Introduction

Genetic Cancer Risk Assessment (GCRA) has become an integral component of colorectal cancer (CRC) care. GCRA refers to a structured clinical process that includes: systematic collection of personal and family cancer history, risk stratification based on established clinical criteria and prediction models (e.g., Amsterdam and Bethesda criteria, PREMM5, and tumor-based testing such as immunohistochemistry), and determination of eligibility for germline testing. When indicated, this process is followed by germline testing to identify pathogenic or likely pathogenic variants (PVs) in cancer susceptibility genes. Through this comprehensive approach, GCRA informs risk-adapted surveillance, preventive interventions, therapeutic decision-making, and cascade testing for at-risk relatives [1].

CRC ranks as the third most prevalent cancer worldwide [2]. Although most early-onset cases are sporadic [3,4], and the disease is multifactorial [5], the rising incidence of CRC in younger populations underscores the need to refine strategies for identifying individuals at increased hereditary risk and to expand access to genetic services [6].

The clinical relevance of hereditary CRC syndromes, particularly Lynch syndrome (LS), has driven the development of risk stratification tools to guide germline testing. Early frameworks, including the Amsterdam I criteria, prioritized family history of CRC but failed to capture extracolonic malignancies [7]. Subsequent refinements, such as the Amsterdam II and revised Bethesda criteria, incorporated a broader tumor spectrum and molecular features, including high microsatellite instability (MSI-H), to improve case detection [8–10].

LS arises from germline PVs in DNA mismatch repair (MMR) genes (MLH1, PMS2, MSH2, and MSH6) or EPCAM, resulting in loss of MMR protein function and a hypermutator phenotype. This biological basis underpins the use of immunohistochemistry (IHC) as a tumor-based screening approach [11,12].

The PREdiction Model for gene Mutations 5 (PREMM5) integrates personal and family history to estimate the probability of carrying a germline PV in LS-associated genes. A threshold of ≥2.5% has demonstrated high sensitivity; however, the limited representation of Hispanic/Latino populations in validation cohorts highlights a critical gap in the generalizability of these tools [13].

Given that genetic ancestry and environmental exposures may influence CRC risk, the performance of established prediction models may not be uniform across populations. In this context, we aimed to evaluate the performance of clinical criteria, PREMM5, and IHC for identifying carriers of germline PVs in a selected cohort of patients with CRC at two centers in Mexico.

Methods

Results

Cohort characteristics

A total of 1,081 patients were assessed for eligibility, of whom 242 met the selection criteria. Among these, 211 provided written informed consent, and 208 completed all study procedures (Fig 1). Among the 208 patients included in this analysis, 127 (61.0%) were included at INCMNSZ and 81 (39.0%) at Hospital Zambrano Hellion. The distribution of clinicopathological characteristics is presented in Table 1. The median age at first CRC diagnosis for the entire cohort was 45.0 years (18.0–82.0), with 52.4% of the participants being women. The proportion of probands by stage of CRC were 13.0% for stage I, 25.5% for stage II, 21.6% for stage III, and 18.8% for stage IV. 21.1% had an unknown stage or were undergoing staging at the time of enrollment.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Clinicopathological differences between non-carriers and carriers of germline PVs in Lynch syndrome-associated genes. The characteristics of carriers of PVs in other cancer susceptibility genes (TP53, ATM, CHEK2, BRCA1, BRCA2, PALB2, and BRIP1) are also described. For this analysis, monoallelic PVs in MUTYH were considered a non-actionable result.

https://doi.org/10.1371/journal.pone.0354795.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart of patient selection.

