Independent prognostic value of KRAS mutation in resected colorectal cancer based on multi-gene sequencing

Abstract

BACKGROUND

Colorectal cancer (CRC) shows marked molecular heterogeneity, yet the clinicopathological and prognostic utility of routinely tested driver mutations in resected CRC remains incompletely defined.

AIM

To evaluate the associations of KRAS, PIK3CA, FBXW7, and TP53 mutations with clinicopathological features and overall survival (OS) in patients with curatively resected CRC.

METHODS

We retrospectively analyzed 87 patients who underwent curative-intent CRC resection (2015-2020). Mutations in KRAS, PIK3CA, FBXW7, and TP53 were assessed using next-generation sequencing on formalin-fixed, paraffin-embedded tumor samples. Associations with clinicopathological variables were tested using appropriate categorical/continuous analyses. OS was evaluated by Kaplan-Meier/log-rank methods and Cox regression. Median follow-up was 60 months (cutoff: April 2025).

RESULTS

Mutation frequencies were KRAS of 44.8%, PIK3CA of 13.8%, FBXW7 of 27.6%, and TP53 of 64.4%. KRAS mutation clustered with adverse phenotypes (including perineural invasion, nodal/distant metastasis, multifocality, larger tumor size, and elevated tumor markers). PIK3CA mutation correlated with larger tumor diameter, while FBXW7 and TP53 showed selective clinicopathological associations. The 3-year and 5-year OS rates were 63.2% and 41.4%, respectively. KRAS mutation independently predicted worse OS (hazard ratio = 3.57), whereas PIK3CA mutation showed a protective association (hazard ratio = 0.26); FBXW7 and TP53 were not independent prognostic factors.

CONCLUSION

In resected CRC, KRAS mutation delineates a high-risk subgroup with inferior survival, supporting routine KRAS genotyping to enhance postoperative risk stratification and surveillance planning; larger, prospectively profiled cohorts are warranted to refine multi-gene prognostic models.

Key Words: Colorectal cancer; Gene mutation; KRAS; TP53; Prognosis

Core Tip: This study used multi-gene next-generation sequencing in a resected colorectal cancer cohort to compare the clinicopathological and prognostic impact of KRAS, PIK3CA, FBXW7, and TP53 mutations. We found that KRAS mutation defines an aggressive phenotype and is an independent predictor of poor long-term survival, whereas PIK3CA, FBXW7, and TP53 mutations provide limited standalone prognostic information. These findings support routine KRAS genotyping for perioperative risk stratification and suggest that other recurrent mutations should be interpreted within integrated molecular models rather than in isolation.

INTRODUCTION

Colorectal cancer (CRC), characterized by significant heterogeneity, ranks as the fourth leading cause of cancer-related deaths globally, accounting for nearly 900000 fatalities annually[1]. Recent advancements in precision medicine have propelled molecular oncology research, enabling the routine application of molecular biomarker detection in CRC for early prevention and precancerous screening. Genetic testing has become integral to personalized treatment strategies, prognosis assessment, and therapeutic efficacy evaluation[2-4]. Current CRC management primarily involves surgical resection supplemented by chemotherapy[5,6]. Genomic sequencing of tumor tissue specimens identifies six frequently mutated genes: (1) TP53; (2) BRAF; (3) KRAS; (4) PIK3CA; (5) FBXW7; and (6) APC. Among these, KRAS and PIK3CA are classified as proto-oncogenes, while TP53 and APC function as tumor suppressor genes[7-11]. Detecting these mutations is pivotal for personalized therapy. This article explores the status of KRAS, PIK3CA, FBXW7, and TP53 genes in CRC patients, analyzes their correlations with clinicopathological features and prognosis, and provides guidance for clinical precision therapeutics.

MATERIALS AND METHODS

Study subjects

Clinical and pathological data were retrospectively collected from 87 patients diagnosed with CRC who underwent curative surgical resection at the Fifth Affiliated Hospital of Zhengzhou University between January 2019 and March 2020. The cohort consisted of 63 males and 24 females.

All pathological slides were independently reviewed by two senior pathologists, and diagnoses were made in accordance with the 5^th Edition of the WHO Classification of Digestive System Tumors: Colorectal Tumors[12]. Only specimens with a tumor cell content ≥ 10% were included (all surgical resection specimens met this threshold, with no specimens excluded due to insufficient tumor cell content). Data were extracted from the hospital’s electronic medical record system.

This retrospective study was approved by the Ethics Committee of the Fifth Affiliated Hospital of Zhengzhou University (No. KY2025034-K02). The requirement for informed consent was waived by the committee as the study utilized solely anonymized patient data and was retrospective in nature, involving only pre-existing information.

Inclusion and exclusion criteria

Inclusion criteria: (1) Pathologically confirmed CRC with available genetic testing results; (2) Curative (R0) surgical resection of the primary tumor; and (3) Complete clinical and pathological records.

Exclusion criteria: (1) Receipt of preoperative radiotherapy, chemotherapy, or hormone therapy; (2) Known hereditary CRC syndromes (e.g., Lynch syndrome or familial adenomatous polyposis) or inflammatory bowel disease-associated CRC (such as ulcerative colitis-associated or Crohn’s disease-associated tumors); and (3) Age younger than 18 years.

Gene mutation detection

Mutations in KRAS, PIK3CA, FBXW7, and TP53 were analyzed using next-generation sequencing (NGS) technology.

