Racial Differences in BRAF/KRAS Mutation Rates and Survival in Stage III Colon Cancer Patients

Abstract

Background:

It is unknown if, after controlling for clinicopathologic variables and treatment, racial disparities in colon cancer outcomes persist. Molecular marker analysis in North American patients comparing Asians with other races has not been reported.

Methods:

BRAF ^V600E and KRAS mutations were analyzed in node-positive colon cancer patients (n = 3305) treated with FOLFOX-based chemotherapy in an adjuvant trial (Alliance N0147). Race categories included Asian, black, or white. Cox models were used to estimate disease-free survival (DFS) and time to recurrence (TTR). All statistical tests were two-sided.

Results:

BRAF mutation frequency in tumors from whites (13.9%) was twice that of tumors from Asians or blacks. KRAS mutation rates were highest in tumors from blacks (44.1%). KRAS/BRAF wild-type tumors were most common among Asians (66.7%) (P _overall < .001). The prognostic impact of race differed by age and N stage (both P _interaction < .02). Compared with whites, blacks had shorter DFS among patients younger than age 50 years (hazard ratio [HR] = 2.84, 95% confidence interval [CI] = 1.73 to 4.66) or with N1 disease (HR = 1.54, 95% CI = 1.04 to 2.29), independent of BRAF, KRAS, and other covariates. Findings were consistent using TTR as the outcome. Asians had longer DFS among N2 tumors that was partly mediated by less frequent BRAF mutation.

Conclusions:

Colon cancers from Asians have a lower rate of BRAF and KRAS mutations than blacks or whites. We report a novel interaction of race with age and N stage in node-positive disease, indicating that racial disparities in survival persist despite uniform stage and treatment in a phase III trial.

Colorectal cancer (CRC) incidence varies globally and has increased since the 1980s in economically transitioning countries, including within Asia (1,2). In the United States, CRC incidence rates have risen among individuals younger than age 50 years (3), and outcome disparities between whites and blacks have worsened among patients with node-positive tumors (4). Factors potentially underlying worse outcomes in blacks include more advanced disease at diagnosis, decreased access to care, suboptimal treatment, lower socioeconomic status, and CRC biology (5). However, the impact of race in patients who received adjuvant fluorouracil (5FU) plus oxaliplatin, the standard in node-positive colon cancer, has not been extensively examined (Supplementary Table 1, available online) (6,7).

CRC is driven by (epi)genetic modifications that regulate proliferation, apoptosis, and angiogenesis, including in the epidermal growth factor receptor (EGFR) pathway (8,9). BRAF and KRAS mutations downstream of EGFR lead to its constitutive activation (10,11). BRAF ^V600E mutations forecast adverse prognosis (12–16); KRAS mutations predict resistance to EGFR inhibitors in advanced disease. BRAF and KRAS mutations are mutually exclusive. Sparse data are available on their frequency by race; limitations of published studies include small subgroup sizes, analysis of only one marker, or racial categorization restricted to whites vs blacks (17–23).

The advantages of assessing racial disparities in a randomized trial are well-described (6,7). N0147 was a phase 3 trial that found the addition of cetuximab to adjuvant FOLFOX did not improve disease-free survival (DFS) in stage III colon cancer (24). In prior analyses of N0147, we noted that BRAF mutation rates differed in white vs nonwhite patients (18) and in patients with nonmutated KRAS by race (25). Here, we performed a detailed analysis of race, categorized as black, Asian, or white, in relation to BRAF and KRAS mutation status in the full cohort and, for the first time, with regard to patient survival.

Methods

Study Population

We analyzed data from patients with resected colon adenocarcinoma from N0147, most of whom received 12 cycles of mFOLFOX6, alone or combined with cetuximab (Figure 1) (24,26). Prospective, centralized KRAS mutation testing was required, although random assignment was done irrespective of KRAS status in the original design (Supplementary Methods, available online).

Figure 1.

Study profile of the N0147 (North Central Cancer Treatment Group/Alliance) clinical trial.

