Epidermal growth factor receptor intron 1 CA dinucleotide repeat polymorphism and survival of advanced gastric cancer patients treated with cetuximab plus modified FOLFOX6

Abstract

(Cancer Sci 2010; 101: 793–799)

Cetuximab is a monoclonal antibody targeting epidermal growth factor receptor (EGFR). The present study investigated the association between germline genetic polymorphisms and the treatment outcome of cetuximab plus modified leucovovin, fluorouracil, and oxaliplatin (FOLFOX)6 chemotherapy in advanced gastric cancer (AGC). DNA from peripheral blood mononuclear cells of 38 patients enrolled in a phase II study of cetuximab plus modified FOLFOX6 were analyzed for 16 polymorphisms in eight genes (EGFR, epidermal growth factor, transforming growth factor‐α (TGFA), thymidylate synthase, excision repair cross‐complementation group 1, Xeroderma pigmentosum group D, and fragment c gamma receptors (FCGR)2A and 3A). The EGFR intron 1 CA repeat polymorphism was associated with survival. Twenty‐one patients had low repeats (sum of both alleles ≤37), and 17 patients had high repeats (sum ≥38). Patients with low CA repeats had longer progression‐free survival (adjusted hazard ratio [HR] 0.42 [95% confidence interval [CI] 0.19–0.96], P = 0.040) and overall survival (adjusted HR 0.40 [95% CI 0.16–0.99], P = 0.048) compared with patients with high CA repeats. In addition, the tumor EGFR expression was higher in patients with a lower number of CA repeats. The association between the CA repeat status and survival was not found in a separate cohort of AGC patients (n = 68) treated only with modified FOLFOX6. These results suggest that the EGFR intron 1 CA repeat polymorphism could be a useful, predictive biomarker of cetuximab efficacy in AGC and merits further investigation in randomized studies.

Gastric cancer is frequently associated with poor survival because it often presents as unresectable disease, and chemotherapy shows limited efficacy.^{( 1 )} Therefore, gastric cancer is a major health concern in many countries, including Korea, which has a particularly high incidence.^{( 1} , ^{2 )} In order to improve the treatment outcome of chemotherapy in advanced gastric cancer (AGC), targeted agents are being actively investigated.^{( 3 )} Recently, trastuzumab, a monoclonal antibody targeting human epidormal growth factor receptor 2 (HER2), in addition to fluoropyrimidine and cisplatin, significantly improved the overall survival in HER2‐positive gastric cancer in a phase III study.^{( 4 )}

Cetuximab (Erbitux; Merck KGaA, Darmstadt, Germany) is a monoclonal antibody that binds to and inactivates epidermal growth factor receptor (EGFR).^{( 5 )} Cetuximab improved the treatment outcome of metastatic colorectal cancer patients.^{( 6} , ^{7 )} Interestingly, the benefit of cetuximab was limited to K‐ras wild‐type colorectal cancers.^{( 7} , ^{8 )} These and other similar findings led to the recommendation that metastatic colorectal patients with K‐ras mutant tumors should not receive anti‐EGFR therapy.^{( 9 )}

Cetuximab plus chemotherapy has also shown favorable results as a first‐line treatment of advanced gastric or gastroesophageal junction adenocarcinoma in phase II studies.^{( 10} , ^{11 )} Based on these results, a phase III study to evaluate the benefit of cetuximab in addition to capecitabine and cisplatin in advanced esophagogastric cancer is currently underway.^{( 12 )} In contrast to colorectal cancer, K‐ras mutation is infrequently found in gastric cancer.^{( 13 )} Therefore, other predictive biomarkers should be investigated to aid patient selection for cetuximab in gastric cancer.

We have also conducted a phase II study of cetuximab in AGC.^{( 14 )} Although cetuximab in combination with modified leucovovin fluorouracil and oxaliplatin (FOLFOX)6 failed to meet the prespecified improvement in the response rate, patients with a tumor EGFR expression and low serum ligand levels showed favorable outcomes in the exploratory biomarker analysis.^{( 14 )} In the present study, we investigated candidate genetic polymorphisms and their association with the treatment outcome.

