Age-dependent prognostic value of KRAS mutation in metastatic colorectal cancer

Abstract

Background:

The age-dependent prognostic impact of KRAS status in metastatic colorectal cancer (mCRC) is unknown.

Materials & Methods:

We used the National Cancer Database to evaluate the survival by KRAS status for age-groups <50, 50–69 and ≥70, adjusting for relevant patient and tumor characteristics.

Results:

mCRC patients (n = 26,095; 33.5%) had KRAS status reported, and 11,338 of these patients (43.4%) had mutations in the KRAS gene. Patients with KRAS mutations had worse overall survival than wild-type KRAS patients. In age-groups <50 years (23 vs 29 months; p < 0.001) and 50–69 (21 vs 23.4 months; p < 0.001), KRAS mutations were significantly associated with worse survival, whereas in the ≥70-year age-group, there was no significant association (14 vs 14 months; p = 0.34).

Conclusion:

We conclude that the age of patients influences the prognostic value of KRAS mutation in metastatic colorectal cancer.

Colorectal cancer carcinogenesis is a complex multilevel process characterized by structural and functional DNA mutations [1–3]. Kirsten-rat sarcoma (KRAS) is a proto-oncogene, member of the Ras family, encoding a small 21 kD guanosine triphosphate/guanosine diphosphate binding protein. It has a well-established role in cellular growth, migration, apoptosis and cellular regulation [1,4]. Other important Ras family genes involved in cancer development include the NRAS gene and HRAS gene. KRAS mutation rate is found to be between 30 and 50% in colorectal cancer patients [5,6], and population-based analyses showed an increased risk of recurrence and death in these patients [7,8]. Patients harboring these mutations have a higher incidence of metastases consistent with more aggressive tumor biology [9]. KRAS mutations are a negative predictor of response to anti-EGF receptor (EGFR) monoclonal antibody therapies (cetuximab and panitumumab) [10]. Several meta-analyses have confirmed the negative prognostic role of KRAS mutation in patients with stage IV colon cancer [9,11] including those undergoing definitive resection of liver metastasis [9,12,13]. Determining KRAS mutational status before or during first-line therapy is now considered a standard of care in patients with metastatic colorectal cancer [14,15]. Left-sided colorectal tumors have better response and survival with the early use of anti-EGFR agents in KRAS wildtype patients than right-sided tumors [16]. Age is also an important factor for treatment response, but it is unknown whether the impact of KRAS mutation in metastatic colorectal cancer is different according to patient age. Also, the incidence of early-onset colorectal cancer is rising globally, especially in patients with left-sided tumors. The objective of this study was to determine the prognostic impact of KRAS mutation status in different age-groups of metastatic colorectal cancer patients.

Materials & methods

We queried the data from the National Cancer Database (NCDB), which is the largest nationwide oncology dataset, including more than 1500 Commission on Cancer (CoC)-accredited facilities and ∼70% of cancer cases. Data include ∼34 million historical records, such as socioeconomic status, staging, treatments and outcomes (www.facs.org%/quality-programs/cancer/ncdb). This study was exempt from the University of Texas Southwestern Medical Center Institutional Review Board due to deidentified data.

Study population

We included adult patients (≥18 years) with stage IV colorectal cancer diagnosed from 2010 to 2016. We selected colon adenocarcinoma using “C18.0 (cecum), C18.2-C18.7 (ascending colon, hepatic flexure, transverse colon, splenic flexure, descending colon, sigmoid colon), C19.9 (rectosigmoid junction) and C20.9 (rectum)” ICD-O-3/WHO 2008 site recode for topography, and “8140–8147, 8210–8211, 8220–8221, 8260–8263, 8480–8481 and 8490” ICD-O-3 codes for histology [17,18]. We excluded cases if the tumor's KRAS mutation status was not reported, if patients had more than one primary tumor, had no follow-up data, had other than stage IV disease or did not receive the first course of treatment at the reporting facility (Figure 1). Pathological AJCC (6th and 7th editions) stage group was used for staging.

Figure 1. . Inclusion and exclusion criteria.

NCDB: National Cancer Database.