Genomic DNA was extracted from formalin-fixed, paraffin-embedded tumor tissues and quantified for purity and concentration. Target capture probes were designed to cover the full coding regions and key exon-intron junctions of the four genes. After library preparation, high-throughput sequencing was performed on an NGS platform.

Sequencing reads were aligned to the human reference genome (GRCh38) using a standard bioinformatics pipeline. Variant calling was performed to identify single-nucleotide variants, small insertions and deletions, and copy number variations within the target regions.

Data collection

Clinical variables: The following clinical and pathological data were collected for each patient: (1) Demographic characteristics (sex, age, diabetes, hypertension, hyperlipidemia, smoking history, alcohol consumption, and family history); (2) Tumor-related features (anatomical location, size, number, perineural invasion, lymphovascular invasion, lymph node metastasis, depth of invasion, histological grade, histologic subtype, and TNM stage); and (3) Biomarkers, including preoperative tumor markers carcinoembryonic antigen (< 5.0 µg/L vs ≥ 5.0 µg/L) and carbohydrate antigen 19-9 (≤ 37 U/mLvs > 37 U/mL), surgical approach (open or laparoscopic), as well as microsatellite instability (MSI) status. Follow-up information was also recorded to determine 3-year and 5-year overall survival (OS) rates.

Follow-up procedures: Patients were followed up every six months after discharge. For those who transferred to other hospitals or missed scheduled visits, follow-up was conducted via telephone interviews or home visits. The cutoff date for follow-up was April 2025.

Tumor location classification: Tumor location was classified according to the embryological origin and blood supply. Right-sided colon cancers were defined as those proximal to the distal two-thirds of the transverse colon (including the cecum, ascending colon, and proximal transverse colon). Left-sided colon cancers encompassed tumors from the distal third of the transverse colon to the rectum (including the distal transverse colon, splenic flexure, descending colon, sigmoid colon, and rectum).

Definition of OS: OS was defined as the period from the date of curative-intent surgery to death from any cause or the last follow-up date, whichever occurred first.

Observation parameters: The study focused on the following endpoints: (1) Mutation status of KRAS, PIK3CA, FBXW7, and TP53; (2) Associations between gene mutations and clinicopathological characteristics; and (3) Prognostic impact of gene mutations on OS.

Statistical analysis

All statistical analyses were performed using SPSS software (version 22.0; IBM Corp., Armonk, NY, United States). Categorical variables were summarized as n (%). χ² or Fisher’s exact tests were used to evaluate associations between gene mutation status (dependent variable) and clinicopathological features (independent variables).

For univariate analysis, binary unconditional logistic regression was applied with each gene’s mutation status as the dependent variable and potential influencing factors entered individually as covariates. Variables with P < 0.10 were subsequently included in multivariate logistic regression to identify independent predictors, with results presented as odds ratios and 95%CI.

For survival analysis, Kaplan-Meier curves were generated and compared using the log-rank test. Variables with P < 0.10 in univariable analyses were entered into multivariable Cox models (backward or enter method; proportional hazards assumption checked). Hazard ratios (HRs) and 95%CI were calculated. A two-sided P < 0.05 was considered statistically significant.

RESULTS

Mutation spectrum

Among the 87 patients included in this study, mutations were detected in KRAS (44.8%), PIK3CA (13.8%), FBXW7 (27.6%), and TP53 (64.4%) (Figure 1). Co-mutation analysis showed that KRAS and TP53 mutations co-occurred in 8 cases (9.2%), TP53 and FBXW7 in 12 cases (13.8%), and KRAS and PIK3CA in 4 cases (4.6%), while other gene pairs exhibited minimal co-occurrence (Figure 2). The clinicopathological correlations and survival analyses are summarized below.

Figure 1 Prognostic analysis of genetic mutations in colorectal cancer. This flowchart outlines the study design for the prognostic analysis of genetic mutations in 87 colorectal cancer cases. The process begins with patient cohort selection, followed by next-generation sequencing detection of mutations in the KRAS, PIK3CA, FBXW7, and TP53 genes. Subsequent steps include clinicopathological correlation analysis and survival analysis using Kaplan-Meier and Cox regression methods. The main findings demonstrate that KRAS mutation is an independent predictor of poor prognosis with an allele-specific impact, while PIK3CA, FBXW7, and TP53 mutations show limited or no independent prognostic value in this cohort. CRC: Colorectal cancer; HR: Hazard ratio; NGS: Next-generation sequencing.

Figure 2 Co-occurrence matrix of KRAS, PIK3CA, FBXW7, and TP53 genetic mutations. The heatmap illustrates the frequency of co-occurring mutations between the genes listed on the Y-axis and X-axis. The color intensity in each cell corresponds to the number of patients harboring mutations in both genes, as indicated by the scale bar (range: 0-12). Key observations include a strong co-occurrence between TP53 and FBXW7 mutations (12 patients), and between TP53 and KRAS mutations (8 patients). KRAS and PIK3CA mutations co-occurred in 4 patients. No patients had mutations in FBXW7 and KRAS, or FBXW7 and PIK3CA. Self-comparisons are zero by definition.

Clinicopathological correlates of gene mutations

KRAS mutation: KRAS mutation was significantly associated with several adverse pathological features, including poor differentiation, perineural invasion, advanced N and M stage, multiple tumors, larger maximum tumor diameter, and elevated preoperative tumor markers (P < 0.05 for all). No significant associations were observed with tumor location, lymphovascular invasion, preoperative pathology, T stage, or MSI status (Table 1).

