The number of CA tandem repeats (CA)n in a highly polymorphic region of EGFR (epidermal growth factor receptor) intron 1 may affect gene transcription; the potential impact of allelic variants on the efficacy of cetuximab in metastatic colorectal cancer (mCRC) patients is debated for long. This study aimed at prospectively estimating the impact of EGFR intron 1 (CA)n variants on clinical outcome in KRAS exon 2 and BRAF wild-type chemo-refractory mCRC patients, receiving cetuximab and irinotecan. Variants presenting<and 20 CA repeats were defined as short (S) and long (L), respectively. One hundred and fifteen patients were included. No differences in progression-free survival or overall survival were observed between L- and SS genotypes (hazard ratio (HR)=1.00 (95% confidence interval (CI): 0.67–1.50), P=0.991; HR=1.32 (95% CI: 0.81–2.16), P=0.261). Exploratory analyses adopting other cutoff values previously described led to similar results. This prospective study did not confirm any previous retrospective finding, reporting a predictive or prognostic effect of EGFR (CA)n repeats allelic variants in chemo-refractory mCRC patients receiving cetuximab and irinotecan.
Results from both retrospective series and post-hoc analyses of randomized trials disclosed that metastatic colorectal cancer (mCRC) patients harboring KRAS (v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog) activating mutations do not achieve benefit from anti-EGFR (anti-epidermal growth factor receptor) monoclonal antibodies.1, 2, 3, 4 Although KRAS mutations are strong predictors of resistance to anti-EGFR agents, only about 50–60% of KRAS wild-type (WT) mCRC patients derive benefit from EGFR inhibitors in advanced lines, thus suggesting a role for additional mechanisms of intrinsic resistance.5 The need to further refine patients’ selection, by identifying new reliable predictive factors, lead to the investigation of other biomarkers, mainly involved in intracellular pathways of signal transduction. Several experiences showed that the presence of BRAF (v-Raf murine sarcoma viral oncogene homolog B1) mutation is associated not only with poor prognosis independently of administered treatments but also that patients bearing this alteration achieve a very slight benefit from anti-EGFR antibodies.6, 7
Nowadays, despite several attempts, no new molecular predictors of sensitivity to these agents have entered clinical practice, as many promising retrospective experiences have not been confirmed in prospectively designed trials. Nevertheless, as side effects and costs of these agents are not negligible, the identification of efficacious predictive biomarkers is still a high priority in translational researchers’ lists.
Different retrospective studies investigated the impact of the number of CA repeats ((CA)n) in a highly polymorphic sequence of the intron 1 of EGFR on the efficacy of cetuximab in mCRC patients.8, 9, 10, 11, 12, 13 Although no clear data are currently available about the prognostic implication of this marker, a potential relation of longer variants with worse outcome to cetuximab was suggested by two retrospective analyses.8, 10, 14 The biological rationale supporting the evaluation of the potential predictive impact of this marker relies on the preclinical finding that increasing (CA)n repeats are associated with decreasing levels of mRNA and protein expression.15, 16 Nevertheless preclinical studies did not provide conclusive results about the relationship between EGFR expression and the number of EGFR intron 1 CA tandem repeats.17 However, it remains an intriguing hypothesis that, conversely, a low number of CA repeats, leading to increased mRNA levels and protein expression, might be associated with a higher sensitivity to anti-EGFR treatment, because of a possible ‘oncogenic addiction’ on the EGFR pathway.
Results from previous clinical experiences are extremely contrasting. As summarized in Table 1, all of these studies adopted different cutoff values to define polymorphic variants and included patients with heterogeneous clinical and molecular characteristics.
The present study moves from the purpose to clarify the potential impact of this ‘biologically sound’ determinant, by recognizing the importance of methodological efforts to translate molecular markers from the bench to the bedside. Therefore, we prospectively evaluated, for the first time, the association of EGFR intron 1 (CA)n repeats allelic variants with clinical outcome in a population of KRAS exon 2 and BRAF WT, chemo-refractory mCRC patients receiving cetuximab and irinotecan in advanced lines of treatment.