The study was approved by the institutional review board (IRB) of Mayo Clinic and the North Central Cancer Treatment Group (now Alliance). Patients signed an IRB-approved consent.

Race

Self-identified race was ascertained via a questionnaire prior to random assignment and classified as white, Asian, black or African American, Native Hawaiian or other Pacific Islander, American Indian or Alaska Native, unknown (patient unsure), or not reported (patient refused or data not available). Ethnicity was separately classified as Hispanic or Latino, or non-Hispanic. Hispanic ethnicity was infrequent among Asian (0/149) or black (6/240) patients, so Hispanic and non-Hispanic ethnicities were combined in our analysis. For whites, analysis combining Hispanic (126/2916) and non-Hispanic (2320/2916) ethnicities had similar results as analyzing them separately, so combined data are shown.

KRAS, BRAF, and DNA Mismatch Repair

KRAS and BRAF ^V600E mutation status and mismatch repair (MMR) protein expression were assessed (Supplementary Methods, available online) (24,26,27). Tumors were classified as MMR deficient (vs MMR proficient) if loss of one or more MMR proteins was detected.

Statistical Analysis

DFS was defined as the time from random assignment to first documented recurrence or any-cause death, whichever occurred first; time to recurrence (TTR), as the time from random assignment to first documented recurrence. Multivariable Cox models were adjusted for age, T and N stage, grade, tumor site, MMR, BRAF/KRAS mutation, cetuximab treatment, body mass index, and smoking, unless noted. The proportional hazard assumption was confirmed by examination of Schoenfeld residuals plot. A correction for multiple comparisons was performed (Supplemental Methods, available online). The age cutpoint of 50 years was prespecified. Continuous age was modeled using restricted cubic splines in Cox models to evaluate possible nonlinearity of age effect on relative hazard of outcomes. Interactions of race with variables were determined. Analyses (SAS v.9.2) utilized data frozen on September 6, 2013. Two-sided P values (<.05 considered statistically significant) are reported.

We utilized a statistical method (28) to determine whether BRAF or KRAS mutation status mediates differences in DFS between races. Mediation effects were analyzed in subgroups where race demonstrated a statistically significant association with outcome. For each analysis, three models were generated. Because this analysis utilizes only binary covariates, effects of BRAF mutation (yes vs no) and KRAS mutation (yes vs no) status were calculated separately. An example using BRAF mutation follows:

Model 1 (Cox): DFS = i₁ + c’Race + b*BRAF + e₁

Model 2 (logistic): BRAF = i₂ + a*Race + e₂

Mediation effect of BRAF = $\widehat{a}\cdot \widehat{b}$

In this example, BRAF was considered to mediate the association between race and DFS if the 95% confidence interval (CI) of the mediation effect value excluded zero. While the size or direction of this value (ie, $\widehat{a}\cdot \widehat{b}$) does not have specific meaning, this method has advantages over other approaches (29,30) when effect sizes are increased after covariate adjustment, as occurred in our analysis.

All statistical tests were two-sided.

Results

Race and Clinicopathologic Characteristics

Of the 3397 patients registered onto the trial, 3305 identified themselves as white, black, or Asian and form the population in which race was analyzed in relation to mutation status and clinicopathologic characteristics (Figure 1).

As shown in Table 1, blacks were more likely to be younger than age 50 years compared with whites, and tumors from blacks showed less aggressive pathologic features, including a higher frequency of T_1-2 (vs T_3-4) stage, N1 (vs N2) stage, and low- (vs high-) grade histology. The frequency of MMR-deficient tumors did not differ in blacks vs whites (10.1% vs 12.1%, respectively, P = .38).

Table 1.