Table 1 Association between KRAS mutation status and clinicopathological features, n (%).

Clinicopathological feature	Total (n = 87)	KRAS mutant (n = 39)	KRAS wild-type (n = 48)	χ² value	P value
Tumor location				4.64	0.098
Left colon	21 (24.14)	7 (17.95)	14 (29.17)
Rectum	46 (52.87)	19 (48.72)	27 (56.25)
Right colon	20 (22.99)	13 (33.33)	7 (14.58)
Differentiation grade				4.52	0.034
Well/moderate	72 (82.76)	36 (92.31)	36 (75.00)
Poor	15 (17.24)	3 (7.69)	12 (25.00)
Lymphovascular invasion				2.86	0.091
Absent	51 (58.62)	19 (48.72)	32 (66.67)
Present	36 (41.38)	20 (51.28)	16 (33.33)
Perineural Invasion				17.132	< 0.001
Absent	75 (86.21)	27 (69.23)	48 (100.00)
Present	12 (13.79)	12 (30.77)	0 (0.00)
Preoperative pathology				0.06	0.811
Adenocarcinoma (non-special)	82 (94.25)	36 (92.31)	46 (95.83)
Mucinous adenocarcinoma	5 (5.75)	3 (7.69)	2 (4.17)
T stage				2.60	0.107
T1 + T2	5 (5.75)	0 (0.00)	5 (10.42)
T3 + T4	82 (94.25)	39 (100.00)	43 (89.58)
N stage				11.31	< 0.001
N0	12 (13.79)	0 (0.00)	12 (25.00)
N1 + N2	75 (86.21)	39 (100.00)	36 (75.00)
M stage				7.22	0.007
M0	71 (81.61)	27 (69.23)	44 (91.67)
M1	16 (18.39)	12 (30.77)	4 (8.33)
Tumor number				17.13	< 0.001
Solitary	75 (86.21)	27 (69.23)	48 (100.00)
Multiple (≥ 2)	12 (13.79)	12 (30.77)	0 (0.00)
Maximum tumor diameter				21.39	< 0.001
< 5 cm	32 (36.78)	4 (10.26)	28 (58.33)
≥ 5 cm	55 (63.22)	35 (89.74)	20 (41.67)
Preoperative tumor markers				16.29	< 0.001
Normal	65 (74.71)	21 (53.85)	44 (91.67)
Elevated	22 (25.29)	18 (46.15)	4 (8.33)
Microsatellite				0.03	0.872
Microsatellite stable	81 (93.10)	37 (94.87)	44 (91.67)
Microsatellite instability	6 (6.90)	2 (5.13)	4 (8.33)

PIK3CA mutation: PIK3CA mutation, identified in 13.8% of cases, correlated significantly only with maximum tumor diameter (P < 0.05). No association was found with tumor location, differentiation, lymphovascular or perineural invasion, TNM stage, tumor number, biomarkers, or MSI status (Table 2).

Table 2 Association between PIK3CA mutation status and clinicopathological features, n (%).

Clinicopathological feature	Total (n = 87)	PIK3CA mutant (n = 12)	PIK3CA wild-type (n = 75)	χ² value	P value
Tumor location				-	0.134
Left colon	21 (24.14)	6 (50.00)	15 (20.00)
Rectum	46 (52.87)	4 (33.33)	42 (56.00)
Right colon	20 (22.99)	2 (16.67)	18 (24.00)
Differentiation grade				1.67	0.197
Well/moderate	72 (82.76)	12 (100.00)	60 (80.00)
Poor	15 (17.24)	0 (0.00)	15 (20.00)
Lymphovascular invasion				0.09	0.769
Absent	51 (58.62)	8 (66.67)	43 (57.33)
Present	36 (41.38)	4 (33.33)	32 (42.67)
Perineural invasion				1.08	0.298
Absent	75 (86.21)	12 (100.00)	63 (84.00)
Present	12 (13.79)	0 (0.00)	12 (16.00)
Preoperative pathology				-	0.533
Adenocarcinoma (non-special)	82 (94.25)	11 (91.67)	71 (94.67)
Mucinous adenocarcinoma	5 (5.75)	1 (8.33)	4 (5.33)
T stage				-	1.000
T1 + T2	5 (5.75)	0 (0.00)	5 (6.67)
T3 + T4	82 (94.25)	12 (100.00)	70 (93.33)
N stage				2.77	0.096
N0	75 (86.21)	8 (66.67)	67 (89.33)
N1 + N2	12 (13.79)	4 (33.33)	8 (10.67)
M stage				1.88	0.171
M0	71 (81.61)	12 (100.00)	59 (78.67)
M1	16 (18.39)	0 (0.00)	16 (21.33)
Tumor number				2.77	0.096
Solitary	75 (86.21)	8 (66.67)	67 (89.33)
Multiple (≥ 2)	12 (13.79)	4 (33.33)	8 (10.67)
Maximum tumor diameter				6.37	0.012
< 5 cm	32 (36.78)	0 (0.00)	32 (42.67)
≥ 5 cm	55 (63.22)	12 (100.00)	43 (57.33)
Preoperative tumor markers				0.11	0.739
Normal	65 (74.71)	8 (66.67)	57 (76.00)
Elevated	22 (25.29)	4 (33.33)	18 (24.00)
Microsatellite				-	0.590
Microsatellite stable	81 (93.10)	12 (100.00)	69 (92.00)
Microsatellite instability	6 (6.90)	0 (0.00)	6 (8.00)

FBXW7 mutation: FBXW7 mutation, detected in 27.6% of patients, was significantly related to tumor location and tumor size (P < 0.05). No significant correlations were observed with differentiation, lymphovascular or perineural invasion, TNM stage, tumor number, biomarkers, or MSI status (Table 3).

