Associations between skin rash, treatment outcome, and single nucleotide polymorphisms in head and neck cancer patients receiving the EGFR-inhibitor zalutumumab: results from the DAHANCA 19 trial

Abstract

Purpose

To study the associations between development of moderate to severe skin rash, clinical outcome, and single nucleotide polymorphisms (SNPs) in candidate genes in head and neck cancer patients from the DAHANCA 19 trial receiving the EGFR-inhibitor zalutumumab concurrently with radiation treatment.

Material and methods

310 patients were included from the zalutumumab-arm of the DAHANCA 19 study. Nine SNPs in the candidate genes EGFR, EGF, AREG, FCGR2A, FCGR3A, and CCND1 were successfully determined in 294 patients. Clinical endpoints were moderate to severe skin rash within the first 3 weeks of treatment, loco-regional failure (LRF), disease-specific survival (DSS), and overall survival (OS).

Results

During the first 3 weeks of treatment, 86% of the patients experienced any grade of rash and 17% experienced a moderate to severe rash. Development of moderate to severe rash was not associated with LRF or DSS but was associated with improved OS, HR 0.40 (95% CI: 0.19–0.82). The effect was similar for patients with p16-negative or p16-positive tumors (p = .90). After adjustment for comorbidity and performance status, the minor alleles of SNPs rs9996584 and rs13104811 located near the AREG gene were significantly associated with increased risk of moderate to severe rash with per-allele odds ratios of 1.61 (1.01–2.54) and 1.56 (1.00–2.44). SNP rs11942466 located close to rs9996584 had a borderline significant association, and none of the other SNPS were significantly associated with risk of skin rash.

Conclusions

Moderate to severe skin rash after zalutumumab during radiation treatment was associated with improved OS, independent of HPV/p16-status. Genetic variants in AREG (member of the EGF family) may be associated with increased risk of skin rash.

Introduction

In recent years, the use of epidermal growth factor receptor (EGFR) inhibitors have played an increasing role in the treatment of several solid tumor types including non-small-cell lung cancer, colorectal cancer, and head and neck squamous cell carcinoma (HNSCC). The main clinical manifestation of this treatment is skin toxicities such as dry skin and acneiform rash [1]. The development of skin toxicity during anti-EGFR treatment has been associated to treatment efficacy [2–6]. In HNSCC, monoclonal antibodies against the EGFR have been reported to improve tumor control and survival in locally advanced HNSCC in combination with radiation treatment (RT) as well as increasing survival in patients with recurrent or metastatic HNSCC receiving combination chemotherapy [7–9]. A recent study in 619 HNSCC patients randomized to addition of zalutumumab, a fully human immunoglobulin G (IgG1k) antibody against the EGFR, to primary RT or chemoradiation (CRT) did not increase loco-regional control (LRC) or overall survival (OS) [10]. Thus, tools for selecting patients for treatment are warranted.

The main modus of action of zalutumumab is inhibition of the ligand activation of the EGFR resulting in reduced angiogenesis, reduced cell proliferation, and reduced cell survival [11,12]. An alternative mode of action is antibody-dependent cell-mediated cytotoxicity (ADCC). The monoclonal antibodies against EGFR can mediate antigen-specific immune responses through direct killing of tumor cells by natural killer cells, monocyte lysis, or phagocytosis leading to antigen processing [13]. The efficacy of ADCC depends on the activation of effector cells after engagement of the IgG fragment C receptors FCGR2A and FCGR3A [14]. This immunotherapeutic approach has demonstrated clinical efficacy in several types of cancer diseases and has been proposed to be important in preventing metastasis [15,16]. Several single nucleotide polymorphisms (SNPs) have previously been associated with toxicity and survival in cancer patients [3,4,6,17–21]. We analyzed the influence of nine candidate SNPs assumed to be related to EGFR signaling, or ADCC, on skin toxicity and outcome in HNSCC patients treated with zalutumumab. We aimed to test the hypotheses that severe skin toxicity is associated with outcome in HNSCC patients from the phase III DAHANCA 19 trial receiving zalutumumab and whether nine previously evaluated SNPs in or near candidate genes potentially related to EGFR signaling (EGFR (rs2227983 and rs2293347), EGF (rs4444903), AREG (rs13104811, rs9996584, rs11942466), CCND1 (rs9344)) or ADCC (FCGR2A (rs1801274), FCGR3A (rs396991)) are associated with skin toxicity, loco-regional failure (LRF), disease-specific survival (DSS), and OS.

