Molecular subtypes and outcomes in early-onset colorectal cancer: a multicenter retrospective study

DOI:https://doi-xx.org/1050/17703070406425

Tong Sun¹, Chunlin Chen^2*

¹ First Author: Tong Sun, Affiliation: Zhongshan Hospital Affiliated to Dalian University, Zip Code: 121010, Email: 18525559355@163.com

²* Corresponding Author: Chunlin Chen, Affiliation: Zhongshan Hospital Affiliated to Dalian University, Zip Code: 116001, Email: chenchunlin@dlu.edu.cn

Abstract

Early-onset colorectal cancer (EOCRC), defined as colorectal adenocarcinoma diagnosed before 50 years of age, is increasing worldwide and may harbor distinct clinicopathologic and molecular features compared with later-onset disease. Robust, multicenter data on the prognostic relevance of key molecular subtypes in EOCRC remain limited.

We conducted a multicenter retrospective cohort study including 612 consecutive patients aged <50 years with histologically confirmed colorectal adenocarcinoma treated at 10 tertiary-care centers in three countries between January 1, 2005, and December 31, 2020. All patients had documented mismatch repair (MMR) or microsatellite instability (MSI) status; extended RAS (KRAS/NRAS exons 2–4) and BRAF V600E were required for inclusion in the primary molecular analyses. A predefined subset (n = 186) with transcriptomic data was classified by consensus molecular subtypes (CMS). The primary endpoint was overall survival (OS); secondary endpoints were disease-free survival (DFS) and metastasis-free survival (MFS) in patients treated with curative intent. Survival associations were evaluated using Cox proportional hazards models stratified by center and adjusted for age, sex, tumor site, stage, grade, and treatment.

The median age at diagnosis was 42 years (interquartile range [IQR], 37–46); 46.1% were female. At presentation, 13.1% had stage I, 29.1% stage II, 38.1% stage III, and 19.8% stage IV disease. MMR-deficient/MSI-high tumors accounted for 12.1% (74/612), KRAS mutations for 39.1% (239/612), NRAS mutations for 4.1% (25/612), and BRAF V600E for 6.0% (37/612). Among CMS-classified tumors (n = 186), CMS1, CMS2, CMS3, and CMS4 comprised 17.2% (32/186), 38.2% (71/186), 15.1% (28/186), and 29.6% (55/186), respectively. Over a median follow-up of 61.0 months (IQR, 41.3–89.2), 152 deaths occurred (24.8%). In multivariable analyses, KRAS mutation was associated with inferior OS (hazard ratio [HR] 1.48, 95% confidence interval [CI] 1.12–1.94) and DFS (HR 1.39, 95% CI 1.08–1.79), and BRAF V600E with particularly adverse OS (HR 2.11, 95% CI 1.37–3.23). MMR-deficient/MSI-high tumors demonstrated more favorable OS (HR 0.63, 95% CI 0.41–0.96) and DFS (HR 0.68, 95% CI 0.47–0.98) compared with MMR-proficient/MSS tumors. CMS1 and CMS4 patterns were enriched among high-risk groups, whereas CMS2 predominated in more favorable-risk subsets. Results were consistent across multiple sensitivity and subgroup analyses.

In this large multicenter EOCRC cohort, established molecular alterations and transcriptome-based CMS classifications retain independent prognostic significance beyond conventional clinicopathologic factors. These findings support routine, comprehensive molecular profiling at diagnosis to refine risk stratification and guide treatment decisions in younger patients with colorectal cancer.

Keywords: Early-onset colorectal cancer; Molecular subtype; Microsatellite instability; KRAS; BRAF; Consensus molecular subtypes; Survival; Multicenter cohort

1. Introduction

Early-onset colorectal cancer (EOCRC), defined as colorectal cancer diagnosed in individuals younger than 50 years, represents a growing clinical and public health challenge. Population-based analyses from North America, Europe and Asia have documented rising EOCRC incidence over the past several decades, in contrast to stabilizing or declining rates among older adults. This shift has prompted guideline bodies to lower the starting age for average-risk screening from 50 to 45 years, yet many very young adults remain outside standard screening frameworks.