Table 3 Association between FBXW7 mutation status and clinicopathological features, n (%).

Clinicopathological feature	Total (n = 87)	FBXW7 mutant (n = 24)	FBXW7 wild-type (n = 63)	χ² value	P value
Tumor location				14.20	< 0.001
Left colon	21 (24.14)	4 (16.67)	17 (26.98)
Rectum	46 (52.87)	20 (83.33)	26 (41.27)
Right colon	20 (22.99)	0 (0.00)	20 (31.75)
Differentiation grade				0.01	1.000
Well/moderate	72 (82.76)	20 (83.33)	52 (82.54)
Poor	15 (17.24)	4 (16.67)	11 (17.46)
Lymphovascular invasion				0.88	0.347
Absent	51 (58.62)	16 (66.67)	35 (55.56)
Present	36 (41.38)	8 (33.33)	28 (44.44)
Perineural invasion				0.02	0.895
Absent	75 (86.21)	20 (83.33)	55 (87.30)
Present	12 (13.79)	4 (16.67)	8 (12.70)
Preoperative pathology				0.02	0.901
Adenocarcinoma (non-special)	82 (94.25)	22 (91.67)	60 (95.24)
Mucinous adenocarcinoma	5 (5.75)	2 (8.33)	3 (4.76)
T stage				0.82	0.365
T1 + T2	5 (5.75)	0 (0.00)	5 (7.94)
T3 + T4	82 (94.25)	24 (100.00)	58 (92.06)
N stage				0.02	0.895
N0	12 (13.79)	4 (16.67)	8 (12.70)
N1 + N2	75 (86.21)	20 (83.33)	55 (87.30)
M stage				3.65	0.056
M0	71 (81.61)	16 (66.67)	55 (87.30)
M1	16 (18.39)	8 (33.33)	8 (12.70)
Tumor number				3.82	0.051
Solitary	75 (86.21)	24 (100.00)	51 (80.95)
Multiple (≥ 2)	12 (13.79)	0 (0.00)	12 (19.05)
Maximum tumor diameter				5.77	0.016
< 5 cm	32 (36.78)	4 (16.67)	28 (44.44)
≥ 5 cm	55 (63.22)	20 (83.33)	35 (55.56)
Preoperative tumor markers				2.87	0.090
Normal	65 (74.71)	21 (87.50)	44 (69.84)
Elevated	22 (25.29)	3 (12.50)	19 (30.16)
Microsatellite				1.20	0.274
Microsatellite stable	81 (93.10)	24 (100.00)	57 (90.48)
Microsatellite instability	6 (6.90)	0 (0.00)	6 (9.52)

TP53 mutation: TP53 mutation was the most prevalent alteration (64.4%) and showed significant associations with lymphovascular invasion, perineural invasion, tumor multiplicity, elevated tumor markers, and MSI status (P < 0.05). However, no significant relationship was found with tumor location, differentiation, or TNM stage (Table 4).

Table 4 Association between TP53 mutation status and clinicopathological features, n (%).

Clinicopathological feature	Total (n = 87)	TP53 mutant (n = 56)	TP53 wild-type (n = 31)	χ² value	P value
Tumor location				1.86	0.394
Left colon	21 (24.14)	11 (19.64)	10 (32.26)
Rectum	46 (52.87)	32 (57.14)	14 (45.16)
Right colon	20 (22.99)	13 (23.21)	7 (22.58)
Differentiation grade				1.93	0.165
Well/moderate	72 (82.76)	44 (78.57)	28 (90.32)
Poor	15 (17.24)	12 (21.43)	3 (9.68)
Lymphovascular invasion				10.63	0.001
Absent	51 (58.62)	40 (71.43)	11 (35.48)
Present	36 (41.38)	16 (28.57)	20 (64.52)
Perineural invasion				6.01	0.014
Absent	75 (86.21)	44 (78.57)	31 (100.00)
Present	12 (13.79)	12 (21.43)	0 (0.00)
Preoperative pathology				0.07	0.786
Adenocarcinoma (non-special)	82 (94.25)	52 (92.86)	30 (96.77)
Mucinous adenocarcinoma	5 (5.75)	4 (7.14)	1 (3.23)
T stage				1.52	0.218
T1 + T2	5 (5.75)	5 (8.93)	0 (0.00)
T3 + T4	82 (94.25)	51 (91.07)	31 (100.00)
N stage				0.00	1.000
N0	12 (13.79)	8 (14.29)	4 (12.90)
N1 + N2	75 (86.21)	48 (85.71)	27 (87.10)
M stage				1.76	0.184
M0	71 (81.61)	48 (85.71)	23 (74.19)
M1	16 (18.39)	8 (14.29)	8 (25.81)
Tumor number				4.38	0.036
Solitary	75 (86.21)	52 (92.86)	23 (74.19)
Multiple (≥ 2)	12 (13.79)	4 (7.14)	8 (25.81)
Maximum tumor diameter				2.49	0.114
< 5 cm	32 (36.78)	24 (42.86)	8 (25.81)
≥ 5 cm	55 (63.22)	32 (57.14)	23 (74.19)
Preoperative tumor markers				13.60	< 0.001
Normal	65 (74.71)	49 (87.50)	16 (51.61)
Elevated	22 (25.29)	7 (12.50)	15 (48.39)
Microsatellite				8.82	0.003
Microsatellite stable	81 (93.10)	56 (100.00)	25 (80.65)
Microsatellite instability	6 (6.90)	0 (0.00)	6 (19.35)