Endpoints

We first described differences in the colorectal cancer patients' baseline characteristics in the database that KRAS status reported compared with patients where this mutation status was not reported. We then focused on patients with reported KRAS mutation status and divided this population into two groups: patients with KRAS mutation and wild-type KRAS. We compared characteristics and outcomes between the two groups based on the age-groups at the time of diagnosis. The age-groups were defined as <50 years, 50–69 years and ≥70 years [18,19]. The age-group <50 was considered young-onset colorectal cancer, age-group 50–69 as a standard colorectal group and ≥70 were considered an older age-group. We compared patient demographics including sex (male, female), facility type (academic, nonacademic, other), race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic and other individuals), facility location (New England, Middle Atlantic, South Atlantic, East North Central, East South Central, West North Central, West South Central, Mountain, Pacific), Charlson-Deyo Score (0–2+) [20], and rurality (metropolitan, nonmetropolitan). We also compared socioeconomic factors including insurance status (uninsured, private insurance, government insurance), education level (rates of patients without high school level ≥17.6%, 10.9–17.5%, 76.3–10.8%, <6.3%), travel distance to treatment facility (<12.5, 12.5–49.9, ≥50 miles) and median income quartiles (<$40,227, $40,227–50,353, $50,354–63,332, ≥$63,333). We used great circle distance (between patient's residential zip code and reporting facility location zip code) to define travel distance [21].

Tumor location was classified as right-sided colon (cecum, ascending colon, hepatic flexure, transverse colon) and left-sided colon (splenic flexure, descending colon, sigmoid colon, rectosigmoid junction and rectum) [13,18]. We reported the carcinoembryonic antigen (CEA) as normal or elevated using collaborative stage site-specific factor 1. Treatments were radiotherapy (yes, no), chemotherapy (yes, no) and surgery (yes, no), coded as “partial colectomy (30), subtotal colectomy (40), total colectomy (60),” total proctocolectomy, colectomy or coloproctotectomy with resection of contiguous organs (70) [22].

Statistical analysis

The chi-square test or Fisher's exact test was used to compare patients' categorical variables between mutated KRAS and wild-type KRAS patients. Overall survival was evaluated using the Kaplan–Meier method and compared with the log-rank test. Multivariable analysis was performed using the Cox regression method (with hazard ratio [HR] [23] and 95% CI) after adjusting for sex, age, race and ethnicity, facility type, comorbidity score, rurality, travel distance, income, insurance, tumor side and chemotherapy. We evaluated the interaction term between KRAS status and age-groups in the multivariable Cox regression analysis. All statistical tests were provided with a two-sided p-value using the SPSS version 25.0, and p < 0.05 was considered statistically significant.

Results

Baseline characteristics

From 2010 to 2016, there were 77,719 metastatic colorectal cancer patients reported in the NCDB database. Of the total population, 26,095 cases reported status for KRAS mutation. The remaining patients who did not have any KRAS status reported (28,569 patients), such patients were more likely to be older (40 vs 23%; p < 0.001), had medicare insurance (50 vs 35%, p < 0.001), had comorbidity score ≥2 (8 vs 5%; p < 0.001) and were treated at the nonacademic facility (69% vs 63%; p < 0.001) compared with patients with known KRAS status. Also, such patients with no reported KRAS status were less likely to have received chemotherapy (60 vs 88%; p < 0.001) and surgery (46 vs 59%; p < 0.001). In the population where KRAS mutation was reported, 14,757 (56.6%) had wild-type KRAS and 11,338 (43.4%) had KRAS mutation. Patients with KRAS mutation were more likely to be older than 70 years (24 vs 22.4%; p = 0.001), female (48 vs 43.8%; p < 0.001), and non-Hispanic Black individuals (17.8 vs 12.2%; p < 0.001). A total of 21% of patients with KRAS mutation were younger than 50 years, and among the younger group, 42% of patients had KRAS mutation. The patients with KRAS mutation belonged more frequently to the lower-income group <40,000 (21.1 vs 19%; p < 0.001) and were treated at an academic facility (38 vs 35.9%; p = 0.001). Similarly, right-sided cancers (35.6 vs 33.9%; p < 0.001) and CEA level elevation (87.1 vs 81.5%; p < 0.001) were more frequent in KRAS mutation patients. Other demographics and socioeconomic variables, comorbidity score, MSI positivity, rate of chemotherapy, radiotherapy administration and underwent surgery were similar between groups (Table 1).