https://doi.org/10.1371/journal.pone.0354795.g001

Variant detection and clinicopathological characteristics by carrier status

Germline PVs were identified in 32.2% (67/208) of the probands. Among these, 77.6% (n = 52/67) were carriers of PVs in LS-related genes (30 in MLH1, 14 in MSH2, 6 in MSH6, and 2 in PMS2). Of note, no PVs were found in EPCAM. 22.4% (n = 15/67) of carriers harbored PVs in other cancer susceptibility genes (4 in ATM, 3 in BRCA1, 3 in CHEK2, 2 in TP53, 1 in BRCA2, 1 in BRIP1, and 1 in PALB2). Additionally, one individual harbored a monoallelic PV in MUTYH, and was classified within the subgroup of patients with negative results. Fig 2.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Distribution of germline pathogenic and likely pathogenic variants among Mexican patients with CRC referred for GCRA.

https://doi.org/10.1371/journal.pone.0354795.g002

Comparing non-carriers, patients with LS presented with their first CRC at an earlier age (47.0 vs. 41.0 years, p = 0.011) and had a higher proportion of primary tumors in the right colon (39.0% vs. 63.5%, p = 0.005).

IHC results for MMR proteins in tumor tissue were available for 107 individuals. Compared to individuals without germline PVs, those with LS exhibited a significantly higher frequency of MMR-deficient tumors (31.3% vs. 91.2%, p < 0.001). Individuals with LS also showed an increased occurrence of second primary malignancies (20.6% vs. 42.3%, p = 0.003), primarily due to a higher incidence of metachronous CRC (0.7% vs. 13.5%, p = 0.005) and gastric cancer (0.7% vs. 7.7%, p = 0.019). In patients with LS and known MMR status, IHC was concordant with germline test results in 91.2% of cases (31/34). The three discordant cases included one carrier each of PVs in MLH1, MSH2 and MSH6, all of whom exhibited MMR-proficient tumors.

In the subgroup of patients with PVs in other cancer susceptibility genes, 86.7% (n = 13/15) harbored PVs in genes involved in the homologous recombination repair pathway. Among these individuals, the median age at diagnosis of the first CRC was 43.0 years; 60.0% were female, and 66.7% had tumors located in the left colon or rectum. Additionally, 20.0% presented with metastatic disease at diagnosis, and all patients with available IHC results exhibited MMR-proficient tumors.

Clinical genetic referral criteria and prediction models performance

Family history data revealed that 38.0%, 11.5%, and 9.6% of the entire cohort reported familial occurrences of colorectal, gastric, and endometrial cancers, respectively. Compared to non-carriers, individuals with LS were significantly more likely to report a family history of colorectal (24.8% vs. 75.0%, p < 0.001), endometrial (5.0% vs. 23.0%, p = 0.005), gastric (4.3% vs. 32.7%, p < 0.001), and pancreatic cancer (3.5% vs. 13.5%, p = 0.018). Among individuals carrying PVs in other genes, 33.3% reported a family history of CRC. Table 2.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Differences in self-reported family cancer history limited to first-degree and second-degree relatives on the affected side between non-carriers and carriers of PVs in LS genes.

https://doi.org/10.1371/journal.pone.0354795.t002

Compared to non-carriers, a greater proportion of individuals with LS met the Amsterdam II criteria (7.8% vs. 53.8%). Similarly, a higher percentage of LS carriers fulfilled the revised Bethesda criteria (80.1% vs. 98.0%). Additionally, 88.5% of patients with LS had a PREMM5 score ≥2.5%, as detailed in S2 Table.

The Amsterdam II criteria demonstrated good accuracy (Sensitivity 56.0%, 95% CI 41.3–70.0; Specificity 91.9%, 95% CI 85.9–95.9), while the Revised Bethesda criteria (Sensitivity 98.1%, 95% CI 89.7–99.9), PREMM5 score ≥2.5% (Sensitivity 92.0%, 95% CI 80.7–97.8), and IHC for MMR proteins (Sensitivity 91.2%, 95% CI 76.3–98.1) exhibited high sensitivity. Table 3.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Diagnostic test indicators for Lynch syndrome.