Association of race with clinicopathologic characteristics in stage III colon cancer patients treated in a phase III trial (n = 3305)

Variable	TOTAL (n = 3305)	White (n = 2916)	Black (n = 240)		Asian (n = 149)
	No. (%)	No. (%)	No. (%)	P (vs white)*	No. (%)	P (vs white)*	P (vs black)*
Age, y
<50	738 (22.3)	630 (21.6)	75 (31.3)	<.001	33 (22.1)	.88	.05
≥50	2567 (77.7)	2286 (78.4)	165 (68.8)		116 (77.9)
Sex
Female	1557 (47.1)	1357 (46.5)	124 (51.7)	.13	76 (51.0)	.29	.90
Male	1748 (52.9)	1559 (53.5)	116 (48.3)		73 (49.0)
Body mass index
Underweight	139 (4.2)	102 (3.5)	14 (5.9)	.13	23 (15.4)	<.001	<.001
Normal	846 (25.7)	714 (24.6)	59 (24.7)		73 (49.0)
Overweight	1175 (35.7)	1057 (36.4)	74 (31.0)		44 (29.5)
Obese	1131 (34.4)	1030 (35.5)	92 (38.5)		9 (6.0)
Missing	14	13	1		0
Smoking
Never	1059 (46.9)	925 (46.1)	82 (49.1)	.08	52 (59.1)	.01	.01
Former	1032 (45.7)	931 (46.4)	66 (39.5)		35 (39.8)
Current	169 (7.5)	149 (7.4)	19 (11.4)		1 (1.1)
Missing	1045	911	73		61
T stage
T1 or T2	496 (15.0)	432 (14.8)	48 (20.1)	.03	16 (10.7)	.17	.02
T3 or T4	2807 (85.0)	2483 (85.2)	191 (79.9)		133 (89.3)
Missing	2	1	1		0
Positive nodes, No.
1–3 (N1 stage)	1949 (59.0)	1706 (58.5)	164 (68.3)	.003	79 (53.0)	.19	.002
4 or more (N2 stage)	1356 (41.0)	1210 (41.5)	76 (31.7)		70 (47.0)
Nodes examined, No.
Mean (SD)	19.9 (11.3)	20.0 (11.5)	19.0 (9.2)	.84†	20.1 (10.4)	.65†	.68†
Grade
High	824 (24.9)	753 (25.8)	37 (15.4)	<.001	34 (22.8)	.41	.07
Low	2481 (75.1)	2163 (74.2)	203 (84.6)		115 (77.2)
Tumor location
Right	1679 (51.6)	1500 (52.1)	132 (56.4)	.21	47 (32.2)	<.001	<.001
Left	1578 (48.4)	1377 (47.9)	102 (43.6)		99 (67.8)
Missing	48	39	6		3
Mismatch repair status
Proficient	2811 (88.3)	2468 (87.9)	204 (89.9)	.38	139 (93.9)	.03	.17
Deficient	372 (11.7)	340 (12.1)	23 (10.1)		9 (6.1)
Missing	122	108	13		1

* Two-sided Chi-square test, unless noted.

† Two-sided Wilcoxon rank sum test.

With whites or blacks as reference, Asian patients were less likely to be overweight or obese, or to be current or past smokers. Tumors from Asians showed a predilection for the distal (vs proximal) colon and were least likely among races to be MMR deficient (6.1% in Asians vs 12.1% in whites, P = .03) (Table 1). Compared with blacks, Asians showed a higher frequency of T_3-4 (vs T_1-2) stage (P = .02) and N2 (vs N1) stage (P = .002).

A difference between races was not observed with regard to the number of lymph nodes examined (median 17 to 18) or chemotherapy cycles administered (median 12 for each race; data not shown).

Race and BRAF/KRAS

As shown in Figure 2, BRAF ^V600E mutation frequency in tumors from whites (13.9%) was twice that of tumors from Asians (5.6%, P < .001) or blacks (6.4%, P = .009). KRAS mutation rates were highest in tumors from blacks (44.1%), which differed compared with tumors from Asians (27.8%, P = .001) and trended toward a difference compared with tumors from whites (34.9%, P = .07). Tumors that were wild-type for both BRAF and KRAS were most common among Asians (66.7%), and the frequency differed compared with tumors from blacks (49.5%, P = .001) or whites (51.2%, P < .001) (P _overall < .001). While the MMR-deficient subgroup was too small for meaningful comparisons, we found a similar racial distribution of BRAF/KRAS mutation frequency among MMR-proficient tumors as in the overall cohort (Supplementary Table 2, available online).