Materials and Methods

Patients and treatment.  Patients who were enrolled in the Korean Cancer Study Group prospective multicenter phase II study of cetuximab in combination with modified FOLFOX6 were included in the present analysis. The main inclusion criteria of the study were age ≥18 years; Eastern Cooperative Oncology Group (ECOG) performance status (PS) ≤2; histologically‐confirmed adenocarcinoma of the stomach; recurrent or metastatic disease, no prior chemotherapy, radiotherapy, immunotherapy, or EGFR pathway‐targeting therapy; adequate bone marrow, hepatic, and renal function; and at least one measurable lesion. Patients received an initial dose of 400 mg/m² cetuximab, followed by weekly doses of 250 mg/m². Modified FOLFOX6 was comprised of 100 mg/m² oxaliplatin and 100 mg/m² leucovorin administered intravenously over 2 h on day 1, followed by a 46‐h infusion of 2400 mg/m² 5‐fluorouracil (5‐FU), which was repeated every 2 weeks. Patients received a maximum of 12 cycles of modified (m) FOLFOX6. Cetuximab was continued as a monotherapy until disease progression. A response evaluation was performed following the RECIST criteria.^{( 15 )} Detailed results of the efficacy and toxicity have been reported previously.^{( 14 )} Among the 40 patients enrolled in the phase II study, the study included 38 patients, excluding two patients whose responses were not evaluable (Table 1). In addition, two patients with unconfirmed partial response (PR) were considered to be responders in the present study. Survival data were last updated in May 2009. Another group of AGC patients in a phase II study of modified FOLFOX6 were also analyzed for the EGFR polymorphism.^{( 16 )} Among the 73 patients enrolled in the original study, 68 patients were included in the present study because five patients had no remaining DNA sample (Table S1). All patients, including the patients who were enrolled in the mFOLFOX6 study,^{( 16 )} gave written, informed consent prior to study entry for the clinical study and biomarker analysis. The study protocol was reviewed and approved by the Institutional Review Boards at the participating institutions. Recommendations of the Declaration of Helsinki for biomedical research involving human participants were also followed.

Table 1.

 Baseline characteristics

Characteristic	No. patients (n = 38)	%
Sex
Male	28	73.7
Female	10	26.3
Age, years
Median	56.5	–
Range	41–74	–
Performance status (ECOG)
0	7	18.4
1	28	73.7
2	3	7.9
Lauren classification
Intestinal	12	31.6
Diffuse	26	68.4
No. organs involved
1	5	13.2
2	7	18.4
≥3	26	68.4
Site of metastasis
Lymph node	34	89.5
Peritoneum	19	50.0
Liver	15	39.5
Others (lung, bone etc.)	9	23.7

ECOG, Eastern Cooperative Oncology Group.

Genotype analysis.  For the analysis of germline genetic polymorphisms, genomic DNA was extracted from pretreatment peripheral blood samples using the QIAmp DNA blood kit (Qiagen, Valencia, CA, USA). Sixteen polymorphisms in eight genes were investigated. The following polymorphisms were analyzed using polymerase chain reaction–restriction fragment length polymorphism methods: epidermal growth factor (EGF; G61A), TGFA (TaqI, RsaI, BamHI), thymidylate synthase (TS; 28‐bp repeat in the enhancer region, G/C polymorphism in the second repeat, 6‐bp deletion in the 3′‐untranslated region [UTR]), excision repair cross‐complementation group 1 (ERCC1; Asn118Asn, C8092A), Xeroderma pigmentosum group D (XPD; Arg156Arg, Asp312Asn), and fragment c gamma receptor (FCGR)2A (His166Arg). Polymorphisms in transforming growth factor α (TGFA) (C3296T, C3827T) and FCGR3A (Val212Phe) were analyzed by direct sequencing. Primer sequences and restriction enzymes are listed in Table 2. The CA dinucleotide simple sequence repeat (SSR) polymorphism in intron 1 of EGFR was analyzed with a fragment length analysis using fluorescently‐labeled primers, as described previously.^{( 17 )} All genotypes were in the Hardy–Weinberg equilibrium (data not shown).