https://doi.org/10.1371/journal.pone.0354795.t003

The area under the ROC curve for PREMM5 was 0.822 (95% CI 0.746–0.898) with a median score of 28.5% (95% CI 23.4–33.5) in patients with LS and of 5.9% (95% CI 4.1–7.6) in non-carriers. Fig 3.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Receiver operating characteristic (ROC) curve and area under the curve (AUC) for the PREMM5 predictive model.

https://doi.org/10.1371/journal.pone.0354795.g003

Discussion

The yield of germline PVs in our cohort of individuals with CRC who met clinical criteria for GCRA was 32.2%. In cohorts of patients with CRC selected for age younger than 50 years (with a 10% incidence of MMR-deficient tumors) a prevalence of PVs ranging from 16% to 25% has been reported [3,4]. In contrast, studies using universal multigene panel testing strategies (regardless of age at onset or family history) have reported germline PV frequencies ranging from 5% to 15% in CRC cohorts [19–21]. The high frequency of germline PVs in our cohort can likely be attributed to the highly selected nature of the population, characterized by a median age of 45.0 years at first CRC diagnosis, 48.5% presenting with MMR-deficient tumors among those with available IHC results, and 38.0% reporting a family history of CRC. Our cohort represents a referral-enriched population, which likely contributed to the high prevalence of germline PVs and may overestimate the diagnostic performance of the evaluated tools. Therefore, these findings should not be extrapolated to unselected CRC populations.

In our study, 77.6% of selected hereditary CRC cases were attributed to Lynch syndrome, predominantly involving PVs in MLH1 (57.7% of LS cases). This aligns with a retrospective analysis of Mexicans referred to GCRA, where LS was the most confirmed hereditary syndrome in the subgroup of patients with CRC, mostly linked to MLH1 PVs [22]. Our findings also reinforce prior evidence from Mexican LS families, in which MLH1 was the most frequently altered gene across a broad tumor spectrum. In these families, CRC was the most prevalent cancer, highlighting its role as a frequent trigger for GCRA in this population [23].

Resources for germline testing in Mexico are limited; therefore, determining the most effective screening strategies is essential. A pivotal prospective study of 1,066 unselected individuals with newly diagnosed CRC reported that 12.7% had high microsatellite instability, and IHC demonstrated a sensitivity of 93.2% (95% CI: 88.9–97.4) for detecting tumors with this phenotype [24]. In patients with MLH1/PMS2 deficiency identified through tumor tissue IHC, other etiologies beyond LS may be responsible, such as somatic variants [25], MLH1 promoter methylation [26], and MLH1 constitutional methylation [27]. Among our patients with LS and available IHC results, 91.2% had MMR-deficient CRC, consistent with prior reports indicating that approximately 90.0% of individuals with LS had MMR-deficient tumors [28,29]. In contrast, 8.8% of individuals with LS-associated CRCs in our cohort were MMR-proficient, consistent with reports from other studies showing rates as high as 10.0%. These findings suggest that such patients may remain at risk of CRC through alternative pathways [29].

Universal screening through IHC for MMR proteins in tumor tissue not only identifies individuals at risk for hereditary cancer but also has significant clinical implications, such as assessing recurrence risk to guide adjuvant chemotherapy decisions in early-stage CRC [30], and selecting eligible patients for immunotherapy in neoadjuvant [31,32], adjuvant [33], and metastatic settings [34,35]. Looking ahead, the use of immunotherapy in MMR-deficient CRC is expected to play a key role in the development of organ-preserving strategies, with the goal of optimizing patient outcomes while minimizing treatment-related morbidity [36]. A high proportion of our enrolled patients had deficient-MMR tumors, suggesting a referral bias, with patients displaying this biomarker being more frequently referred to GCRA.

Among the predictors evaluated, the Amsterdam II criteria demonstrated the lowest sensitivity, likely due to their stringent requirements, which necessitate simultaneous fulfillment of multiple criteria to identify an individual as an at-risk proband. Consequently, this may result in the non-identification of individuals with limited family history [37], young carriers who have not yet manifested symptoms, carriers with lower penetrance PVs [38], and in rare instances, de novo germline PVs [39].