Figure 2.

Association of race with BRAF and KRAS mutation status in stage III colon cancer patients. * Reference is BRAF or KRAS mutated. † Reference is wild-type for both BRAF and KRAS. Chi-square or Fisher’s exact test was used. All statistical tests were two-sided.

By multivariable analysis, the frequency of BRAF ^V600E mutation (vs BRAF/KRAS wild-type) remained associated with tumors from whites vs blacks independent of MMR, grade, tumor site, and T and N stage (OR = 2.11, 95% CI = 1.14 to 3.89, P = .02) (data not shown). By contrast, KRAS mutation frequency did not differ in whites vs blacks (OR = 0.87, 95% CI = 0.64 to 1.17, P = .35). A higher frequency of BRAF/KRAS wild-type vs KRAS mutation was found in tumors from Asians vs blacks (OR = 1.70, 95% CI = 1.04 to 2.79, P = .04) and in Asians vs whites (OR = 1.41, 95% CI = 0.95 to 2.09, P = .09). BRAF/KRAS wild-type vs BRAF ^V600E mutation was also associated with Asian vs white race (OR = 2.14, 95% CI = 0.98 to 4.69, P = .06).

Race and Survival

Outcomes were analyzed in 2931 patients randomly assigned to FOLFOX +/- cetuximab (Figure 1). Given a lack of interaction between race and treatment for DFS (P _interaction = .17), study arms were pooled. We examined interactions of race with selected covariates for DFS. Statistically significant interactions were found for age (P _interaction = .01) and N stage (P _interaction = .005), but not for T stage_, grade, tumor site, MMR, or BRAF/KRAS. We analyzed outcomes within age and N stage strata.

Among patients younger than age 50 years, blacks had shorter DFS (HR = 1.80, 95% CI = 1.21 to 2.66, P = .004) and TTR (HR = 1.77, 95% CI = 1.18 to 2.65, P = .006) compared with whites (Figure 3). The racial disparity among younger patients increased in multivariable analysis after adjustment for covariates, including BRAF and KRAS mutation status (HR for DFS = 2.84, 95% CI = 1.73 to 4.66, P < .001; HR for TTR = 2.71, 95% CI = 1.61 to 4.54, P < .001) (Table 2). By contrast, among patients age 50 years and older, no difference in DFS or TTR between blacks and whites was observed. Likewise, blacks younger than age 50 years (vs ≥50) had shorter DFS and TTR (Table 2). Differential outcomes between Asians and whites were not observed in either age group, although among patients younger than age 50 years, blacks demonstrated shorter DFS compared with Asians (Supplementary Table 3, available online).

Figure 3.

Survival by race and age in stage III colon cancer patients treated with FOLFOX-based adjuvant chemotherapy in a phase 3 trial. Disease-free survival (DFS) and/or time to recurrence (TTR) are shown by race among patients younger than age 50 years (A, B) and age 50 years and older (C). DFS is shown by age among white (D), black (E), and Asian (F) patients. Univariate hazard ratios and 95% confidence intervals (likelihood ratio test) are shown. Interaction of race with age (cutpoint at 50 years) was statistically significant (P = .01 for DFS, P = .03 for TTR). All statistical tests were two-sided. DFS = disease-free survival; TTR = time to recurrence.

Table 2.