Table 2.

 Primer sequences and restriction enzymes

Gene	Polymorphisms	Primers (5′–3′)	Restriction enzyme
EGF	G61A (rs4444903)	Forward: TGTCACTAAAGGAAAGGA Reverse: TTCACAGAGTTTAACAGCCC	AluI
TGFA	TaqI (rs11466267)	Forward: TTGTTTTGTTTTTTGAGACGG Reverse: GTGTGAGACTTTTCCAGCCCTGT	TaqI
	RsaI (rs3732248)	Forward: TGCCTCACCACGACAGACACA Reverse: TGAATAACCCCAAGCAGACGG	RsaI
	BamHI (rs11466297)	Forward: ACAGATGGCGGAAGCAGAGGT Reverse: CTAAAGGGCAAGGAAACACAG	BamHI
	C3296T (rs2166975)	Forward: GCTCTGCCATCTCCAAGT Reverse: ATCTCTGGCAGTGCTGTCCT	—
	C3827T (rs1058213)	Forward: TGGGGAAGAAAGTGAAGGAG Reverse: ATCTCCAAGGGTGGCGATAG	—
TS	28‐bp repeat in enhancer region (rs45445694)	Forward: GTGGCTCCTGCGTTTCCCCC Reverse: GCTCCGAGCCGGCCACAGGCATGGCGCGG	—
	G/C SNP in second repeat (rs2853542)	Forward: GTGGCTCCTGCGTTTCCCCC Reverse: GCTCCGAGCCGGCCACAGGCATGGCGCGG	HaeIII
	6‐bp deletion in 3′‐UTR (rs16430)	Forward: CAAATCTGAGGGAGCTGAGT Reverse: CAGATAAGTGGCAGTACAGA	DraI
ERCC1	Asn118Asn (rs11615)	Forward: TCATCCCTATTGATGGCTTCTGCCC Reverse: GACCATGCCCAGAGGCTTCTCATAG	BsrDI
	C8092A (rs3212986)	Forward: CAGAGACAGTGCCCCAAGAG Reverse: GGGCACCTTCAGCTTTCTTT	MboII
XPD	Arg156Arg (rs238406)	Forward: CACACCTGGCTCATTTTTGTAT Reverse: TCATCCAGGTTGTAGATGCCA	TfiI
	Asp312Asn (rs1799793)	Forward: CTGTTGGTGGGTGCCCGTATCTGTTGGTCT Reverse: (TAATA)TCGGGGCTCACCCTGCAGCACTTCCT	StyI
FCGR2A	His166Arg (rs1801274)	Forward: GGAAAATCCCAGAAATTCTCGC Reverse: CAGCGTGTAGCCTATGTTTCC	BstU1
FCGR3A	Val212Phe (rs396991)	Forward: TGGCAAAGGCAGGAAGTATT Reverse: ATTGCAGGTTCCACACACAG	—

EGF, epidermal growth factor; ERCC1, excision repair cross‐complementation group 1; FCGR, fragment c gamma receptor; SNP, single nucleotide polymorphism; TGFA, transforming growth factor‐α; TS, thymidylate synthase; UTR, untranslated region; XPD, Xeroderma pigmentosum group D.