Survival analysis

At a median follow-up of 60 months, the 3-year OS rate was 63.2% (55/87), and the 5-year OS rate was 41.4% (36/87). Univariate analysis identified KRAS mutation, diabetes, hypertension, family history, lymphovascular invasion, N stage, open surgical approach, tumor number, and elevated tumor markers as significant predictors of 3-year survival (P < 0.05). Conversely, gender, age, PIK3CA/FBXW7/TP53 mutations, hyperlipidemia, smoking, alcohol use, tumor location, differentiation, T/M stage, and MSI status were not significant (Table 5).

Table 5 Association between genetic mutations, clinicopathological features and prognosis (3-year survival), n (%).

Variable	Total (n = 87)	Death (n = 32)	Survival (n = 55)	χ² value	P value
Gender				0.17	0.681
Female	24 (27.59)	8 (25.00)	16 (29.09)
Male	63 (72.41)	24 (75.00)	39 (70.91)
Age				3.15	0.076
< 60 years	20 (22.99)	4 (12.50)	16 (29.09)
≥ 60 years	67 (77.01)	28 (87.50)	39 (70.91)
KRAS mutation				6.39	0.011
Mutant	39 (44.83)	20 (62.50)	19 (34.55)
Wild-type	48 (55.17)	12 (37.50)	36 (65.45)
PIK3CA mutation				0.00	1.000
Mutant	12 (13.79)	4 (12.50)	8 (14.55)
Wild-type	75 (86.21)	28 (87.50)	47 (85.45)
FBXW7 mutation				2.49	0.115
Mutant	24 (27.59)	12 (37.50)	12 (21.82)
Wild-type	63 (72.41)	20 (62.50)	43(78.18)
TP53 mutation				2.49	0.114
Mutant	56 (64.37)	24 (75.00)	32 (58.18)
Wild-type	31 (35.63)	8 (25.00)	23 (41.82)
Diabetes				12.32	< 0.001
Absent	71 (81.61)	20 (62.50)	51 (92.73)
Present	16 (18.39)	12 (37.50)	4 (7.27)
Hypertension				6.39	0.011
Absent	48 (55.17)	12 (37.50)	36 (65.45)
Present	39 (44.83)	20 (62.50)	19 (34.55)
Hyperlipidemia				1.17	0.279
Absent	71 (81.61)	28 (87.50)	43 (78.18)
Present	16 (18.39)	4 (12.50)	12 (21.82)
Smoking				1.17	0.280
Absent	63 (72.41)	21 (65.62)	42 (76.36)
Present	24 (27.59)	11 (34.38)	13 (23.64)
Alcohol use				2.46	0.116
Absent	62 (71.26)	26 (81.25)	36 (65.45)
Present	25 (28.74)	6 (18.75)	19 (34.55)
Tumor location				0.94	0.625
Left colon	21 (24.14)	8 (25.00)	13 (23.64)
Rectum	46 (52.87)	15 (46.88)	31 (56.36)
Right colon	20 (22.99)	9 (28.12)	11 (20.00)
Differentiation grade				0.80	0.372
Well/moderate	72 (82.76)	28 (87.50)	44 (80.00)
Poor	15 (17.24)	4 (12.50)	11 (20.00)
Lymphovascular invasion				5.60	0.018
Absent	51 (58.62)	24 (75.00)	27 (49.09)
Present	36 (41.38)	8 (25.00)	28 (50.91)
Perineural invasion				0.00	1.000
Absent	75 (86.21)	28 (87.50)	47 (85.45)
Present	12 (13.79)	4 (12.50)	8 (14.55)
Preoperative pathology				0.00	1.000
Adenocarcinoma (non-special)	82 (94.25)	30 (93.75)	52 (94.55)
Mucinous adenocarcinoma	5 (5.75)	2 (6.25)	3 (5.45)
T stage				1.64	0.201
T1 + T2	5 (5.75)	0 (0.00)	5 (9.09)
T3 + T4	82 (94.25)	32 (100.00)	50 (90.91)
N stage				6.37	0.012
N0	12 (13.79)	0 (0.00)	12 (21.82)
N1 + N2	75 (86.21)	32 (100.00)	43 (78.18)
M stage				1.47	0.225
M0	71 (81.61)	24 (75.00)	47 (85.45)
M1	16 (18.39)	8 (25.00)	8 (14.55)
Surgical approach				5.93	0.015
Laparoscopic	60 (68.97)	17 (53.12)	43 (78.18)
Open	27 (31.03)	15 (46.88)	12 (21.82)
Tumor number				3.96	0.047
Solitary	75 (86.21)	24 (75.00)	51 (92.73)
Multiple (≥ 2)	12 (13.79)	8 (25.00)	4 (7.27)
Maximum tumor diameter				3.02	0.082
< 5 cm	32 (36.78)	8 (25.00)	24 (43.64)
≥ 5 cm	55 (63.22)	24 (75.00)	31 (56.36)
Preoperative tumor markers				4.00	0.046
Absent	65 (74.71)	20 (62.50)	45 (81.82)
Present	22 (25.29)	12 (37.50)	10 (18.18)
Microsatellite				0.38	0.535
Microsatellite stable	81 (93.10)	31 (96.88)	50 (90.91)
Microsatellite instability	6 (6.90)	1 (3.12)	5 (9.09)