Material and methods

Patients and study design

In total, 310 patients with biopsy-proven HNSCC from the zalutumumab-arm of the DAHANCA 19 study were included from November 2007 to June 2012 [10]. The inclusion criteria were age above 18 years, no prior cancer treatment, eligibility for curatively intended radiation treatment, histologically verified squamous cell carcinoma of the pharynx, larynx (except stage I larynx and stage I–II glottic larynx cancer), WHO performance status 0–2, ability to go through the treatment and follow-up, and use of contraceptives for fertile women. Exclusion criteria were cancer of the nasopharynx or cancer of unknown origin, distant metastases, or other malignant diseases except for squamous cell skin cancer.

Of the 310 patients, 8 withdrew consent and 1 patient died before commencing treatment. Of the remaining 301 patients, 294 had material available for SNP analysis and were included in the present study. The patients were treated with primary accelerated RT 66–68 Gy, 2Gy/fx, 6fx/week, and concomitant hypoxic radiosensitization with nimorazole. Patients with disease stage III–IV (91%) received additional weekly cisplatin 40 mg/m² during RT. Zalutumumab 8 mg/kg was given weekly during RT with the first dose administered in the week before initiation of RT. Skin toxicity was prospectively evaluated by the Common Terminology Criteria for Adverse Events (CTCAE v3.0) modified scale (www.dahanca.dk). Grades 1 and 2 were considered mild, and grades 3 or 4 were considered moderate to severe. Patient and tumor characteristics are listed in Table 1.

Table 1. Patient, tumor, and treatment characteristics (N = 294).

Factor	Number of patients	%
Age (years)
Median	58
Range =31–79
≤60 years	181	62
>60 years	113	38
Sex
Female	55	19
Male	239	81
Smoking status
≤10 pack years	75	26
>10 pack years	219	74
Tumor site
Oropharynx	208	71
Other	86	29
Tumor stage
T1–2	179	61
T3–4	115	39
Nodal stage
N0	55	19
N+	239	81
Disease stage
I–II	27	9
III–IV	267	91
Differentiation
Moderate or well	134	51
Poor	131	49
Unknown	29	11
p16/HPV
Negative	130	45
Positive	161	55
Unknown	3	1
Performance
0	230	78
1	55	19
2	9	3
Comorbidity
No	13	4
Yes	281	96
Chemotherapy
No chemo	81	28
<5 cycles	65	22
≥5 cycles	148	50

DNA extraction and genotyping

Genomic DNA was manually extracted from leukocytes from pretreatment buffy coat samples using the DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA). DNA concentration was determined by the Qubit™ 3.0 Fluorometer. The polymorphisms were analyzed with TaqMan SNP genotyping assays (Taqman^® SNP Genotyping Assays, Thermo Fischer Scientific, Waltham, MA, USA) using the 7900 HT Fast Real-Time PCR System and the SDS 2.3 and TaqMan Genotyper Software (Thermo Fischer Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. The polymorphisms analyzed are presented in Supplementary Table S1.