Statistical analysis.  The statistical analysis of genetic status, baseline characteristics, and response rate was carried out using Pearson’s chi‐squared test, Fisher’s exact test, or linear‐by‐linear association test where appropriate. The correlation between the CA repeat number and tissue EGFR expression was examined using the Mann–Whitney U‐test or Spearman’s rank correlation. Serum ligand levels were compared using the Mann–Whitney U‐test or Kruskal–Wallis test. Median durations of progression‐free‐survival (PFS) and overall survival (OS) were calculated using the Kaplan–Meier method. Unadjusted comparisons of PFS and OS were made with log–rank tests. A multivariate analysis of response was performed with the backward stepwise logistic regression model. The multivariate analysis of PFS and OS was carried out using the backward stepwise Cox regression model. The following covariates were included to adjust for baseline characteristics: sex, age (older vs younger than median), ECOG PS (0 vs 1–2), Lauren classification, and additional characteristics with P < 0.20 (peritoneal seeding [response, PFS, OS], liver metastasis [OS], and number of organs involved [1–2 vs≥3, OS]). Same covariates were included in the multivariate analysis of the mFOLFOX6‐only patient cohort, but the ECOG PS was categorized as 1 versus 2 because no patient had a PS of zero in the study. In the backward stepwise model, the covariate selection was performed using likelihood ratio statistics based on the conditional parameter estimate. The criteria for entry and removal were 0.05 and 0.10, respectively. Two‐sided P‐values of <0.05 were considered significant. All analyses were performed using SPSS for Windows, version 12.0 (SPSS, Chicago, IL, USA).

Results

EGFR CA–SSR polymorphism.  The most frequent genotype of the EGFR intron 1 CA–SSR polymorphism was 16/20 repeats found in 11 patients, followed by 20/20 repeats in 10 patients. Repeat lengths in the remaining patients were 19/20 in four patients, 16/16, 15/20, and 15/16 in two patients each, and 20/22, 20/21, 18/20, 16/17, 15/19, 14/20, and 14/16 in one patient each. For the statistical analysis, we classified patients as having either low or high CA repeats, according to the sum of repeat numbers in both alleles based on our previous study in non‐small cell lung cancer.^{( 17 )} Twenty‐one patients had low repeats (sum ≤37), and 17 patients had high repeats (sum ≥38).

The tumor EGFR expression, determined by the immunohistochemistry score (intensity X percentage of positive cells), was higher in patients with low CA repeats compared to patients with high CA repeats (P = 0.011 by Mann–Whitney U‐test). We analyzed the association between the CA repeat length and EGFR expression in patients with at least one 20 repeat allele (n = 31), which is the most frequent allele, to see how the CA repeat number in a single allele affects the EGFR expression. Patients with longer CA repeats in the remaining allele tended to have a lower EGFR expression (Spearman’s ρ = −0.46, ρ = 0.010) (Fig. 1). No correlation was found between the CA repeat length and serum EGFR level (Spearman’s ρ = −0.060, P = 0.75).

Figure 1.

 CA repeat length and tissue epidermal growth factor receptor (EGFR) expression. Tissue EGFR expression determined by immunohistochemistry was analyzed according to CA repeat length in the remaining allele in patients with at least one 20 repeat allele. Immunohistochemistry (IHC) score was derived by multiplying the staining intensity and percentage of positive cells. Spearman’s ρ‐value was –0.46 (ρ = 0.010).

The CA repeat status (low vs high) was not associated with baseline characteristics. Although no significant association was found between the CA repeat status and response, PFS, or OS in the unadjusted analysis (Table 3), there were significant associations between CA repeat status and PFS and OS after adjusting for baseline characteristics. Patients with low CA repeats had longer PFS (adjusted HR 0.42 [95% confidence interval [CI] 0.19–0.96], P = 0.040) and OS (adjusted HR 0.40 [95% CI 0.16–0.99], P = 0.048) compared to patients with high CA repeats (Fig. 2A; Table 4). In addition, patients with low CA repeats were more likely to develop skin rashes (≥grade 2) compared with high‐repeat patients (66.7%vs 35.3%, respectively; P = 0.054).

Table 3.