Regarding 5-year survival, significant prognostic variables included gender, KRAS mutation, alcohol use, family history, tumor location and differentiation, lymphovascular invasion, T and N stage, tumor size, number, and preoperative biomarkers (P < 0.05). No significant associations were found with age, PIK3CA/FBXW7/TP53 mutations, diabetes, hypertension, hyperlipidemia, smoking, M stage, surgical approach, or MSI status (Table 6).

Table 6 Association between genetic mutations, clinicopathological features and prognosis (5-year survival), n (%).

Variable	Total (n = 87)	Death (n = 51)	Survival (n = 36)	χ² value	P value
Gender				8.74	0.003
Female	24 (27.59)	8 (15.69)	16 (44.44)
Male	63 (72.41)	43 (84.31)	20 (55.56)
Age				1.99	0.159
< 60 years	20 (22.99)	9 (17.65)	11 (30.56)
≥ 60 years	67 (77.01)	42 (82.35)	25 (69.44)
KRAS mutation				12.69	< 0.001
Mutant	39 (44.83)	31 (60.78)	8 (22.22)
Wild-type	48 (55.17)	20 (39.22)	28 (77.78)
PIK3CA mutation				2.56	0.110
Mutant	12 (13.79)	4 (7.84)	8 (22.22)
Wild-type	75 (86.21)	47 (92.16)	28 (77.78)
FBXW7 mutation				1.02	0.314
Mutant	24 (27.59)	12 (23.53)	12 (33.33)
Wild-type	63 (72.41)	39 (76.47)	24 (66.67)
TP53 mutation				0.14	0.707
Mutant	56 (64.37)	32 (62.75)	24 (66.67)
Wild-type	31 (35.63)	19 (37.25)	12 (33.33)
Diabetes				2.17	0.141
Absent	71 (81.61)	39 (76.47)	32 (88.89)
Present	16 (18.39)	12 (23.53)	4 (11.11)
Hypertension				3.28	0.070
Absent	48 (55.17)	24 (47.06)	24 (66.67)
Present	39 (44.83)	27 (52.94)	12 (33.33)
Hyperlipidemia				0.60	0.438
Absent	71 (81.61)	43 (84.31)	28 (77.78)
Present	16 (18.39)	8 (15.69)	8 (22.22)
Smoking				2.04	0.153
Absent	63 (72.41)	34 (66.67)	29 (80.56)
Present	24 (27.59)	17 (33.33)	7 (19.44)
Alcohol use				12.48	< 0.001
Absent	62 (71.26)	29 (56.86)	33 (91.67)
Present	25 (28.74)	22 (43.14)	3 (8.33)
Tumor location				6.07	0.048
Left colon	21 (24.14)	13 (25.49)	8 (22.22)
Rectum	46 (52.87)	22 (43.14)	24 (66.67)
Right colon	20 (22.99)	16 (31.37)	4 (11.11)
Differentiation grade				12.79	< 0.001
Well/moderate	72 (82.76)	36 (70.59)	36 (100.00)
Poor	15 (17.24)	15 (29.41)	0 (0.00)
Lymphovascular invasion				5.09	0.024
Absent	51 (58.62)	35 (68.63)	16 (44.44)
Present	36 (41.38)	16 (31.37)	20 (55.56)
Perineural invasion				0.09	0.769
Absent	75 (86.21)	43 (84.31)	32 (88.89)
Present	12 (13.79)	8 (15.69)	4 (11.11)
Preoperative pathology				0.28	0.595
Adenocarcinoma (non-special)	82 (94.25)	47 (92.16)	35 (97.22)
Mucinous adenocarcinoma	5 (5.75)	4 (7.84)	1 (2.78)
T stage				5.17	0.023
T1 + T2	5 (5.75)	0 (0.00)	5 (13.89)
T3 + T4	82 (94.25)	51 (100.00)	31 (86.11)
N stage				17.02	< 0.001
N0	12 (13.79)	0 (0.00)	12 (33.33)
N1 + N2	75 (86.21)	51 (100.00)	24 (66.67)
M stage				2.17	0.141
M0	71 (81.61)	39 (76.47)	32 (88.89)
M1	16 (18.39)	12 (23.53)	4 (11.11)
Surgical approach				2.23	0.136
Laparoscopic	60 (68.97)	32 (62.75)	28 (77.78)
Open	27 (31.03)	19 (37.25)	8 (22.22)
Tumor number				7.95	0.005
Solitary	75 (86.21)	39 (76.47)	36 (100.00)
Multiple (≥ 2)	12 (13.79)	12 (23.53)	0 (0.00)
Maximum tumor diameter				9.31	0.002
< 5 cm	32 (36.78)	12 (23.53)	20 (55.56)
≥ 5 cm	55 (63.22)	39 (76.47)	16 (44.44)
Preoperative tumor markers				9.34	0.002
Absent	65 (74.71)	32 (62.75)	33 (91.67)
Present	22 (25.29)	19 (37.25)	3 (8.33)
Microsatellite				0.76	0.382
Microsatellite stable	81 (93.10)	49 (96.08)	32 (88.89)
Microsatellite instability	6 (6.90)	2 (3.92)	4 (11.11)