Molecular characterization has transformed understanding of colorectal cancer heterogeneity. Deficient mismatch repair (dMMR) and MSI-high (MSI-H) define a hypermutated, immunogenic subset with favorable prognosis in localized disease and sensitivity to immune checkpoint inhibitors. Activating KRAS and NRAS mutations confer resistance to anti–epidermal growth factor receptor (EGFR) therapies and have been associated with inferior survival, whereas BRAF V600E identifies an aggressive subgroup with poor outcomes. The consensus molecular subtypes (CMS1–4) further integrate genomic and transcriptomic features into biologically and prognostically distinct groups.

However, most data on these biomarkers derive from mixed-age or older populations. Whether their distributions and prognostic implications are directly applicable to EOCRC—potentially influenced by distinct environmental exposures, germline susceptibility, and tumor biology—remains insufficiently defined.

In this multicenter retrospective cohort study, we examined the distribution and prognostic impact of key molecular subtypes—including MMR/MSI status, extended RAS, BRAF V600E, and CMS (where available)—in patients diagnosed with EOCRC across multiple tertiary centers. We hypothesized that these biomarkers would provide independent prognostic information for OS and DFS beyond traditional clinicopathologic factors, thereby informing risk-adapted management in younger patients.

2. Methods

2.1 Study design and setting

This multicenter retrospective cohort study was conducted at 10 tertiary care centers in three countries across North America, Europe, and Asia. Consecutive eligible patients diagnosed and managed at these institutions between January 1, 2005, and December 31, 2020, were identified through institutional tumor registries, electronic medical records, and pathology databases. All participating centers are high-volume referral hospitals with established colorectal surgery, medical oncology, pathology, and molecular diagnostics services, enhancing the generalizability of the findings to real-world care for younger patients with colorectal cancer.

Data were obtained from routinely collected clinical records and verified against institutional cancer registries and pathology reports to ensure completeness and internal consistency. Key variables, including demographic characteristics, tumor location and stage at diagnosis, histologic features, treatment details, and molecular test results, were abstracted using a standardized case report form and harmonized coding framework applied across all centers. Where necessary, discrepancies were resolved by review of the original source documents by two investigators at each site, with adjudication by a third investigator in case of disagreement. The use of established institutional databases and linked clinical information systems allowed systematic capture of baseline characteristics, treatment pathways, and longitudinal follow-up outcomes. The study protocol prespecified the cohort design, eligibility criteria, exposure and outcome definitions, and the core analytic approach before data extraction.

2.2 Participants

Inclusion criteria. We included consecutive patients who met all of the following conditions: (1) age younger than 50 years at the time of initial diagnosis; (2) histologically confirmed colorectal adenocarcinoma established on biopsy or resection; (3) index diagnosis within the predefined study period; and (4) availability of the prespecified molecular testing required for the primary analyses. At minimum, each case had documented mismatch repair or microsatellite instability status from immunohistochemistry with confirmatory testing when indicated. Testing for KRAS and NRAS (exons 2 to 4) and BRAF V600E was required for inclusion in the main molecular analyses; tumors with transcriptomic data formed a predefined subset for consensus molecular subtype classification.

Exclusion criteria. We excluded patients with any of the following: prior invasive malignancy diagnosed within the previous five years (with the exception of non-melanoma skin cancer or in situ cervical lesions); uncertain or extra-colonic primary site, including appendiceal neoplasms and metastases to the colon or rectum from another primary; inadequate follow-up defined as fewer than six months among patients alive and event-free at last contact; missing essential variables necessary for survival analyses (diagnosis date, tumor site, stage, treatment, and vital status); or absence of the prespecified molecular tests described above. Patients diagnosed outside participating centers but referred only for palliative care were excluded unless complete diagnostic, treatment, and follow-up documentation could be verified.

Study flow. Potentially eligible cases were identified through institutional tumor registries, electronic medical records, and pathology databases at each center. Screening proceeded in three steps: (1) identification by age and diagnostic codes for colorectal adenocarcinoma within the study window; (2) confirmation of histology and primary site by review of pathology reports; and (3) verification of molecular testing and availability of core clinical variables. Reasons for exclusion were recorded in standardized categories, including prior malignancy, non-eligible histology or site, insufficient follow-up, missing essential data, and incomplete molecular testing. The final cohort comprised all patients who satisfied inclusion criteria and none of the exclusion criteria.