 Genetic polymorphisms and treatment outcomes

	Criteria (no. patients)	Responders (%)	P‐value*	Median PFS (months)	P‐value**	Median OS (months)	P‐value**
EGFR CA–SSR†	≤37 (21)	10 (58.8)	0.69	5.5	0.64	14.4	0.22
	≥38 (17)	11 (52.4)		5.3		7.6
EGF 61	GG (20)	11 (55.0)	0.97	5.5	0.77	13.5	0.38
	GA, AA (18)	10 (55.6)		5.5		8.2
TGF‐α RsaI	CC (20)	10 (50.0)	0.49	5.6	0.55	16.9	0.60
	CT, TT (18)	11 (61.1)		5.5		8.5
TGF‐α TaqI	TAAT/TAAT(32)	18 (56.3)	1.00	5.5	0.87	9.9	0.81
	TAAT/– (6)	3 (50.0)		5.3		8.2
FCGR2A	HH (21)	14 (66.7)	0.12	5.6	0.61	13.5	0.89
	HR, RR (17)	7 (41.2)		5.5		9.9
FCGR3A	VV (18)	9 (50)	0.54	5.3	0.22	9.9	0.89
	VF (20)	12 (60)		5.5		9.2
TSER	3R/3R (28)	16 (57.1)	0.73	5.5	0.91	16.9	0.12
	2R/2R, 2R/3R (10)	5 (50)		5.5		5.6
TSER‡	High type (27)	13 (48.1)	0.28	5.5	0.50	18.8	0.024
	Low type (11)	8 (72.7)		5.5		8.5
TS 3′‐UTR	–6 bp/–6 bp (21)	11 (52.4)	0.69	5.5	0.41	16.9	0.65
	+6 bp/+6 bp, +6 bp/–6 bp (17)	10 (59.8)		5.5		9.9
ERCC1 118	CC (23)	12 (52.2)	0.64	5.6	0.81	14.4	0.94
	CT, TT (15)	9 (60)		5.0		9.2
ERCC1 8092	CC (28)	15 (53.6)	1.00	5.5	0.64	9.9	0.13
	AA, AC (10)	6 (60)		5.3		8.5
XPD 156	CC (16)	9 (56.3)	0.92	5.6	0.53	7.6	0.35
	AA, AC (22)	12 (54.5)		5.5		9.9

*P‐values by chi‐squared test or Fisher’s exact test; **P‐values by log–rank test; †sum of number of repeats in both alleles; ‡high type includes 2R/3G, 3C/3G, and 3G/3G; low type includes 2R/2R, 2R/3C, and 3C/3C. CA–SSR, CA simple sequence repeat; EGF, epidermal growth factor; ERCC1, excision repair cross‐complementation group 1; FCGR, fragment c gamma receptor; TGF‐α, transforming growth factor‐α; TS, thymidylate synthase; TSER, thymidylate synthase enhancer region; UTR, untranslated region; XPD, Xeroderma pigmentosum group D.

Figure 2.

 Kaplan–Meier curves of overall survival in patients treated with cetuximab plus modified leucovovin, fluorouracil, and oxaliplotion 6 (mFOLFOX6) (A) and in another cohort of patients treated only with mFOLFOX6 (B). Adjusted hazard ratio (adjusted HR) and P‐values were calculated with the backward stepwise Cox regression analysis with baseline characteristics as covariates. (A) Adjusted HR 0.40 (0.16–0.99), P = 0.048. (B) P = not significant. Adjusted HR and P‐value are not given in patients treated only with mFOLFOX6 because the CA repeat status was removed during the stepwise analysis.

Table 4.

 Multivariate analysis of survival

	Criteria (no. patients)	Adjusted hazard ratio (95% confidence interval)	P‐value
Progression‐free survival
EGFR CA–SSR	≤37 (21)	0.42 (0.19–0.96)	0.040
	≥38 (17)	1
Sex	Male (28)	1	0.050
	Female (10)	2.39 (1.00–5.69)
Age	<56.5 years (19)	1	<0.001
	>56.5 years (19)	0.17 (0.066–0.45)
Performance status (ECOG)	0 (7)	1	0.001
	1–2 (31)	7.30 (2.27–23.5)
Overall survival
EGFR CA–SSR	≤37 (21)	0.40 (0.16–0.99)	0.048
	≥38 (17)	1
Age	<56.5 years (19)	1	0.020
	>56.5 years (19)	0.37 (0.16–0.86)
Peritoneal seeding	No (19)	1	0.002
	Yes (19)	4.15 (1.68–10.2)
Liver metastasis	No (23)	1	0.003
	Yes (15)	3.57 (1.55–8.24)