Multivariable Cox proportional hazards models of survival in stage III colon cancer patients by race, age, and N stage

Variables	Disease-free survival			Time to recurrence
	5-year EFS, %	HR (95% CI)	P*	5-year EFS, %	HR (95% CI)	P*
Race and age†
Age < 50 y
White	70	(reference)		71	(reference)
Black	50	2.84 (1.73 to 4.66)	<.001‡	53	2.71 (1.61 to 4.54)	<.001‡
Asian	78	0.97 (0.38 to 2.47)	NS	78	1.02 (0.40 to 2.61)	NS
Age ≥ 50 y
White	65	(reference)		69	(reference)
Black	68	0.82 (0.55 to 1.21)	NS	72	0.90 (0.60 to 1.35)	NS
Asian	73	0.58 (0.31 to 1.06)	NS	74	0.67 (0.36 to 1.23)	NS
White
< 50 y	70	0.82 (0.65 to 1.04)	NS	71	0.91 (0.71 to 1.16)	NS
≥ 50 y	65	(reference)		69	(reference)
Black
< 50 y	50	4.07 (2.16 to 7.66)	<.001‡	53	3.69 (1.92 to 7.08)	<.001‡
≥ 50 y	68	(reference)		72	(reference)
Asian
< 50 y	78	1.77 (0.37 to 8.48)	NS	78	1.77 (0.37 to 8.48)	NS
≥ 50 y	73	(reference)		74	(reference)
Race and N stage§
N2
White	52	(reference)		55	(reference)
Black	56	0.95 (0.60 to 1.51)	NS	60	0.98 (0.61 to 1.58)	NS
Asian	73	0.50 (0.25 to 0.98)	.04	73	0.53 (0.27 to 1.05)	NS
N1
White	76	(reference)		80	(reference)
Black	66	1.54 (1.04 to 2.29)	.03	69	1.71 (1.13 to 2.58)	.01
Asian	75	0.94 (0.44 to 2.03)	NS	76	1.18 (0.54 to 2.57)	NS
White
N2	52	2.17 (1.81 to 2.59)	<.001‡	55	2.30 (1.90 to 2.79)	<.001‡
N1	76	(reference)		80	(reference)
Black
N2	56	1.45 (0.74 to 2.83)	NS	60	1.45 (0.74 to 2.84)	NS
N1	66	(reference)		69	(reference)
Asian
N2	73	0.38 (0.08 to 1.91)	NS	73	0.38 (0.08 to 1.91)	NS
N1	75	(reference)		76	(reference)

* Two-sided Likelihood ratio test. CI = confidence interval; EFS = event-free survival; HR = hazard ratio; NS = not statistically significant (P > .05).

† Adjusted for T and N stage, tumor grade, tumor site, mismatch repair status, BRAF and KRAS mutation status, cetuximab treatment, body mass index, and smoking.

‡ P value met correction for multiple comparisons at a statistical significance level of .004 or less (see Methods; also, Supplemental Methods, available online).

§ Adjusted for age, T stage, tumor grade, tumor site, mismatch repair status, BRAF and KRAS mutation status, cetuximab treatment, body mass index, and smoking.

Age, as a continuous variable, demonstrated a univariate association with DFS (P _overall < .001) that was nonlinear (P = .005), where the youngest and oldest patients showed modestly worse outcomes than patients of middle age (Figure 4A). The association between age and TTR (P _overall = .15) showed a similar trend (P = .07 for nonlinearity) (Figure 4B). An interaction between race and continuous age was found (P = .002 for DFS, P = .008 for TTR). The risk for a DFS or TTR event among blacks increased as age decreased further below approximately 55 years (Figure 4, C and D). By contrast, outcomes for whites or Asians were relatively monotonic across ages. In multivariable analysis that included continuous age and all other covariates, black (vs white) race was associated with shorter DFS (HR = 2.80, 95% CI = 1.71 to 4.61, P < .001) and TTR (HR = 2.67, 95% CI = 1.59 to 4.48, P < .001) among patients younger than age 50 years; but was nonprognostic among patients age 50 years and older (data not shown). In the same models, age as a continuous variable was not associated with DFS or TTR among patients younger than age 50 years (data not shown). However, among patients age 50 years and older, each year of increased age was associated with shorter DFS (HR = 1.01, 95% CI = 1.00 to 1.03, P = .04) but not with differential TTR (data not shown).

Figure 4.