Table 1. . Baseline characteristics.

Variables	KRAS wildtype (%)	KRAS mutated (%)	p-value
	14,757 (56.6)	11,338 (43.4)
Gender			<0.001
Male	8289 (56.2)	5899 (52.0)
Female	6468 (43.8)	5439 (48.0)
Age at diagnosis			0.001
<50 years	3338 (22.6)	2391 (21.1)
50–69 years	8119 (55.0)	6226 (54.9)
≥70 years	3300 (22.4)	2721 (24.0)
Race and ethnicity‡			<0.001
Non-Hispanic white	10,866 (73.8)	7776 (68.8)
Non-Hispanic black	1794 (12.2)	2016 (17.8)
Hispanics	996 (6.6)	751 (6.6)
Others	1088 (7.4)	757 (6.7)
Comorbidity score			0.36
0	11,506 (78.0)	8766 (77.3)
1	2457 (16.6)	1963 (17.3)
2+	794 (5.4)	609 (5.4)
Facility type‡			0.001
Academic	4926 (35.9)	4061 (38.0)
Nonacademic	8812 (64.1)	6622 (62.0)
Facility location‡			0.28
New England	786 (5.7)	594 (5.6)
Middle Atlantic	2274 (16.6)	1860 (17.4)
South Atlantic	2878 (20.9)	2313 (21.7)
East North Central	2383 (17.3)	1851 (17.3)
East South Central	844 (6.1)	613 (5.7)
West North Central	1174 (8.5)	891 (8.3)
West South Central	1079 (7.9)	859 (8.0)
Mountain	694 (5.1)	512 (4.8)
Pacific	1626 (11.8)	1190 (11.1)
Insurance status‡			0.63
Uninsured	900 (6.2)	711 (6.3)
Private	6842 (46.9)	5150 (46.0)
Medicaid	1621 (11.1)	1261 (11.3)
Medicare	5041 (34.6)	3942 (35.2)
Other government	173 (1.2)	141 (1.3)
Income‡			<0.001
<$40,227	2760 (19.0)	2361 (21.1)
$40,227–50,353	3378 (23.2)	2510 (22.4)
$50,354–63,332	3469 (23.9)	2551 (22.8)
≥$63,333	4938 (33.9)	3765 (33.7)
Education†,‡			0.008
≥17.6%	3149 (21.6)	2616 (23.3)
10.9–17.5%	3877 (26.6)	2927 (26.1)
6.3–10.8%	4194 (28.8)	3102 (27.7)
<6.3	3351 (23.0)	2559 (22.8)
Rurality‡			0.03
Metropolitan	12,040 (83.6)	9,376 (84.6)
Non-metropolitan	2358 (16.4)	1703 (15.4)
Travel distance‡			0.01
<12.5 miles	8531 (58.0)	6757 (59.7)
12.5–49.9 miles	4567 (31.0)	3323 (29.4)
≥50 miles	1621 (11.0)	1234 (10.9)
CEA‡			<0.001
Normal	2166 (18.5)	1148 (12.9)
Elevated	9521 (81.5)	7775 (87.1)
Tumor side			<0.001
Right	4712 (33.9)	4851 (45.6)
Left	9175 (66.1)	5780 (54.4)
Chemotherapy‡			0.82
No	1684 (11.6)	1282 (11.6)
Yes	12,779 (88.4)	9816 (88.4)
Chemotherapy types‡			0.37
Single-agent	1082 (8.7)	864 (9.1)
Multi-agent	11,330 (91.3)	8666 (90.9)
Radiotherapy‡			0.67
No	12,958 (88.2)	9931 (88.0)
Yes	1730 (11.8)	1348 (12.0)
Surgery			0.038
No	5908 (40.3)	4686 (41.6)
Yes	8739 (59.7)	6575 (58.4)
Overall survival			<0.001
Median time (months)	22.2	19.9

^†Education: percentage of patients without high school level.

^‡Unknown data excluded.

CEA: Carcinoembryonic antigen.