Multivariate analysis was performed using the backward stepwise Cox regression model. Covariates entered were epidermal growth factor receptor (EGFR) CA simple sequence repeat (CA–SSR) (sum ≤37 vs≥38), sex, age (older vs younger than median), ECOG PS (0 vs 1–2), Lauren classification, peritoneal seeding in the progression‐free survival analysis and EGFR CA–SSR, sex, age, Eastern Cooperative Oncology Group (ECOG) performance status, Lauren classification, peritoneal seeding, liver metastasis, and number of organs involved (1–2 vs≥3) in the overall survival analysis. Covariates in the final models are shown.

In order to examine whether the longer PFS and OS in patients with low CA repeat was due to an innate good prognosis associated with low CA repeat regardless of cetuximab treatment, we evaluated the impact of the CA repeat status in a separate AGC patient cohort treated only with modified FOLFOX6.^{( 16 )} None of these patients received EGFR‐targeted treatment after disease progression. Sixty‐eight patients were assessable for CA repeat statuses. There was no significant difference in the response rate (37% in low repeat and 46.3% in high repeat, P = 0.45) and PFS (median 6.1 months in low repeat and 5 months in high repeat, P = 0.33). The OS was not different between the two groups (median 14 months in low repeat and 15.3 months in high repeat, P = 0.74; Fig. 2B). In the multivariate backward stepwise Cox regression analysis of PFS and OS for the adjustment of baseline characteristics, the CA repeat status was removed during the stepwise analysis. In summary, the CA repeat status was associated with survival only in patients who received cetuximab plus chemotherapy, but not in patients who were treated only with chemotherapy.

Ligand polymorphisms.  The genotype frequency of EGF G61A single nucleotide polymorphisms (SNP) was GG in 20 patients, GA in 17, and AA in one. There was no association between the presence of the A allele and baseline characteristics. The serum EGF level was not different between the GG and GA/AA genotypes (P = 0.75 by Mann–Whitney U‐test). Response, PFS, and OS were not affected by EGF G61A SNP.

Among the transforming growth factor‐α (TGF‐α) polymorphisms, all patients had the AA genotype in the BamHI SNP site. Genotypes in the RsaI, C3296T, and C3827T SNP sites were completely identical. The genotype was CC in 20 patients, CT in 13, and TT in five. No association was found between the genotype and baseline characteristics. There was no significant difference in the serum TGF‐ α level between the three genotypes (P = 0.37 by Kruskal–Wallis test). The response rate, PFS, and OS were not significantly different between the CC and CT/TT genotype. In the TaqI polymorphic site of TGF‐α, 32 patients had the TAAT/TAAT (C1C1) genotype, and six patients had TAAT/– (C1C2) genotype. There was no significant association between the genotype and baseline characteristics or treatment outcomes. The serum TGF‐α level was not significantly different in the two genotypes (P = 0.14 by Mann–Whitney U‐test).

FCGR polymorphisms.  The distribution of the FCGR2A polymorphism in H166R was HH in 21 patients, HR in 15, and RR in two patients. There was no difference in the baseline characteristics according to the genotype. There was a trend towards a higher response in the HH genotype patients compared with patients with the HR or RR genotype (adjusted odds ratio 3.40 [95% CI 0.82–14.08], P = 0.092). However, the difference in the response rate did not translate into difference in PFS or OS. There was no significant association between FCGR3A V212F SNP and baseline characteristics or treatment outcomes.