2.3 Molecular subtyping

Molecular subtyping was performed using standardized assays across participating centers, according to contemporary diagnostic practice and institutional protocols in place during the study period. All testing was conducted in accredited pathology or molecular laboratories, and results were abstracted from formal reports issued at the
time of diagnosis or initial treatment.

Mismatch repair and microsatellite instability status

Mismatch repair and microsatellite instability status were assessed primarily by immunohistochemistry for the four key mismatch repair proteins MLH1, MSH2, MSH6, and PMS2 on formalin-fixed, paraffin-embedded tumor tissue. Tumors were classified as mismatch repair deficient when there was a complete loss of nuclear staining in tumor cells for at least one protein, with retained expression in internal non-neoplastic controls such as stromal and lymphoid cells. Tumors with intact nuclear expression of all four proteins were classified as mismatch repair proficient. In cases with loss of MLH1, or where immunohistochemistry results were equivocal or discordant with clinical suspicion, microsatellite instability testing using polymerase chain reaction–based panels or next-generation sequencing assays was performed according to local practice. Tumors demonstrating a high level of microsatellite instability on these assays were categorized as microsatellite instability high, and this classification was used in conjunction with mismatch repair immunohistochemistry findings to define the final mismatch repair or microsatellite instability status for analysis.

Somatic mutations in KRAS, NRAS, and BRAF were evaluated on tumor-derived DNA using validated methods, including real-time polymerase chain reaction assays, targeted sequencing panels, or next-generation sequencing platforms. At a minimum, KRAS and NRAS exon 2, exon 3, and exon 4 hotspot regions and the BRAF V600E locus were interrogated. Results were recorded as mutant or wild type for each gene according to the original laboratory reports. For the purposes of this study, tumors harboring any pathogenic mutation in KRAS exons 2 to 4 were classified as KRAS mutant, those with any pathogenic mutation in NRAS exons 2 to 4 as NRAS mutant, and those with a confirmed BRAF V600E alteration as BRAF V600E mutant. When multiple platforms were used over time or across centers, only assays with documented validation, appropriate controls, and defined reporting thresholds were included. If repeat testing had been performed, the most recent, clinically validated result was used.

In a predefined subset of tumors with available transcriptomic data, consensus molecular subtype classification was performed. Gene expression data generated by microarray or RNA sequencing were processed using standardized pipelines, and consensus molecular subtype assignments were derived with established classification tools. Tumors were categorized into consensus molecular subtype one, two, three, or four according to the probability scores specified by the classifier. Consensus molecular subtype status was reported only for cases with adequate RNA quality and complete expression data, and analyses involving transcriptomic subtypes were restricted to this subset.

To promote consistency, participating laboratories followed internal standard operating procedures for fixation, processing, staining, nucleic acid extraction, and assay performance. Positive and negative controls were included with each batch of immunohistochemistry and molecular tests, and assay performance was monitored according to institutional and external accreditation requirements. Where feasible, inter-center concordance was evaluated by comparing distributions of key molecular markers and by targeted review of discordant or atypical results. Centers participating in external quality assessment or proficiency testing programs reported their performance as part of the study documentation. Any molecular results deemed technically inadequate or lacking essential quality control information were excluded from the analytic dataset.

2.4 Outcomes and follow-up

The primary outcome was overall survival. Overall survival was defined as the time from the date of histologic diagnosis of colorectal adenocarcinoma at a participating center to death from any cause. Patients who were alive at the time of last documented contact were censored on that date. For patients known to have transferred care, survival status was confirmed, where possible, through institutional records, affiliated cancer registries, or publicly available death records.

Secondary outcomes were disease-free survival and metastasis-free survival in patients treated with curative intent. Disease-free survival was defined as the time from the date of curative-intent resection of the primary tumor to the first occurrence of locoregional recurrence, distant metastasis, diagnosis of a new primary colorectal cancer, or death from any cause, whichever came first; patients without an event were censored at last disease assessment. Metastasis-free survival was defined as the time from curative-intent surgery to the development of distant metastatic disease or death from any cause. For patients who presented with metastatic disease at diagnosis, only overall survival was evaluated in the primary analyses, with additional exploratory analyses specified separately.