Patients and methods
Main inclusion criteria were the following: histologically confirmed diagnosis of colorectal adenocarcinoma; measurable lesions according to RECIST 1.018 (Response Evaluation Criteria In Solid Tumors); two previous lines of treatment for the metastatic disease, including oxaliplatin and irinotecan; irinotecan-refractory disease (that is, progressed during or within 3 months after treatment with an irinotecan-based regimen); no previous treatment with anti-EGFR monoclonal antibodies; KRAS exon 2 (codons 12 and 13) and BRAF (V600E) WT status, as assessed by means of PCR and High Resolution Melting, followed by direct sequencing, as per clinical practice on the primary tumor; and available blood samples for EGFR intron 1 genotyping.
Treatment consisted of biweekly cetuximab and irinotecan (cetuximab 500 mg per sq.m intravenous (IV), day 1; irinotecan 180 mg per sq.m IV, day 1).19, 20, 21, 22 Tumor response was evaluated every 8 weeks by computerized tomographic scan according to RECIST 1.0. The toxicity profile was assessed by means of NCI CTCAE version 3.
Investigators performing molecular analyses were blind to clinical data. The Ethics Committee of Livorno Hospital approved the protocol (Approval Number 2008-0004). Patients provided their written informed consent for blood sampling and genetic analyses.
In a subset of patients, paired plasma samples were collected in order to investigate the modulation of EGFR endogenous ligands during the treatment (data presented elsewhere).23
EGFR intron 1 genotyping
Venous blood samples were collected to genotype EGFR intron 1 (CA)n repeats polymorphism. Genomic DNA was extracted from blood using the QIAamp Kit (Qiagen, Valencia, CA, USA). EGFR (CA)n repeats polymorphism was determined by a 5′-end 33pγ ATP-labeled PCR protocol.
Briefly, 100 ng gDNA, 200 μM dNTPs, 1.0 μM 5' 33pγ ATP end-labelled reverse primer, 1.0 μM unlabelled forward primer, 0.75 U Taq polymerase (Perkin Elmer Inc., Shelton, CT, USA) and PCR buffer (10 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl2) were used together in a final PCR volume of 15 μl. The reaction was carried out for 28 cycles with denaturation at 94 °C (1 min), annealing at 55 °C (1 min) and extension at 72 °C (2 min). The reaction products were separated on a 6% denaturing polyacrylamide DNA sequencing gel, which was vacuum blotted for 1 h at 80 °C and exposed to an XAR film (Eastman-Kodak Co., Rochester, NY, USA) overnight. The exact number of CA repeats on homozygous samples was confirmed by direct sequencing.
For the primary end point analysis, EGFR intron 1 (CA)n repeats polymorphism was defined according to a previously adopted criterion,10, 24 defining allelic variants with<and 20 CA repeats as short (S) and long (L), respectively.
Exploratory analyses were conducted adopting other cutoff values reported in previous studies. According to the definition used by Graziano et al.,8 allelic variants presenting<and 17 CA repeats were defined as S and L, respectively. According to the definition adopted by Dahan et al.,12 allelic variants presenting a sum <and 35 repeats were defined as S and L, respectively.
The current study was prospectively designed to test the effect of EGFR intron 1 (CA)n repeats polymorphism on progression-free survival (PFS) in patients with KRAS exon 2 and BRAF WT chemo-refractory mCRC, treated with biweekly cetuximab and irinotecan. PFS, the primary end point, was defined as the time from the first administration of study treatment to the first documentation of objective disease progression according to RECIST 1.0 or death due to any cause, whichever occurred first. Tumor response per RECIST, disease control and overall survival (OS) were the secondary end points. Tumor response was categorized as binary: response (complete or partial response) or non-response (stable or progression disease). Disease control was dichotomous as yes (complete or partial response or stable disease) or no (progressive disease). OS was defined as the period from starting study treatment to death. OS was censored at the latest date of being known alive.
The null hypothesis was that there was no difference in PFS between patients carrying two short EGFR intron 1 (CA)n repeats polymorphism (SS) and those carrying any long EGFR intron 1 (CA)n repeats polymorphism (L-) (hazard ratio, HR=1). Ninety-nine events were required to have 80% power to detect with a two-sided 0.05 level log-rank test, a HR of 1.75 for L- (prevalence of 40%), compared with SS genotype, that was set on the basis of previous publications.