Statistical analysis

The primary endpoint in the survival analysis was LRF defined as recurrent or persistent disease in the T- and N-site, with failure being local recurrence, regional recurrence, or persistence of the primary tumor. Secondary endpoints were OS estimated from the date of randomization to the date of death from any cause and DSS calculated from the date of randomization to the date of death from the primary cancer. Patients alive at the end of follow-up or lost to follow-up were censored at that time. The correlation between treatment-related skin toxicity and SNP genotypes to LRF, DSS, and OS was analyzed. The follow-up time was calculated from the date of randomization to the date of death or the cutoff date for the end of the study. Univariate analysis was performed to determine the association between covariates and the outcome measure. Multivariate analysis was performed including the covariates with a p value less than .05, and subsequently stepwise backward regression was performed. Kaplan–Meier curves were used to illustrate the association between the skin toxicity and OS. The log-rank test was used to evaluate the association between skin toxicity and OS. Hazard ratios were estimated using the Cox proportional hazards model, and the proportional hazards assumption was tested graphically using log-minus-log plots and corroborated by Schoenfelds’s residuals. LRF and DSS were based on cumulative incidence rates and illustrated with cumulative incidence rate plots with competing risk. Significance was evaluated by the risk difference at 60 months by a Wald test. A p value less than .05 was considered significant. When analyzing the endpoint of skin toxicity in association to the evaluated SNPs, logistic regression was performed using a log-additive genetic model. Multivariate analysis was performed including the covariates with a p value less than .05 and subsequently stepwise backward regression was performed. A likelihood ratio test was used to test SNP contribution in the multivariate logistic regression model. In the analyses performed, no correction for multiple testing has been performed. All analyses were performed using Stata version 14.2 (StataCorp, College Station, TX, USA).

Ethics

Patient samples were used with permission from the Danish Research Ethics Committee (EudraCT number 2007-002097-61). The procedures were followed in accordance with the Helsinki Declaration of 1975 (revised in 1983). Data handling procedures were approved by the Danish Data Protection Agency (case number 2005-41-4802). The registry for use of tissue was consulted before the use of patient material.

Results

Clinical characteristics

Patient and tumor characteristics are listed in Table 1. Among the 294 patients, most tumors were oropharynx (71%) and the remaining were larynx (14%), hypopharynx (12%), and oral cavity (3%).

Skin toxicity data were available for 287 patients. Two hundred and seventy two of the patients (95%) experienced any skin toxicity (above grade 0), and 85 patients (30%) experienced a moderate to severe skin toxicity (grade 3 or 4) at some point during the treatment course. The frequency of patients with moderate to severe skin toxicity at the individual visits before, during, and after start of radiotherapy treatment is shown in Figure 1.

Figure 1. Frequency of moderate to severe skin toxicity at the individual visits before, during, and after start of radiotherapy treatment.

The maximum frequency of moderate to severe skin toxicity was observed at the third visit after start of radiotherapy, corresponding to the third week. According to protocol, treatment with zalutumumab was discontinued in case of intolerable skin toxicity (grade 3/4). Thus, the following analyses are based on the frequency of moderate to severe skin toxicity (grade 3/4) during the first 3 weeks of treatment. During this period, 248 (86%) of the patients experienced any grade of skin toxicity and 50 (17%) experienced moderate to severe skin toxicity.

Skin toxicity and association to treatment outcome

Grade 3/4 skin toxicity within the first 3 weeks of treatment was found to be significantly associated with OS in univariate analysis (Figure 2) with skin toxicity grade 3/4 resulting in a better OS (83% vs. 63% after 5 years), (HR: 0.40, 95% CI: 0.19–0.82). The association of OS and skin toxicity was similar for patients with p16-positive and p16-negative tumors (test for interaction, p = .90). No association was found between skin toxicity, DSS (HR: 0.63, 0.29–1.40), and LRF (HR: 0.99, 0.55–1.76) in univariate analysis.

Figure 2. Associations between moderate to severe skin toxicity and loco-regional failure (A), disease-specific survival (B), and overall survival (C).