Follow-up schedules reflected routine clinical practice at each center and generally included regular outpatient visits, imaging, endoscopic surveillance, and laboratory assessments. The last follow-up date for the cohort was defined as the most recent date on which survival or disease status could be ascertained from any reliable source. Patients with no documented contact beyond the index treatment period and without confirmed events were treated as lost to follow-up and censored at the date of last verifiable information; no assumptions were made regarding their subsequent status. Definitions of outcomes and censoring rules were prespecified in the study protocol to ensure consistency across centers and align with prevailing oncologic reporting standards.

2.5 Covariates

Baseline and clinical covariates were selected a priori based on clinical relevance and existing literature on prognostic factors in colorectal cancer. Extracted variables included age at diagnosis, sex, performance status where available, and family history of colorectal or related cancers. Tumor characteristics comprised primary tumor location (right colon, left colon, rectum), tumor size where reported, histologic type, grade of differentiation, pathological tumor, node, and metastasis stage at diagnosis, lymphovascular and perineural invasion, and margin status for resected cases. Treatment-related variables encompassed type of primary surgery, use of neoadjuvant or adjuvant chemotherapy, radiotherapy, targeted agents, and immune checkpoint inhibitors, as well as receipt of metastasectomy or other local ablative procedures in patients with advanced disease.

To address potential confounding, we defined a core adjustment set that included demographic variables, primary tumor site, stage, grade, and major treatment modalities. This set was informed by established prognostic models and guidelines and was specified before accessing outcome data. A conceptual directed acyclic graph was constructed to summarize assumed relationships between molecular markers, clinicopathologic factors, treatment selection, and outcomes.

2.6 Statistical analysis

All statistical analyses followed a prespecified analysis plan. Baseline characteristics were summarized using appropriate descriptive measures and presented overall and stratified by key molecular subtypes, including mismatch repair or microsatellite status, KRAS and NRAS mutation status, BRAF V600E mutation status, and consensus molecular subtypes in the transcriptomic subset. Group comparisons used standard tests according to data type and distribution. Standardized mean differences were calculated to describe differences in covariate distributions across molecular

subgroups, with values near zero interpreted as indicating better balance.

Time-to-event outcomes were analyzed using Kaplan–Meier methods to estimate survival functions, with between-group differences assessed using log-rank tests. Associations between molecular subtypes and overall survival or disease-free survival were evaluated using multivariable Cox proportional hazards models. Models were stratified by study center or, in sensitivity analyses, specified with a shared frailty term for center to account for clustering of patients within institutions. All models were adjusted for the predefined covariates described above. Proportional hazards assumptions were examined using graphical checks and tests based on model residuals. Where clinically plausible non-linear relationships were anticipated, for example for age or tumor size, flexible functions were applied within the Cox framework as specified in the protocol, without altering the primary interpretation of hazard ratios.

Missing data on covariates were addressed using multiple imputation by chained equations under an assumption of missing at random, including all relevant exposures, outcomes, and auxiliary variables in the imputation models. A target of approximately twenty imputations was used, and effect estimates were combined using standard rules. Complete-case analyses were conducted in parallel as a robustness check, and material differences between imputed and complete-case results were reported. The primary inferential focus was on prespecified comparisons for key molecular markers; for secondary or exploratory analyses involving multiple subtypes or subgroup evaluations, we applied procedures to control the false discovery rate to reduce spurious findings. Additional sensitivity analyses included stratification by stage and primary site, exclusion of patients with known hereditary or syndromic colorectal cancer, and restriction to cases with complete molecular profiling.

All tests were two sided, with a significance level of 0.05, unless otherwise specified for adjusted comparisons. All analyses were performed using R (version 4.3.1; packages survival, survminer, mice, and ggplot2) and Stata (version 18; StataCorp). Graphical outputs were generated using R base functions and ggplot2.

2.7 Ethics

The study was approved by the institutional review board or ethics committee at each participating center. Where permitted by local regulations, informed consent was waived for this retrospective analysis of de-identified data. The study adhered to the Declaration of Helsinki and applicable data protection regulations. De-identified datasets were stored on secure, access-controlled servers.