The association of allelic variants with PFS and OS was analyzed using Kaplan–Meier curves and the log-rank test. The association of allelic variants with tumor response rate and disease control rate was examined using the contingency tables and the two-sided Fisher’s exact test.
All analyses were conducted using the SAS version 9.3 (SAS Institute, Cary, NC, USA). The tests were two-sided at a significance level of 0.05. P-values were not adjusted for multiple comparisons for the secondary end points.
Between April 2008 and June 2011, one hundred and fifteen mCRC patients with KRAS exon 2 and BRAF WT disease, treated with biweekly cetuximab and irinotecan, were included. Main patients’ characteristics are summarized in Table 2.
EGFR (CA)n repeats polymorphism was successfully genotyped in 113 cases. L- and SS variants were found in 45 (40%) and 68 (60%) cases, respectively. Consistently with literature data, homozygous 16–16 CA repeats was the most frequent genotype (35%), followed by heterozygous 16–20, 16–17 and 16–18 CA repeats (19%, 6% and 6%, respectively).12 All bi-allelic genotypes agreed with those predicted by the Hardy–Weinberg equilibrium.
According to RECIST criteria, out of the 113 patients, 31 were partial responders and 1 was complete responder, for a global response rate of 28%. Forty-four (39%) patients achieve stable disease, with a disease control rate of 67%. At a median follow-up of 21.9 months, 104 PFS events and 75 deaths were observed. Median PFS and OS were 5.2 and 13.3 months, respectively.
No differences in PFS and OS were observed between L- and SS genotype (median PFS: 4.4 versus 5.3 months, HR=1.00 (95% confidence interval (CI): 0.67–1.50), P=0.991; median OS: 11.3 versus 14.2 months, HR=1.32 (95% CI: 0.81–2.16), P=0.261; Table 3; Figures 1a and b). Among 112 evaluable patients, 10 (22%) out of the 45 L- patients achieved response compared with 22 (33%) out of the 67 SS patients (Fisher’s exact test: P=0.287, Table 3).
With regard to the relation of skin toxicity with clinical outcome, among 62 (55%) patients developing G2 or G3 skin rash, longer PFS was reported compared with those (N=53) with G1 or no skin rash (median PFS: 6.2 vs 3.8 months, HR=0.68 (95% CI: 0.43–0.99), P=0.046). A trend toward longer OS (median OS=15.7 vs 10.9 months, HR=0.66 (95% CI: 0.40–1.03), P=0.067) was also reported. No significant difference in terms of response rate (23 vs 34%, P=0.221) was observed. No correlation of skin rash with the prevalence of EGFR intron 1 (CA)n repeats variants was found (P=0.179).
Patients with the left-sided primary location (N=73) had a better outcome in terms of PFS, compared with the right-sided ones (N=18; median PFS 5.3 vs 2.6 months, HR=0.53 (95% CI: 0.22–0.83), P=0.013). No differences were observed in OS (median OS 13.1 vs 11.7 months, HR=0.67 (95% CI 0.29–1.35), P=0.23), according to tumor site.
Furthermore, we conducted exploratory analyses adopting two different cutoff values to define L and S variants of EGFR intron 1 (CA)n repeats. As reported in Table 3, no association of EGFR intron 1 (CA)n repeats polymorphic variants with clinical outcome was evidenced.
In the landscape of potentially useful biomarkers, germ-line polymorphisms are excellent candidates. In fact, they can be easily assessed in normal tissues, such as venous blood, and in most cases they can be determined by means of simple, accessible and well-standardized techniques.
In order to identify actually relevant alterations, a rational strategy may consist in the investigation of genetic determinants with a biological impact in terms of gene expression modulation or protein aberrations.
This is the case of EGFR intron 1 (CA)n repeats, whose allelic variants seem to be associated with significant differences in mRNA transcription and protein expression in preclinical experiences, so that a biological rationale seems to support the investigation of EGFR intron 1 (CA)n repeats as a potential predictor of benefit from anti-EGFR monoclonal antibodies.15, 16 Some evidences seemed to confirm this hypothesis by reporting a consistent relation of longer variants, associated with lower expression levels, with poor clinical outcome.10 Nevertheless, contrasting findings came out of other retrospective series.8, 9, 11, 12
The present study has been conceived with the aim to prospectively challenge the potential role for EGFR intron 1 (CA)n repeats as a marker of benefit from anti-EGFRs. Our results in terms of treatment’s activity and efficacy are in line with previous cohorts in literature in the same molecular and clinical setting 6. No association with clinical outcome was reported, neither by adopting our previous definition of allelic variants nor by applying other cutoff values as reported by other experiences.