Polymorphisms related to 5‐FU and oxaliplatin.  In the analysis of genotype and baseline characteristics, the XPD 156 A allele was associated with diffuse type cancer: 100% (6/6) in the AA genotype, 75% (12/16) in the AC genotype, and 50% (8/16) in the CC genotype (P = 0.020 by linear‐by‐linear association). Patients with the 2R allele in the thymidylate synlhase enhancer region (TSER) or low‐type TSER genotype more frequently had liver metastasis (70% in 2R/2R or 2R/3R vs 28.6% in 3R/3R, P = 0.030; 72.7% in low type [2R/2R, 2R/3C, and 3C/3C] vs 25.9% in high type (2R/3G, 3C/3G, and 3G/3G), P = 0.012). In contrast, patients with the –6‐bp/–6‐bp genotype in the TS 3′‐UTR had a lower frequency of liver metastasis (23.8% in –6 bp/–6 bp vs 58.8% in +6 bp/+6 bp or +6 bp/–6 bp, P = 0.028).

Patients with the high‐type TSER genotype had a significantly longer OS compared to those with a low‐type genotype (P = 0.024 by log–rank test; Table 3), but this was not statistically significant in the multivariate analysis adjusting for the baseline characteristics listed earlier. The TSER genotype was removed during the stepwise analysis, whereas older age (HR 0.47, 95% CI 0.21–1.04), peritoneal seeding (HR 3.04, 95% CI 1.35–6.87), and liver metastasis (HR 4.09, 95% CI 1.79–9.36) were in the final model. In the case of ERCC1 C8092A SNP, patients with the CC genotype had a longer OS compared with those with the CA or AA genotype (adjusted HR 0.39 [95% CI 0.15–0.97], P = 0.044). However, ERCC1 C8092A was not retained in the stepwise multivariate analysis with baseline characteristics and CA repeat status as covariates, whereas CA repeat statuses remained in the final model.

Discussion

The identification of a predictive biomarker has become an important issue in the era of molecular‐targeted treatment. Treatment that is effective in a certain subgroup of patients can be futile in others. In the case of cetuximab in colorectal cancer, adding cetuximab to first‐line chemotherapy can only improve the treatment outcome in patients with K‐ras wild‐type tumors.^{( 7} , ^{18 )} In contrast, there was no additional benefit of cetuximab compared with chemotherapy alone in patients with K‐ras mutant tumors.^{( 7} , ^{18 )} Therefore, cetuximab is recommended for colorectal cancer patients only if they have a K‐ras wild‐type tumor.^{( 9 )} Other biomarkers potentially associated with the efficacy of cetuximab in colorectal cancer include B‐raf mutation, phosphartidylinositol 3‐kinase/phosphatase and tensin homolog (PI3KCA/PTEN) deregulation, EGFR gene amplification, EGFR ligand (epiregulin and amphiregulin) expression, and polymorphisms in EGFR, EGF, and FCGR.^{( 19} , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ^{25 )}

Trastuzumab, an anti‐HER2 antibody, has recently shown activity in AGC.^{( 4 )} Trastuzumab was the first molecular‐targeted agent to prove efficacy in gastric cancer, opening a new era of gastric cancer treatment. Patient inclusion of the study was limited to HER2‐positive tumors, which could have been the key to the study’s success. Whether cetuximab can improve the treatment outcome of gastric cancer will be addressed by the ongoing phase III study.^{( 12 )} However, patients are not selected based on molecular markers in the cetuximab study. In fact, there is no biomarker to predict the differential effect of cetuximab in gastric cancer. K‐ras mutation, which is a good predictor of the lack of benefit from cetuximab in colorectal cancer, is an uncommon genetic event in gastric cancer.^{( 13} , ¹⁴ , ^{26 )}

In order to find candidates for a predictive biomarker of cetuximab efficacy in gastric cancer, we performed an exploratory biomarker analysis in a phase II study of cetuximab. In our previous study focusing on tumor tissue and serum, patients with an EGFR expression and low ligand levels had better outcomes with cetuximab/mFOLFOX6 treatment.^{( 14 )} In the present study, germline genetic polymorphisms analyzed with peripheral blood mononuclear cell DNA are presented.