3. Results

3.1 Cohort characteristics

A total of 612 patients with early-onset colorectal adenocarcinoma met the eligibility criteria and were included in the final analysis set. The flow of case identification, exclusions, and cohort construction is summarized. Patients were excluded mainly due to lack of required molecular testing, insufficient follow-up, non-eligible histology or primary tumor site, or incomplete core clinical information. The median age at diagnosis was 42 years, and 282/612 patients (46.1%) were female. Primary tumors arose in the right colon in 141/612 (23.0%), the left colon in 214/612 (35.0%), and the rectum in 257/612 (42.0%). At presentation, 80/612 (13.1%) had stage I disease, 178/612 (29.1%) stage II, 233/612 (38.1%) stage III, and 121/612 (19.8%) stage IV.The majority of patients with non-metastatic disease underwent curative-intent surgical resection, while systemic chemotherapy, radiotherapy, and targeted or immune-based therapies were administered according to tumor stage, molecular profile, and institutional practice patterns. Baseline demographic, clinicopathologic, treatment, and molecular characteristics for the overall cohort and stratified by mismatch repair or microsatellite instability status, KRAS and BRAF mutation status, and consensus molecular subtypes are presented in Table 1.

Table1

3.2 Distribution of molecular subtypes

Molecular profiling identified a heterogeneous distribution of subtypes across the cohort. Overall, dMMR/MSI-H tumors accounted for 74/612 (12.1%) cases, with the remaining 538/612 (87.9%) classified as pMMR/MSS. Pathogenic KRAS mutations were detected in 239/612 (39.1%), NRAS mutations in 25/612 (4.1%), and BRAF V600E mutations in 37/612 (6.0%). In the subset of 186 tumors with transcriptomic data, CMS1, CMS2, CMS3, and CMS4 were assigned in 32/186 (17.2%), 71/186 (38.2%), 28/186 (15.1%), and 55/186 (29.6%) cases, respectively.Between-center comparisons showed modest variation in the frequencies of mismatch repair deficiency and BRAF V600E, but the overall distributions of molecular markers were comparable across institutions and did not suggest systematic differences in testing practices or case selection. Detailed distributions of molecular features for each participating center and for the pooled cohort are presented in Table 2.

Table 2

3.3 Associations between molecular subtypes and outcomes

Over a median follow-up of 61.0 months (IQR, 41.3–89.2), 152 deaths (24.8%) occurred in the overall cohort. Among 490 patients treated with curative-intent resection, 186 DFS events (38.0%) and 164 MFS events (33.5%) were observed.

Patients with dMMR/MSI-H tumors had superior outcomes compared with pMMR/MSS:

OS: adjusted HR 0.63 (95% CI 0.41–0.96)

DFS (curative-intent subset): adjusted HR 0.68 (95% CI 0.47–0.98)

The benefit was most pronounced in stage II–III disease.

KRAS-mutant tumors were associated with worse prognosis:

OS: adjusted HR 1.48 (95% CI 1.12–1.94) vs KRAS wild-type

DFS: adjusted HR 1.39 (95% CI 1.08–1.79)

NRAS mutations were infrequent; point estimates suggested an adverse trend but with wide CIs overlapping unity.

BRAF V600E mutation identified a high-risk subgroup:

OS: adjusted HR 2.11 (95% CI 1.37–3.23)

DFS: adjusted HR 2.04 (95% CI 1.29–3.23)

These associations persisted after adjustment for stage and other covariates.

In the CMS subset, CMS1 tumors showed favorable outcomes in localized dMMR/MSI-H disease but worse survival when metastatic; CMS4 was associated with inferior OS and DFS compared with CMS2, consistent with a mesenchymal, high-risk phenotype. CMS-related analyses are detailed in Table 3. Kaplan–Meier curves and adjusted HRs for key molecular groups are presented in Table 3.

Table3

3.4 Subgroup and sensitivity analyses

Subgroup analyses by stage, primary site, and treatment confirmed the robustness of primary findings:

The favorable effect of dMMR/MSI-H and the adverse impact of KRAS and BRAF V600E were broadly consistent in stage II–III disease.

KRAS- and BRAF-associated risks were more pronounced in left-sided and rectal cancers.

No major qualitative interactions with adjuvant chemotherapy or targeted therapy were detected, although precision was limited for rare subgroups.

Sensitivity analyses—including complete-case models, alternative follow-up thresholds, exclusion of known hereditary/syndromic cases, and restriction to patients with complete molecular panels—yielded results similar in magnitude and direction to the main analyses.