Survival analyses

Patients with KRAS mutation had worse overall survival (OS) than patients with wild-type KRAS (median OS 19.9 vs 22.2 months; p < 0.001). In multivariable analysis, KRAS mutation status was an independent poor prognostic factor after adjusting for available confounding factors (HR: 0.92 [95% CI: 0.89–0.95]; p < 0.001; Table 2). The interaction term between KRAS status and age-groups was significant (p < 0.001), which showed that the prognosis of KRAS status is different based on age-groups. Therefore, we reanalyzed OS using univariable and multivariable methods by KRAS status for each age-group (Table 3). The univariable analysis showed that patients with KRAS mutation had worse OS among <50 years old (23.5 vs 28.8 months; p < 0.001) and in 50–69 years old (21 vs 23.4 months; p < 0.001) metastatic colorectal cancer patients (Figure 2). The overall survival between KRAS mutation versus KRAS wild-type patients among ≥70 old patients was not significantly different (14.3 vs 13.9 months; p = 0.34). This age-related differences remained significant after multivariable analysis (HR: 0.80 [95% CI: 0.75–0.86] for <50; HR: 0.93 [95% CI: 0.89–0.97] for 50–69, respectively; Table 3).

Table 2. . Multivariable Cox regression analysis for overall survival.

Characteristics	Overall survival
	HR (95% CI)	p-value
KRAS status
KRAS mutated	Ref
KRAS wildtype	0.92 (0.89–0.95)	<0.001
Gender
Male	Ref
Female	0.96 (0.94–0.99)	0.011
Age at diagnosis
<50 years	Ref
50–69 years	1.07 (1.03–1.12)	0.001
≥70 years	1.30 (1.23–1.38)	<0.001
Race and ethnicity
Non-Hispanic white	Ref
Non-Hispanic black	0.97 (0.93–1.02)	0.20
Hispanics	0.74 (0.69–0.78)	<0.001
Other/unknown	0.94 (0.89–0.99)	0.038
Comorbidity score
0	Ref
1	1.06 (1.02–1.10)	0.004
2+	1.17 (1.10–1.25)	<0.001
Facility type
Academic	Ref
Nonacademic	1.19 (1.15–1.22)	<0.001
Unknown	1.12 (1.04–1.20)	0.002
Travel distance
<12.5 miles	Ref
12.5–49.9 miles	0.94 (0.90–0.97)	<0.001
≥50 miles	0.84 (0.80–0.89)	<0.001
Unknown	0.95 (0.68–1.31)	0.74
Income
<$40,227	Ref
$40,227–50,353	0.95 (0.91–0.99)	0.037
$50,354–63,332	0.96 (0.92–1.00)	0.075
≥$63,333	0.91 (0.87–0.95)	<0.001
Unknown	1.07 (0.94–1.23)	0.29
Insurance status
Private	Ref
Uninsured	1.22 (1.14–1.29)	<0.001
Medicaid	1.14 (1.09–1.20)	<0.001
Medicare	1.18 (1.13–1.23)	<0.001
Other government/unknown	1.09 (0.99–1.20)	0.08
Rurality
Metropolitan	Ref
Non-metropolitan	1.09 (1.04–1.14)	<0.001
Unknown	1.05 (0.95–1.15)	0.36
Tumor side
Right	Ref
Left	0.68 (0.66–0.71)	<0.001
Unspecified	0.94 (0.88–1.00)	0.05
Chemotherapy
Yes	Ref
No	2.47 (2.36–2.58)	<0.001
Unknown	1.21 (1.09–1.34)	<0.001
Radiotherapy
Yes	Ref
No	1.08 (1.03–1.14)	0.001
Unknown	0.85 (0.69–1.05)	0.14
Surgery
Yes	Ref
No	2.10 (2.04–2.17)	<0.001
Unknown	1.12 (0.94–1.33)	0.22
Interaction term of age and KRAS status		<0.001
Age (<50 vs 50–69)* KRAS status (KRAS mutated vs KRAS wildtype)	1.14 (1.06–1.23)	0.001
Age (<50 vs ≥70)* KRAS status (KRAS mutated vs KRAS wildtype)	1.20 (1.10–1.30)	<0.001

HR: Hazard ratio.

Table 3. . Log-rank tests and multivariable Cox regression analysis for overall survival among age-groups.