Previous reports indicate that approximately 72.0% of patients with LS identified through a CRC diagnosis met the revised Bethesda criteria [28]. In our cohort, 86.0% of all referred patients and 98.0% of those diagnosed with LS met these criteria, highlighting their sensitivity and suggesting they can be effectively used as referral criteria for germline genetic testing by healthcare professionals managing individuals diagnosed with CRC.

In our study, PREMM5 effectively distinguished carriers of PVs in LS-genes, achieving an area under the ROC curve (AUC) of 0.822 (95% CI 0.746–0.898) in a Mexican population. This performance is comparable to the AUC of 0.83 (95% CI: 0.75–0.92) reported by Kastrinos et al. in a cohort composed predominantly of non-Hispanic White individuals [13].

Screening strategies combining IHC for MMR proteins with germline genetic testing have been recognized as cost-effective, particularly as the number of family members undergoing genetic testing increases [40]. Enhancing access to cascade testing for targeted variant identification poses a critical challenge in Mexico, especially as the diagnosis of hereditary CRC in probands continues to rise. It may be that population genomic screening of high-evidence genes linked to hereditary conditions, including LS, could prove cost-effective with adequate access to risk-reduction measures [41]. This highlights the critical need for establishing specialized clinics to manage patients at high risk for CRC and other neoplasms.

In our cohort, 22.4% of the identified PVs were located in non-LS genes, predominantly involving genes related to homologous recombination pathway. Analyzing this subgroup was challenging due to the heterogeneous CRC risk associated with these genes. Among those identified with PVs in non-Lynch syndrome genes, only TP53 has an established association with CRC risk. In the LIFT UP study, it was reported that TP53 PV carriers had a 7.2% cumulative risk of CRC by age 80, with an elevated relative risk peaking at age 20 and declining after age 70 [42]. These findings support early screening colonoscopy in young adults with Li-Fraumeni syndrome. In contrast, the NCCN recently revised its guidelines [1], no longer recommending high-risk colon cancer screening for CHEK2 PV carriers. The presence of germline PVs in non-LS genes underscores the increasing complexity of GCRA in the era of multigene panel testing. This expanded spectrum challenges traditional risk prediction tools and highlights the need to validate broader models—such as PREMMplus [43]—in our population. Such tools may improve the identification of individuals carrying PVs in a wider array of cancer susceptibility genes beyond those classically linked to CRC.

Conclusion

In our cohort, composed entirely of patients self-identified as having Mexican-Mestizo ancestry, we validated the accuracy of the PREMM5 model, supporting its use as a tool to guide patient selection for germline genetic testing. While no predictive model is infallible, our findings suggest that in Mexican patients with CRC, decisions regarding genetic testing should be informed—but not solely determined—by risk models.

We recommend a comprehensive clinical evaluation incorporating detailed family cancer history, physical examination, and relevant endoscopic and histological findings. In this context, use of the PREMM5 score and the revised Bethesda criteria—both demonstrating high sensitivity—may help reduce missed high-risk individuals who could benefit from germline testing. Although genetic testing for all patients with CRC should ideally become standard practice in Mexico, optimizing referral criteria remains essential to improve risk stratification in settings with limited access.

Finally, genetic testing should not be restricted to patients with MMR-deficient tumors. Although Lynch syndrome is the most common hereditary cancer syndrome in Mexican patients with CRC, our findings demonstrate a broader spectrum of germline pathogenic variants.

Supporting information

Acknowledgments

We thank the American Society of Clinical Oncology Virtual Mentoring Program 2023 for fostering collaboration with international researchers. We would like to express our deepest gratitude to the patients who participated in the study, along with their families and caregivers, for their invaluable support and contributions.