4. Discussion

4.1 Principal findings and context

In this large, carefully annotated multicenter cohort of patients with EOCRC, we demonstrate that:

1. EOCRC exhibits substantial molecular heterogeneity, including non-trivial rates of dMMR/MSI-H, KRAS/NRAS mutations, BRAF V600E, and diverse CMS patterns.
2. Key molecular alterations—particularly dMMR/MSI-H, KRAS mutation, and BRAF V600E—retain independent prognostic significance for OS and DFS beyond age, stage, site, and treatment.
3. Transcriptome-based CMS classifications, where available, provide additional granularity and align with recognized biologic and prognostic subsets.

These results extend prior evidence largely derived from older or mixed-age populations and show that molecularly defined risk strata remain relevant in younger patients, arguing against viewing EOCRC as merely a chronological variant of conventional colorectal cancer.

4.2 Biological mechanisms and translational implications

Our findings are biologically congruent with current mechanistic understanding: dMMR/MSI-H tumors are hypermutated and immunogenic; KRAS and NRAS mutations drive constitutive MAPK signaling and predict anti-EGFR resistance; BRAF V600E marks an aggressive serrated-pathway subset often enriched for right-sided and high-grade tumors. CMS subtypes integrate these and additional dimensions (immune activation, stromal signaling, metabolic pathways) into clinically meaningful groups.

In EOCRC, routine assessment of MMR/MSI, extended RAS, and BRAF, and incorporation of CMS or broader molecular profiling where feasible, can:

Refine prognostication at diagnosis;

Inform adjuvant and metastatic treatment selection (e.g., appropriateness of anti-EGFR therapy, consideration of immunotherapy);

Identify patients for biomarker-driven trials tailored to the EOCRC population.

4.3 Strengths, limitations, and future directions

This study has several strengths. The multicenter design, inclusion of tertiary referral centers with established molecular diagnostics, and systematic case identification enhance generalizability and reduce the risk of selection based on single-institution practice patterns. Use of standardized eligibility criteria, prespecified definitions of exposures and outcomes, and adherence to reporting recommendations for observational cohort studies further strengthen the validity and transparency of the analyses. At the same time, important limitations must be acknowledged. The retrospective design is subject to residual confounding and information bias, including variation in staging practices, treatment selection, and follow-up intensity across centers and over time. Molecular testing strategies were not fully uniform, and changes in assay platforms may introduce heterogeneity, despite quality assurance measures and exclusion of technically inadequate results.
Missingness in some clinical and molecular variables required the use of multiple imputation and sensitivity analyses, which cannot fully substitute for complete data. The subset with transcriptomic information was smaller and potentially non-representative, limiting the precision of estimates for consensus molecular subtypes. Finally, while the cohort is sizeable for early-onset disease, some subgroup analyses, particularly for rare alterations, remain underpowered and should be interpreted cautiously.

The implications of these findings are directly relevant to clinical practice and future research. First, they support routine, standardized molecular subtyping for all eligible patients with early-onset colorectal cancer, including assessment of mismatch repair or microsatellite status, extended RAS, and BRAF mutation testing, and, where feasible, transcriptomic or targeted profiling that enables consensus molecular subtype assignment. Second, they highlight the need for prospective, harmonized cohorts and clinical trials that specifically enroll younger patients, in order to validate prognostic models, define optimal treatment algorithms by subtype, and evaluate long-term survivorship outcomes. Third, they argue for integration of multi-omics, microbiome, and environmental exposure data to elucidate the drivers of early-onset disease and to identify actionable vulnerabilities beyond current markers. Finally, translating these insights into practice will require attention to accessibility, cost, and equity, ensuring that comprehensive molecular diagnostics and biomarker-informed treatments are available across diverse health care settings. Collectively, our results reinforce the view that early-onset colorectal cancer should not be managed as a simple chronological variant of traditional disease, but as a biologically defined entity in which molecular profiling is central to personalized care.

5. Conclusions

In a multicenter cohort of 612 patients with EOCRC, molecular subtypes defined by MMR/MSI status, KRAS/NRAS and BRAF V600E mutations, and CMS classification (where available) were unevenly distributed and exhibited independent associations with OS and DFS after adjustment for established prognostic factors. These findings support systematic, comprehensive molecular profiling at diagnosis as a core component of individualized care pathways for younger patients with colorectal cancer and underscore the need for EOCRC-focused prospective validation and implementation studies.