Age as a continuous variable in relation to patient survival with and without regard to race. Risk of (A) disease-free survival (DFS) or (B) time to recurrence (TTR) event as nonlinear function of age, with highest risk at young and old extremes; risk of (C) DFS or (D) TTR event as function of age according to race. Interaction of race with age (continuous variable) was statistically significant (P = .002 for DFS, P = .008 for TTR). All statistical tests were two-sided. DFS = disease free survival; TTR = time to recurrence.

Among patients with N1 tumors, blacks had shorter DFS (HR = 1.50, 95% CI = 1.10 to 2.06, P = .01) and TTR (HR = 1.59, 95% CI = 1.14 to 2.23, P = .007) compared with whites (Figure 5), and this disparity persisted in multivariable analysis (DFS in blacks vs whites in N1 disease: HR = 1.54, 95% CI = 1.04 to 2.29) (Table 2). By contrast, a difference in DFS or TTR between blacks and whites with N2 tumors was not observed. Asians showed improved DFS (HR = 0.49, 95% CI = 0.30 to 0.79, P = .004) and TTR (HR = 0.53, 95% CI = 0.33 to 0.86, P = .01) compared with whites among patients with N2 tumors, but not among those with N1 tumors. In multivariable analysis the difference in outcomes between Asians and whites with N2 disease remained statistically significant for DFS but was not statistically significant for TTR (Table 2; Supplementary Table 3, available online).

Figure 5.

Survival by race and N stage in stage III colon cancer patients treated with FOLFOX-based adjuvant chemotherapy in a phase 3 trial. Disease-free survival (DFS) and/or time to recurrence (TTR) are shown by race among patients with N1 (A, B) and N2 (C) stage. DFS is shown by N stage among white (D), black (E), and Asian (F) patients. Univariate hazard ratios and 95% confidence intervals (likelihood ratio test) are shown. Interaction of race with N stage was statistically significant (P = .005 for DFS, P = .005 for TTR). All statistical tests were two-sided. DFS = disease-free survival; TTR = time to recurrence.

The association of blacks (vs whites) with shorter DFS or TTR among younger patients remained statistically significant after adjustment for multiple comparisons, whereas the other outcome differences between races did not (Table 2; Supplementary Table 3, available online).

We utilized a statistical model to determine whether differential BRAF ^V600E or KRAS mutation frequency may partly explain the observed racial differences in DFS (Supplementary Table 4, available online). Among patients younger than age 50 years, neither BRAF nor KRAS mediated the shorter DFS of blacks vs whites. Among patients with N2 tumors, a mediation effect for KRAS was not observed, yet the lower BRAF ^V600E mutation frequency in tumors from Asians vs whites was found to mediate, at least partially, the longer DFS in Asians compared with whites (mediation effect = -0.179, 95% CI = -0.434 to -0.004).

Discussion

We examined the association of race with molecular and clinicopathologic characteristics and survival in a large cohort of patients with stage III colon cancer treated with adjuvant FOLFOX chemotherapy in a phase 3 trial. Tumors from whites showed more than twice the frequency of BRAF ^V600E mutation than tumors from Asians or blacks. KRAS mutation rates were highest in tumors from blacks, and tumors that were wild-type for both BRAF and KRAS were most common among Asians. As previously reported, patients with BRAF ^V600E or KRAS mutation had worse DFS rates in the overall cohort from the N0147 trial (13), and a recent analysis from the N0147 trial found that patients with BRAF/KRAS wild-type tumors had similar DFS rates as those with MMR-deficient tumors (31). Though our MMR-deficient subgroup was too small for analysis by race, race-based differences in BRAF and KRAS frequency were similar in MMR-proficient tumors. To our knowledge, these are the first data in colon cancers to show statistically significant differences in BRAF or KRAS mutation frequency across races. In the largest study to date of CRCs from black patients, which included stage I-IV and rectal tumors, the BRAF ^V600E mutation rate was 4% among 566 black patients, consistent with our findings, but was only 7% among the 328 white patients examined (21). In a population-based study of more than 1200 colon cancers from white patients, the BRAF ^V600E mutation rate was 10.3%, which is similar to our results (23).