The main strong point of this study lies in its prospective design, based on a well-established statistical hypothesis. As strongly recommended by methodologists and statisticians, the prospective validation of preliminary results is an essential step to clarify the reliability of retrospective suggestions that should be uniquely regarded as hypothesis-generating.25 We recognize that the approach adopted (single-arm, not randomized vs control) does not allow to discriminate the predictive vs prognostic role of a candidate biomarker and may require a further step of validation. However, our negative results can be easily interpreted even in the absence of a control arm. By a methodological point of view, we also believe that in the real life, the ‘retrospective-prospective’ approach may provide an acceptable way out in order to challenge preliminary findings and to push or not their further development toward clinical practice.
Moreover, the study population includes irinotecan-refractory patients who had already received two previous treatment lines, prospectively enrolled to receive biweekly cetuximab plus irinotecan. This choice allows to avoid the potential confounding effect of both different combined chemotherapy regimens and different lines of treatment, affecting other retrospective series. Differently from first-line trials, this setting allows to appreciate the actual impact of cetuximab without the confounding effect of the associated chemotherapy backbone. However, as this is a germ-line SNP, its impact should not be influenced by changes in tumor biology across subsequent lines of treatment.
Finally, only patients with KRAS and BRAF WT patients were included. Although only KRAS WT status is formally required by regulatory authorities to administer cetuximab,26 the choice not to include patients with BRAF-mutated disease is mainly founded on the minimal benefit achieved by this subgroup and on current clinical attitudes at the participating centers. Moreover, as a peculiar molecular profile identifies BRAF-mutated tumors as a specific subset, though some differences between microsatellite stable and unstable tumors exist, this choice allows to select a rather homogenous population, where the potential impact of the investigated biomarker might have been more properly appreciated.
In the past months, the role of rare KRAS and NRAS mutations, beyond KRAS exon 2 ones, as predictors of resistance to anti-EGFR monoclonal antibodies, was ascertained. Even if tumor samples of included patients were not extensively characterized with regard to the RAS panel, its quite unlikely that results of the present trial have been somehow affected: a correlation of somatic RAS mutations with germ-line EGFR intron 1 allelic variants has no rational basis, and the small percentage of RAS mutant patients in our cohort may have theoretically lowered the power of our study, but the HR=1.00 in PFS comparing L- to SS genotypes does not suggest any minimal difference.
The analysis according to tumor site confirms the better outcome of the left-sided compared with the right-sided KRAS and BRAF WT tumors when treated with cetuximab. Although extremely exploratory in nature, this observation is consistent with the previous finding by Missiaglia et al.27 reporting different pathological and molecular patterns according to tumor site, with potential therapeutic implications.
With regard to the adoption of a cutoff value, we recognize that the choice of a ‘preferred’ value is certainly arbitrary until reproducible results, attesting its reliability, are achieved. Nevertheless, the choice to use a cutoff value to dichotomize a continuous variable is quite common in translational research and is justified by the attempt to identify a biomarker easily transferable to clinical practice.
In conclusion, negative results coming from this prospective experience highlight the crucial relevance of applying a rigorous research methodology to clarify the potential role of apparently promising markers and to stop or prosecute their way toward clinical practice.25, 28, 29 The complexity of cellular networks makes rather unlikely that a single germ-line polymorphism might have a noteworthy impact on tumors’ sensitivity to a targeted agent. New translational approaches, based on innovative technologies, such as next-generation sequencing and genome-wide association studies, will probably be useful in achieving a comprehensive overview of tumor biology and in further optimizing the use of anti-EGFR monoclonal antibodies.
About this article
Association betweenEGFRgene polymorphisms, skin rash and response to anti-EGFR therapy in metastatic colorectal cancer patients
Experimental Dermatology (2014)