Kaplan-Meier survival curves: As illustrated in Figure 3A, patients harboring KRAS mutations exhibited markedly poorer prognosis than those with wild-type KRAS. The log-rank test confirmed a statistically significant difference in OS between the two groups (P = 0.012), and the Cox proportional hazards model further established KRAS mutation as a strong negative prognostic indicator, with a HR of 3.57 (95%CI: 1.85-6.89). Similarly, a significant survival difference was observed for PIK3CA mutation status (Figure 3B). The log-rank P value was 0.027. Notably, the Cox model indicated that patients with mutant PIK3CA had a significantly lower risk of death compared to wild-type patients (HR = 0.26, 95%CI: 0.09-0.74). In contrast, survival outcomes did not differ significantly between mutant and wild-type groups for FBXW7 (P = 0.213, HR = 1.86, 95%CI: 0.85-4.05; Figure 3C), or TP53 (P = 0.340, HR = 1.68, 95%CI: 0.87-3.25; Figure 3D).

Figure 3 Overall survival by KRAS, PIK3CA, FBXW7 and TP53 status. A: Overall survival (OS) by KRAS status. Log-rank P = 0.012. Hazard ratio (HR) (mutant vs wild-type) from Cox model: 3.57 (95%CI: 1.85-6.89); B: OS by PIK3CA status. Log-rank P = 0.027. HR (mutant vs wild-type) from Cox model: 0.26 (95%CI: 0.09-0.74); C: OS by FBXW7 status. Log-rank P = 0.213. HR (mutant vs wild-type) from Cox model: 1.86 (95%CI: 0.85-4.05); D: OS by TP53 status. Log-rank P = 0.340. HR (mutant vs wild-type) from Cox model: 1.68 (95%CI: 0.87-3.25).

Multivariable cox regression analysis: Multivariable Cox proportional hazards regression identified KRAS mutation as an independent predictor of reduced survival (HR = 3.567, 95%CI: 1.846-6.893, P < 0.001). Conversely, PIK3CA mutation appeared to exert a protective effect (HR = 0.260, 95%CI: 0.092-0.739, P = 0.011). Gender also emerged as a significant independent variable (HR = 0.296, P = 0.011), suggesting potential sex-related differences in long-term outcomes.

Mutations in TP53 (HR = 1.676, P = 0.126) and FBXW7 (P = 0.120), as well as age (P = 0.316), were not significantly associated with survival after adjustment for confounders (Table 7).

Table 7 Cox regression analysis of KRAS, PIK3CA, FBXW7, and TP53 gene mutation status along with other high-risk factors.

Variable	B	SE	Wald	df	Significance	Exp(B)	95%CI for Exp(B)
							Lower bound	Upper bound
Gender	-1.219	0.476	6.542	1	0.011	0.296	0.116	0.752
Age	-0.432	0.43	1.007	1	0.316	0.649	0.279	1.509
KRAS mutant	1.272	0.336	14.313	1	< 0.001	3.567	1.846	6.893
PIK3CA mutant	-1.345	0.532	6.391	1	0.011	0.26	0.092	0.739
FBXW7 mutant	0.619	0.397	2.423	1	0.12	1.857	0.852	4.046
TP53 mutant	0.516	0.337	2.344	1	0.126	1.676	0.865	3.246

DISCUSSION

Principal findings

In this single-center cohort of 87 surgically treated CRCs, mutation status across four recurrently altered genes – KRAS, PIK3CA, FBXW7, and TP53 – showed distinct clinicopathological correlates, but only KRAS carried clear survival implications. KRAS mutations (44.83%) associated with several adverse features – perineural invasion, nodal and distant metastasis, larger tumor size, multifocality, and elevated preoperative tumor markers – while the distribution of differentiation grade was paradoxically more favorable in the mutant group. By contrast, PIK3CA mutations (13.79%) tracked primarily with larger tumor size, and FBXW7 mutations (27.59%) were enriched in rectal tumors and larger lesions, without detectable survival impact. TP53 mutations were common (64.37%) and linked to perineural invasion but occurred almost exclusively in microsatellite-stable tumors and did not influence OS. On univariable analyses, 3-year survival (63.22%) was shaped by KRAS status and several clinical factors (e.g., diabetes, hypertension, lymphovascular invasion, N stage, tumor multiplicity, surgical approach, and biomarkers), whereas 5-year survival (41.38%) was associated with KRAS, sex, alcohol use, tumor location and differentiation, lymphovascular invasion, T/N stage, tumor burden, and biomarkers.

Context with prior evidence

The adverse prognostic signal of KRAS is coherent with its central role in MAPK pathway activation and with extensive literature linking KRAS to metastatic propensity and resistance to anti-epidermal growth factor receptor (EGFR) therapy[13]. The present data reinforce that – even outside the metastatic setting – KRAS mutation captures biologic aggressiveness reflected by nodal and distant spread and by larger, multifocal tumors[14]. The absence of a survival signal for PIK3CA and FBXW7 in this cohort aligns with prior heterogeneous reports where their effects are often context-dependent (e.g., co-mutational patterns, sidedness, treatment exposures)[15,16]. For TP53, the strong enrichment in microsatellite stable disease and lack of survival separation underscore the molecular heterogeneity of TP53-mutant CRC: Functional consequences vary by mutation class and co-driver background, which a modestly sized cohort and univariable analyses may not fully resolve[17,18].