The possibility that CRC biology differs across races has been suggested by data from population-based studies. Blacks typically present with CRC at an earlier age than whites, and the proportion of blacks diagnosed at younger than age 50 years is higher than that for whites (5,32). Although CRC mortality differences between blacks and whites exist among all age cohorts, they are most pronounced for the youngest cohorts (5). Blacks have a higher incidence of proximal cancers than do whites, a finding that may not be fully explained by differences in the frequency or type of CRC screening (5,33,34). Racial differences among CRC patients have also been described in relationship to genotypes of drug-metabolizing enzymes, tumor response rates, and drug toxicity rates (6,35,36).

Recent studies identify a “serrated” neoplasia pathway characterized by BRAF mutations and a CpG island methylator phenotype that can lead to sporadic CRCs with frequent high-grade histology (37). We found a lower rate of both BRAF mutations and high-grade histology in tumors from blacks vs whites, suggesting the serrated pathway may be less common among blacks (38,39), although further study is needed to address this issue. In addition, smoking has been associated with a subset of sporadic CRCs that are frequently BRAF mutated and MMR deficient because of MLH1 inactivation by methylation (23,40). Interestingly, we found that Asians exhibited substantially lower rates of smoking and obesity compared with whites or blacks, and their tumors tended to be distal, MMR proficient, and BRAF/KRAS wild-type, suggesting an origin from typical adenomatous polyps.

We found a statistically significant interaction of race with age on survival, indicating the prognostic impact of race differed by age. Among patients younger than age 50 years, blacks exhibited a more-than-2.5-fold worsening of DFS and TTR compared with whites independent of BRAF and KRAS status and other prognostic factors. Almost 50% of younger black patients experienced colon cancer recurrence within five years, compared with approximately 30% of black patients older than age 50 years or white or Asian patients regardless of age. The associations were strengthened after adjustment for multiple potential confounders. To our knowledge, a shortened TTR within young black patients with stage III colon cancer has not been previously reported. The racial differences in survival were not evident in patients age 50 years and older. The potential that a race-age interaction exists is supported by two tumor registry studies that found an interaction between black (vs white) race and age in CRC patients of varying stages, with a higher risk for all-cause death observed among younger blacks (41,42). Because the purpose of adjuvant therapy is to extend patient survival by delaying or preventing colon cancer recurrence, TTR is the most sensitive measure of the intended effect of chemotherapy. Unlike DFS, this endpoint would not be biased as a result of disparity in treatment for recurrent colon cancer, disparate care for comorbid conditions, or differential rates of death of causes unrelated to colon cancer. The consistency of results for both TTR and DFS suggest that, despite equivalent treatment with modern chemotherapy in a clinical trial, tumors that occur in young black patients have an intrinsically more aggressive biology or potentially could receive less benefit from adjuvant chemotherapy. While we did not find that BRAF/KRAS status mediated the shorter DFS observed in blacks vs whites, a limitation of this approach is that it allows analysis of only one mutation at a time in a binary fashion (eg, BRAF-mutated patients can only be compared with BRAF wild-type, with the latter including KRAS-mutated patients). However, where feasible, mutation status may be most appropriately analyzed categorizing BRAF/KRAS mutation as a three-level variable (eg, comparing BRAF-mutated vs wild-type for both BRAF/KRAS), as we otherwise did. Therefore, it is possible that our analysis lacked sensitivity to detect a mediation effect.

