development and Validation of a Radiomics Nomogram for Preoperative Prediction of Lymph Node Metastasis in Colorectal Cancer. J Clin Oncol 2016;34:2157-64. [Crossref] [PubMed]
11. Hong T, Cai D, Jin L, et al. Development and validation of a nomogram to predict survival after curative resection of nonmetastatic colorectal cancer. Cancer Med 2020;9:4126-36. [Crossref] [PubMed]
12. Luo P, Li YY, Huang C, et al. A novel conditional survival nomogram for monitoring real-time prognosis of non-metastatic colorectal cancer. Discov Oncol 2024;15:179. [Crossref] [PubMed]
13. Zheng Z, Wang X, Liu Z, et al. Individualized conditional survival nomograms for patients with locally advanced rectal cancer treated with neoadjuvant chemoradiotherapy and radical surgery. Eur J Surg Oncol 2021;47:3175-81. [Crossref] [PubMed]
14. Wancata LM, Banerjee M, Muenz DG, et al. Conditional survival in advanced colorectal cancer and surgery. J Surg Res 2016;201:196-201. [Crossref] [PubMed]
15. Ali ND, Donohue K, Zandieh S, et al. Conditional Survival Analysis of Metastatic Colorectal Cancer Patients Living ≥24 Months: A Single Institutional Study. Am J Clin Oncol 2019;42:512-8. [Crossref] [PubMed]
16. Hagens ERC, Feenstra ML, Eshuis WJ, et al. Conditional survival after neoadjuvant chemoradiotherapy and surgery for oesophageal cancer. Br J Surg 2020;107:1053-61. [Crossref] [PubMed]
17. Biller LH, Schrag D. Diagnosis and Treatment of Metastatic Colorectal Cancer: A Review. JAMA 2021;325:669-85. [Crossref] [PubMed]
18. Mauri G, Sartore-Bianchi A, Russo AG, et al. Early-onset colorectal cancer in young individuals. Mol Oncol 2019;13:109-31. [Crossref] [PubMed]
19. Benson AB, Venook AP, Al-Hawary MM, et al. NCCN Guidelines Insights: Colon Cancer, Version 2.2018. J Natl Compr Canc Netw 2018;16:359-69. [Crossref] [PubMed]
20. Van Cutsem E, Cervantes A, Adam R, et al. ESMO consensus guidelines for the management of patients with metastatic colorectal cancer. Ann Oncol 2016;27:1386-422. [Crossref] [PubMed]
21. Haydu LE, Scolyer RA, Thompson JF. Conditional survival estimates for cancer patients. Oncotarget 2017;8:84639-40. [Crossref] [PubMed]
22. Jung SH, Lee HY, Chow SC. Statistical Methods for Conditional Survival Analysis. J Biopharm Stat 2018;28:927-38. [Crossref] [PubMed]
23. Latenstein AEJ, van Roessel S, van der Geest LGM, et al. Conditional Survival After Resection for Pancreatic Cancer: A Population-Based Study and Prediction Model. Ann Surg Oncol 2020;27:2516-24. [Crossref] [PubMed]
24. Shah MM, Meyer BI, Rhee K, et al. Conditional survival analysis of hepatocellular carcinoma. J Surg Oncol 2020;122:684-90. [Crossref] [PubMed]
25. Shin DW, Bae J, Ha J, et al. Conditional relative survival of cervical cancer: a Korean National Cancer Registry Study. J Gynecol Oncol 2021;32:e5. [Crossref] [PubMed]
26. Kalantzis I, Nonni A, Pavlakis K, et al. Clinicopathological differences and correlations between right and left colon cancer. World J Clin Cases 2020;8:1424-43. [Crossref] [PubMed]
27. Kim K, Kim YW, Shim H, et al. Differences in clinical features and oncologic outcomes between metastatic right and left colon cancer. J BUON 2018;23:11-8.
28. Pulte D, Jansen L, Brenner H. Disparities in Colon Cancer Survival by Insurance Type: A Population-Based Analysis. Dis Colon Rectum 2018;61:538-46. [Crossref] [PubMed]
29. Stelzner S, Radulova-Mauersberger O, Zschuppe E, et al. Prognosis in patients with synchronous colorectal cancer metastases after complete resection of the primary tumor and the metastases. J Surg Oncol 2019;120:438-45. [Crossref] [PubMed]