Characteristics	n (%)	Univariable		Multivariable
		Median time (months)	Log-rank p-value	HR (95% CI)	p-value
<50 years	5,729 (100)		<0.001
KRAS mutated	2,391 (41.7)	23.5		Ref
KRAS wildtype	3,338 (58.3)	28.8		0.80 (0.75–0.86)	<0.001
50–69 years	14,345 (100)		<0.001
KRAS mutated	6,226 (43.4)	21.0		Ref
KRAS wildtype	8,119 (56.6)	23.4		0.93 (0.89–0.97)	<0.001
≥70 years	6,021 (100)		0.34
KRAS mutated	2,721 (45.2)	14.3		Ref
KRAS wildtype	3300 (54.8)	13.9		1.00 (0.95–1.06)	0.89

HR: Hazard ratio; Ref: Reference.

Figure 2. . Overall survival based on KRAS status stratified by age-groups (<50 years, 50–69 years and ≥70 years).

Discussion

This study's major finding is that the prognostic impact of KRAS mutation in metastatic colorectal cancer depends on the age-group of patients. In age-groups <50 and 50–69 years, the presence of KRAS mutation was associated with worse survival, whereas in patients ≥70 years, there was no difference in survival based on the KRAS mutation status. This is a new finding as we did not identify any study that reported the prognostic impact of age and KRAS mutation in this population. In our cohort, 42% of the young patients had KRAS mutation in their tumors, which is higher than previously reported. Liang et al. and Berg et al. reported that 30 and 32.4% of young-onset colorectal cancer patients had KRAS mutations, respectively [24,25]. Because this study is a large cohort, our findings might be closer to the actual KRAS mutation percentage in the younger colorectal cancer patient population.

Our study also demonstrated that KRAS mutation status did not maintain its prognostic value in older age-group patients. The reason for this could be multifactorial, including tumor biology, demographics and treatment characteristics. We believe that the difference may lie in treatment choices in the older age-group compared with the disease biology. Although colorectal cancer is primarily a disease of the elderly, such patients are generally underrepresented in clinical trials. It is difficult to generalize the clinical benefits reported in a relatively younger population from the clinical trials to older age-group patients [26]. In terms of treatment, it is well established that KRAS wild-type status is associated with the response of anti-EGFR therapies [27]. However, there are limited data about the benefits of these targeted agents in older age-group patients. Bouchahda et al. reported the efficacy and safety of cetuximab in 56 elderly patients with heavily pretreated metastatic colorectal cancer [28]. The median age was 76 years (range: 70–84). They reported that efficacy was similar to that demonstrated in younger patients. Similarly, Jehn et al. evaluated cetuximab's efficacy and safety combined with chemotherapy in pretreated metastatic colorectal cancer patients aged >65 versus ≤65 years [29]. They observed a similar efficacy and safety profile between patients aged ≤65 years and their older counterparts. The overall response rate was 37.9% for ages 18–65 and 35.4% for >65 years. Another study examined age, comorbidity and performance status as predictors of outcome in 572 cetuximab-treated patients with advanced colorectal cancer. They reported that age was not associated with OS (p = 0.13) [30]. However, despite the benefit of anti-EGFR in older patients, such targeted therapies are underutilized. Bikov et al. evaluated the patterns of biologics use in elderly medicare patients. They used Surveillance, Epidemiology, and End Results (SEER)-Medicare data of 4545 elderly patients diagnosed with metastatic colon cancer from 2003 to 2009. Bevacizumab was the most frequently used biologic in elderly patients, followed by cetuximab. They showed the adjusted odds of biologics receipt decreased rapidly with age, resulting in a three-fold difference between the youngest and the oldest study participants in the sample (odds ratio: 0.35; p = 0.01) [31]. After adjusting for sociodemographic and clinical differences, age, race, comorbidities and low income had a significant negative effect on the likelihood of receiving biologics among treated patients. Another study with medicare data between 2000 and 2009 observed 4% increased survival per year along with increased bevacizumab utilization in the elderly population (HR: 0.96; 95% CI 0.95–0.97) [32].