Biological and clinical interpretation

KRAS as a composite risk marker. The clustering of nodal/distant spread, larger size, and multiplicity within the KRAS-mutant subset suggests KRAS captures a high-risk biological state that is already phenotypically manifest at presentation. Given its robust association with worse 3-year and 5-year survival, KRAS genotyping should inform perioperative risk discussions and postoperative surveillance intensity[19].

The PIK3CA mutation is an independent significant protective factor for survival outcomes (P = 0.011), and is not merely associated with tumor size. This suggests that it may confer a definite survival advantage through some mechanism. In contrast, the FBXW7 mutation shows no statistically significant association with survival outcomes[20,21].

TP53 and MSI decoupling. The inverse relationship between TP53 mutation and MSI in this cohort (MSI confined to TP53 wild-type) is consistent with the canonical CIN-MSI dichotomy in CRC biology. The mixed pattern for histologic aggressiveness (increased perineural invasion yet lower rates of lymph vascular invasion, multiple lesions, and elevated markers among TP53-mutant cases) suggests subtype heterogeneity and potential confounding by tumor location and co-mutations[22].

Furthermore, beyond the broad prognostic impact of KRAS mutation, emerging evidence highlights the particular clinical significance of specific subtypes, such as the KRAS glycine-to-cysteine mutation at codon 12 (KRAS G12C). KRAS G12C has been associated with poorer responses to conventional chemotherapy and inferior survival compared to other KRAS mutations[23]. The recent development of allele-specific KRAS G12C inhibitors (e.g., sotorasib and adagrasib) represents a paradigm shift in targeting previously ‘undruggable’ oncogenes. Promising results from phase I/II clinical trials in various solid tumors, including CRC, underscore the potential of these agents[24,25]. Although our current cohort size precluded a robust subgroup analysis of KRAS G12C, future studies with larger cohorts should prioritize allele-specific profiling to identify patients who may benefit from these novel targeted therapies. This precision approach could significantly alter the clinical management of KRAS-mutant CRC.

It is also important to note that KRAS-driven tumors often exhibit robust adaptive resistance mechanisms, frequently involving feedback loops and pathway reactivation (e.g., via EGFR, mesenchymal-epithelial transition factor, or Wnt signaling). Therefore, the adequate treatment of KRAS-mutated CRC will likely require rational combination therapies rather than monotherapy approaches. For instance, combining KRAS G12C inhibitors with EGFR inhibitors has shown enhanced efficacy in preclinical models and early clinical trials by preventing adaptive feedback resistance[25]. Similarly, combinations with MEK inhibitors, immunotherapy, or other targeted agents are being actively explored to achieve durable responses. Future therapeutic strategies should focus on leveraging multi-agent regimens tailored to the specific molecular context of the tumor to overcome inherent resistance.

Implications for practice

Risk stratification. Incorporating KRAS status alongside conventional factors (T/N stage, lymph vascular invasion, tumor size/number, and biomarkers) can refine short- and intermediate-term risk estimates after curative resection.

Therapeutic tailoring. In line with current practice standards, anti-EGFR therapy should be avoided in KRAS-mutant disease; where available, clinical-trial enrollment for KRAS-targeted strategies may be prioritized in high-risk adjuvant or oligometastatic contexts. For microsatellite stable/TP53-mutant tumors, the lack of a survival signal here counsels against using TP53 alone for adjuvant decisions; multi-parametric models are preferred.

Strengths and limitations

Strengths include uniform surgical management at a single institution, centralized pathology review, and systematic NGS across four key genes. Limitations include the retrospective design, modest sample size (which constrains statistical power for subgroup and interaction analyses and may introduce selection bias), lack of detailed treatment exposure data (adjuvant regimens, biologics), and absence of broader molecular measurements (e.g., APC/BRAF, tumor mutational burden, transcriptomic subtypes) that could contextualize co-mutational effects and pathways. Although this study made every effort to control for confounding factors, it must be acknowledged that the sample size constitutes a constraint on statistical power. In the multivariable Cox regression model, the ratio of the number of covariates to the number of events suggests that caution is warranted when interpreting the results to avoid the model overfitting the specific characteristics of the current dataset.

Future directions

Co-mutation landscapes: Dissect the prognostic impact of KRAS in the context of APC/TP53/PIK3CA/BRAF and sidedness; examine whether FBXW7 effects emerge in specific genomic backgrounds.

Biology-to-therapy bridges: Map KRAS-driven programs (e.g., invasion/angiogenesis signatures) against therapeutic vulnerabilities to nominate rational adjuvant or perioperative combinations.

Prospective cohorts: Embed comprehensive molecular profiling and standardized adjuvant pathways to separate biology from treatment effects on survival.

CONCLUSION

In resected CRC, KRAS mutation robustly delineates a high-risk phenotype and portends inferior 3-year and 5-year survival, whereas FBXW7, and TP53 – despite specific clinicopathological associations – do not independently stratify outcomes in this cohort. Routine inclusion of KRAS testing in perioperative decision-making and surveillance planning is justified, while the prognostic roles of PIK3CA likely depend on broader molecular context and should be revisited in larger, prospectively profiled datasets.