We found a statistically significant interaction of race with N stage on survival, indicating the prognostic impact of race differed based on N stage. Among patients with N1 disease, blacks (vs whites) demonstrated shorter DFS and TTR, whereas a difference was not observed in patients with N2 disease. Population-based studies suggest that racial disparities in survival were greatest in stage II or III disease (vs stage IV) in earlier years (pre-1994) (4,43–45). Recent SEER data analysis found that, by 2006 to 2008, the survival disparity between blacks vs whites was primarily driven by their higher rates of presentation with advanced disease, though differences between races in rates of regional disease continued to have a contributing role (4). We found that, among N2 tumors, Asians showed statistically significantly longer DFS and a trend toward longer TTR, compared with whites, that was at least partly mediated by a lower BRAF ^V600E mutation frequency in tumors from Asians (vs whites). These findings are consistent with national surveillance data indicating a lower CRC incidence and mortality rate among Asians in the United States compared with blacks or whites (32,46,47). In a prior study from this cohort, we found that BRAF ^V600E mutations were associated with poor tumor differentiation, increased N2 (vs N1) stage, and adverse outcome (25). Furthermore, BRAF ^V600E mutations have been associated with increased rates of peritoneal metastases (48) and adverse outcome after colon cancer recurrence (16,49). BRAF ^V600E knockdown in xenografts not only stalled tumor growth but also induced glandular structures with marked expression of CDX2, a tumor suppressor and master transcription factor of intestinal differentiation (50).

A major strength of our study was utilizing data from a randomized controlled trial. Advantages include protocol-specified assessment of eligibility, treatment and supportive measures, and follow-up, thereby controlling for factors that may influence care after diagnosis and enabling reliable measurement of disease-specific outcome. The adequacy of surgical resections and pathology examination is reflected in the high number of lymph nodes examined that did not differ by race. Molecular markers were analyzed prospectively using uniform methodology in a single Clinical Laboratory Improvement Amendments–certified laboratory.

Our study has limitations. Although the percentages of Asian and black patients are comparable with their frequency reported elsewhere (6,7,32), we acknowledge their smaller sample size compared with whites. The proportion of young-onset cases was similar to that of other adjuvant colon cancer trials (7) yet was higher than reported in national statistics (3,51), suggesting our results may not generalize to non–clinical trial populations. To date, we report the largest study of CRC tumors from Asian or black patients whose molecular features and survival have been analyzed in a trial. Currently, a suitable validation cohort is mostly lacking, although the potential exists to pool data from ongoing and completed adjuvant trials. Although the observation that blacks (vs whites) demonstrated shorter DFS and TTR among younger patients remained statistically significant after adjustment for multiple comparisons, we cannot rule out the possibility of a type 1 error. Self-reported race was utilized, which lacks grandparental racial origins or ancestry-informative genetic evaluation, though studies suggest that self-defined race may be as informative as genetic ancestry analysis (52–55).

Funding

This work was supported by a National Cancer Institute Senior Scientist Award (K05CA-142885 to FAS) and the North Central Cancer Treatment Group (NCCTG) Biospecimen Resource Grant (CA-114740) from the National Institutes of Health. Support for correlative studies was also provided by unrestricted funds from Bristol-Myers Squibb, ImClone Systems, Sanofi-Aventis, and Pfizer. The study was conducted as a collaborative trial of the North Central Cancer Treatment Group, Mayo Clinic, and was supported in part by Public Health Service grants CA-25224 and CA-37404 from the National Cancer Institute, Department of Health and Human Services. The study was also supported, in part, by grants from the National Cancer Institute to the Alliance for Clinical Trials in Oncology (U10CA180821) and to the Alliance Statistics and Data Center (U10CA180882).

H. H. Yoon, Q. Shi, S. R. Alberts, R. M. Goldberg, D. J. Sargent, and F. A. Sinicrope were responsible for study concept and design. H. H. Yoon, Q. Shi, D. J. Sargent, and F. A. Sinicrope were responsible for analysis and interpretation of data. H. H. Yoon, Q. Shi, and F. A. Sinicrope were responsible for the drafting of the manuscript. H. H. Yoon, Q. Shi, S. R. Alberts, R. M. Goldberg, S. N. Thibodeau, D. J. Sargent, and F. A. Sinicrope were responsible for critical revision of the manuscript for important intellectual content. Q. Shi and D. J. Sargent were responsible for statistical analysis.

The authors declare no conflicts of interest. The content is solely the responsibility of the authors and does not necessarily represent the views of the National Cancer Institute or the National Institutes of Health. The study funders had no role in the design of the study, the collection, analysis, or interpretation of the data, the writing of the manuscript, nor the decision to submit the manuscript for publication.