This study summarizes data on 26,095 metastatic colorectal cancer patients with known KRAS status and is the largest report on outcomes of KRAS mutation on survival in metastatic colorectal cancer patients. It confirms in a large dataset, previously reported findings that these mutations are more frequent in females, the black population and patients with right-sided tumors. Previous studies, such as the international multicenter RASCAL study (the Kirsten Ras in Colorectal Cancer Collaborative Group), with 2721 patients with metastatic and nonmetastatic disease, detected an increased risk of recurrence and death by 26% [7]. The expanded RASCAL II study with 3439 patients showed that KRAS mutations were associated with a 50% increased risk of recurrence or death in nonmetastatic node-positive patients [8]. Lee et al. evaluated the KRAS status in a cohort of 388 higher-risk stage II and stage III Korean patients who were treated with adjuvant FOLFOX regimen [33]. They observed that KRAS mutation was associated with a worse survival outcome and a higher risk of recurrence. Previously reported data also showed that KRAS-mutated colorectal cancers were more frequently observed in black patients and right-sided locations [8,23]. In a meta-analysis of 20 studies, Staudacher et al. showed that Caucasians have 36% less KRAS mutation than African Americans [34]. Several studies described that black patients have unfavorable survival compared with white patients [35–37]. Tumor biology and socioeconomic differences have a potential role in this disparity [38]. Similarly, several studies have suggested that right-sided tumor location could be a proxy for genetic mutations and associated with a worse prognosis [13,39,40].

Our study has several strengths, including a large cohort and a significant population with reported KRAS mutation status. Age-related prognostic impact of a mutation that is widely used in clinics to determine therapy is relevant because it identifies a high-risk group, where new approaches are needed to improve outcomes. This is especially true for young-onset colorectal cancer patients where we showed that KRAS mutation was associated with significantly worse outcomes. It also highlights the need for utilization of appropriate targeted therapies in other groups, such as elderly patients in whom underutilization of targeted therapies likely explain the lack of prognostic impact. The coverage of NCDB can be different for some cancer types, but this difference decreases with increased facility location [41,42]. We also acknowledge several limitations, particularly with being retrospective, which carries the risk of selection bias. Further, 66.4% of patients with metastatic colorectal cancer did not have a KRAS status reported. We believe that a high rate of nonreporting for KRAS status is associated with patients and facility-related factors including older age, high comorbidity score and treatment at the nonacademic facility. Also, using large databases can cause potential coding errors. NCDB does not have information regarding certain variables that could affect overall survival, including chemotherapy regimen, EGFR-directed antibodies and recurrence rate. Another limitation is that this only reports KRAS but does not report any data about KRAS mutation types, NRAS, and other mutations.

Conclusion

We showed that the prognostic value of KRAS mutation in metastatic colorectal cancer depends on the age-group of the patients. Younger KRAS mutated colorectal cancer patients had a worse prognosis than wild-type KRAS in the tumor. However, KRAS mutation status did not maintain its prognostic value among older age-group patients. We believe that the difference seen in this population-based database could be due to older patients being more vulnerable and receiving less intense chemotherapy regimens. Also, the underutilization of targeted therapies in the older population also could be an important factor. However there may be other factors that may explain the current findings. Further studies are needed on characteristics of young-onset colorectal cancer.

Summary points.

• The objective of this study was to determine the prognostic impact of KRAS mutation status among different age-groups in metastatic colorectal cancer patients.

• We queried the National Cancer Database, including adult patients (≥18 years) with stage IV colorectal cancer diagnosed from 2010 to 2016 with known KRAS mutation status.

• We compared characteristics and overall survival in age groups (<50, 50–69 and ≥70 years) by KRAS mutation status. The multivariable analysis assessed the interaction between KRAS status and age-groups using Cox regression analysis.

• Patients with KRAS mutation had worse OS than patients with wild-type KRAS (median OS 19.9 vs 22.2 months; p < 0.001). In multivariable analysis, both age and KRAS mutation status were associated with worse overall survival.

• The interaction term between KRAS status and age-groups was significant (p < 0.001), which showed that the prognosis of KRAS status is impacted by the age-group.

• KRAS mutation was prognostic for survival among <50 years old (23.5 vs 28.8 months; p < 0.001) and in 50–69 years (21 vs 23.4 months; p < 0.001) but lost its prognostic impact in ≥70-year-old patients (14.3 vs 13.9 months; p = 0.34).

• We conclude that the age of patients impacts the prognostic value of KRAS mutation in metastatic colorectal cancer.