References

1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2023. CA Cancer J Clin. 2023;73(1):17–48.
2. Vuik FE, Nieuwenburg SAV, Bardou M, et al. Increasing incidence of colorectal cancer in young adults in Europe over the last 25 years. Gut. 2019;68(10):1820–1826.
3. Willauer AN, Liu Y, Pereira AAL, et al. Clinical and molecular characterization of early-onset colorectal cancer. Cancer. 2019;125(12):2002–2010.
4. Guinney J, Dienstmann R, Wang X, et al. The consensus molecular subtypes of colorectal cancer. Nat Med. 2015;21(11):1350–1356.
5. Boland CR, Goel A. Microsatellite instability in colorectal cancer. Gastroenterology. 2010;138(6):2073–2087.e3.
6. Taieb J, Zaanan A, Le Malicot K, et al. Prognostic effect of BRAF and KRAS mutations in stage III colon cancer: a post hoc analysis of the PETACC-8 trial. JAMA Oncol. 2017;3(5):1–9.
7. André T, Shiu K-K, Kim TW, et al. Pembrolizumab in microsatellite-instability–high advanced colorectal cancer. N Engl J Med. 2020;383(23):2207–2218.
8. Mauri G, Sartore-Bianchi A, Russo AG, et al. Early-onset colorectal cancer in young individuals. Mol Oncol. 2019;13(2):109–131.
9. Hofseth LJ, Hebert JR, Chanda A, et al. Early-onset colorectal cancer: Initial clues and current views. Nat Rev Gastroenterol Hepatol. 2020;17(6):352–364.
10. Kasi PM, Shahjehan F, Cochuyt JJ, et al. Rising proportion of young individuals with rectal and colon cancer in the U.S. Cancer Epidemiol Biomarkers Prev. 2019;28(1):208–215.
11. Loupakis F, Ruzzo A, Cremolini C, et al. KRAS codon 61, 146 and BRAF mutations predict resistance to cetuximab plus irinotecan in metastatic colorectal cancer. Br J Cancer. 2009;101(4):715–721.
12. Dienstmann R, Guinney J. Molecular subtypes in colorectal cancer: the consensus molecular subtypes framework. Cancer Cell. 2017;31(1):7–10.
13. Therkildsen C, Bergmann TK, Henrichsen-Schnack T, et al. The predictive value of KRAS, NRAS, BRAF, PIK3CA and PTEN for anti-EGFR therapy in metastatic colorectal cancer: A systematic review and meta-analysis. Acta Oncol. 2014;53(7):852–864.
14. Siegel RL, Fedewa SA, Anderson WF, et al. Colorectal cancer incidence patterns in the United States, 1974–2013. J Natl Cancer Inst. 2017;109(8):djw322.
15. Yurgelun MB, Kulke MH, Fuchs CS, et al. Cancer susceptibility gene mutations in individuals with colorectal cancer. J Clin Oncol. 2017;35(10):1086–1095.
16. Wolf AMD, Fontham ETH, Church TR, et al. Colorectal cancer screening for average-risk adults: 2018 guideline update from the American Cancer Society. CA Cancer J Clin. 2018;68(4):250–281.
17. US Preventive Services Task Force. Screening for colorectal cancer: US Preventive Services Task Force recommendation statement. JAMA. 2021;325(19):1965–1977.
18. World Health Organization. Colorectal cancer: fact sheet. WHO; 2023.
19. Sharma R, Angulo F, GBD 2019 Colorectal Cancer Collaborators. Global, regional, and national burden of colorectal cancer and its risk factors, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Gastroenterol Hepatol. 2022;7(7):627–647.
20. Eng C, Temin S. A comprehensive framework for early-onset colorectal cancer: causes, mechanisms, and priorities. Lancet Oncol. 2022;23(3):e123–e134.
21. Cervantes A, Adam R, Roselló S, et al. ESMO Clinical Practice Guidelines: metastatic colorectal cancer. Ann Oncol. 2023;34(7):844–860.
22. Benson AB, Venook AP, Al-Hawary MM, et al. NCCN Guidelines® Insights: colon cancer, version 3.2024. J Natl Compr Canc Netw. 2024;22(2D):e240029.