A shorter repeat of the EGFR intron 1 CA dinucleotide SSR polymorphism has been associated with poor survival in non‐small cell lung cancer and pancreatic cancer, suggesting its association with poor prognosis.^{( 27} , ^{28 )} In contrast, a short CA repeat was associated with better treatment results with either the EGFR tyrosine kinase inhibitor (gefitinib and erlotinib) or anti‐EGFR monoclonal antibody (cetuximab).^{( 17} , ²⁵ , ²⁹ , ^{30 )} We have previously reported that a low CA repeat was associated with a better response in non‐small cell lung cancer patients treated with gefitinib, independent of EGFR mutational status.^{( 17 )} Because there is no general agreement in the cut‐off of CA repeats, we applied the same cut‐off (sum ≤37 vs≥38) that was used in our previous study of gefitinib in non‐small cell lung cancer and found that a shorter CA repeat was independently associated with longer survival in AGC patients treated with cetuximab. Collectively, it is possible that the EGFR intron 1 CA repeat polymorphism is a common predictive biomarker for the treatment outcome of EGFR‐targeted agents in various types of cancers. In contrast, the CA repeat status did not impact survival among AGC patients not receiving cetuximab. This finding suggests that this polymorphism could be a predictive biomarker of cetuximab efficacy in gastric cancer, which merits further investigation in randomized studies. Moreover, as this study is the first to examine the polymorphism in gastric cancer, further studies regarding its prognostic role in gastric cancer is warranted.

The CA repeat polymorphism of EGFR has interethnic variability that Asians have higher repeat numbers compared with Caucasians.^{( 27} , ³¹ , ^{32 )} In the present study, a 20 repeat allele was the most frequent allele, and the distribution of repeat length was similar to that of Asian patients in previous studies.^{( 27} , ³¹ , ^{32 )} The longer repeat length in Korean patients could be one of the reasons why cetuximab/mFOLFOX6 showed disappointing results in previous study.^{( 14 )} In contrast, phase II studies of cetuximab in gastric cancer performed in Caucasian patients, who generally have shorter CA repeats, showed positive results.^{( 10} , ^{11 )} Therefore, it would be interesting to determine whether the interethnic difference in CA repeat distribution could affect the treatment outcome of cetuximab in different ethnicities. Intriguingly, the benefit of cetuximab was evident among Caucasians, whereas Asian patients did not benefit from cetuximab in a phase III study of cetuximab plus chemotherapy in non‐small cell lung cancer.^{( 33 )} Although other factors, such as imbalance in subsequent treatment, might have led to such a difference, interethnic genetic difference could also be considered.

The inverse association between the CA repeat number and tumor EGFR expression is in line with previous reports in breast and head and neck cancers, showing a higher EGFR expression in low CA repeats.^{( 34} , ^{35 )} We could not find association between genetic polymorphisms and tumor expression or serum level in other polymorphisms.

The biomarker selection for mFOLFOX6 was based on our previous pharmacogenomic study.^{( 16 )} However, the polymorphisms associated with poor response to mFOLFOX6 (TS 3′‐UTR and XPD 156) were not associated with outcome in this study.^{( 16 )} It is possible that the addition of cetuximab to mFOLFOX6 increased the response rates in patients with poor response alleles. This possibility could only be examined in a randomized study. The ERCC1 C8092A SNP was the only mFOLFOX6‐related polymorphism associated with outcome in the adjusted analysis in the present study. It has been reported that patients with the CC genotype have a longer OS compared to those with the CA or AA genotype in other cancers treated with platinum‐containing chemotherapy.^{( 36} , ^{37 )} However, the association was not significant in the multivariate analysis, with the EGFR CA repeat status as a covariate.

The present study was an exploratory biomarker study performed in a small‐sized single‐arm phase II study. Therefore, whether the EGFR intron 1 CA repeat polymorphism can predict the benefits of cetuximab in gastric cancer needs to be investigated in a randomized study.

In conclusion, a low repeat of the EGFR intron 1 CA repeat polymorphism was associated with longer survival in AGC patients treated with cetuximab plus mFOLFOX6, but not in patients treated only with mFOLFOX6.

Supporting information

Table S1. Baseline characteristics of modified leucovorin, fluorouracil, and oxaliplatin 6 (mFOLFOX6)‐only cohort.

Please note: Wiley‐Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

Supporting info item