In univariate analysis, performance status, smoking status, p16-status, site, T-stage, histological differentiation, and treatment (with or without chemotherapy) were significantly associated with OS (Supplementary Table S2). Stepwise backward selection was applied leaving three significant covariates in the final model for OS: p16-status, T-stage, and performance status. Higher T-stage and poor performance status were associated with increased hazard, while p16-positivity was associated with decreased hazard. The same three covariates were included in multivariate analysis for LRF and DSS.

In multivariate analysis, the association between skin toxicity and OS was not statistically significant (HR: 0.52, 0.25–1.08), and skin toxicity was not associated with LRF (HR: 1.22, 0.67–2.21) or DSS (HR: 0.82, 0.37–1.86) (Table 2).

Table 2. Multivariate Cox analysis with moderate to severe skin toxicity, patient, tumor, and treatment characteristics for endpoints loco-regional failure, disease-specific survival, and overall survival.

	Loco-regional failure			Disease-specific survival			Overall survival
	HR	95% CI	p value	HR	95% CI	p value	HR	95% CI	p value
Skin toxicity (Grade 3–4 vs. 0–2)	1.22	(0.67–2.21)	.52	0.82	(0.37–1.86)	.64	0.52	(0.25–1.08)	.078
p16 status (positive vs. negative)	0.61	(0.36–1.02)	.057	0.35	(0.18–0.66)	.0013	0.34	(0.21–0.56)	<.0001
T-stage (T3–4 vs. T1–2)	1.44	(0.90–2.30)	.13	2.10	(1.19–3.69)	.010	1.67	(1.09–2.57)	.018
Performance status (1/2 vs. 0)	1.80	(1.05–3.09)	.032	1.77	(0.97–3.23)	.063	1.74	(1.09–2.75)	.019

SNPs and association with skin toxicity

SNPs were successfully amplified in 96.9% (2565/2646) of the samples with the majority of missing values from rs11942466. Characteristics of each SNP are listed in Supplementary Table S1. The genotyping frequency was found to be in Hardy–Weinberg equilibrium for each SNP by the chi² test (p > .05).

In univariate logistic regression analysis, none of the SNPs were significantly associated to skin toxicity (Supplementary Table S3). Among the covariates tested, poor performance status, presence of comorbidity, a smoking history of >10 packyears, and higher T stage were all associated with decreased risk of skin toxicity. After stepwise backward selection, only the presence of comorbidity and poor performance status were significantly associated with skin toxicity and were included in the final multivariate logistic regression analysis.

In multivariate analysis, two SNPs (rs9996584 and rs13104811) located near the AREG gene were significantly associated with skin toxicity with the minor allele being associated with increased risk. Per-allele odds ratios were 1.61 (1.01–2.54) and 1.56 (1.00–2.44), respectively (Table 3). Although not statistically significant, a third SNP (rs11942466) located near AREG also showed a trend toward association with skin toxicity, but with the minor allele being associated with decreased risk (OR: 0.70, 0.40–1.21). For all three SNPs and in both univariate and multivariate analysis, the largest effect size was observed in patients homozygous for the minor allele, with intermediate effect sizes in heterozygous (data not shown).

Table 3. Multivariate logistic regression of SNPs for endpoint moderate to severe skin toxicity.

	OR	95% CI	p value*
rs2227983	1.06	(0.64–1.77)	.81
rs2293347	1.04	(0.51–2.10)	.92
rs4444903	1.00	(0.64–1.56)	1.00
rs13104811	1.56	(1.00–2.44)	.050
rs9996584	1.61	(1.01–2.54)	.043
rs11942466	0.70	(0.40–1.21)	.19
rs1801274	0.90	(0.56–1.42)	.64
rs396991	0.89	(0.55–1.43)	.62
rs9344	0.76	(0.49–1.20)	.24

* p values from likelihood ratio tests comparing the goodness of fit of two multivariate models, one with performance status (OR 0.19 [0.06–0.63]) and comorbidity (OR 0.24 [0.07–0.84]), and one with performance status and comorbidity, and the individual SNP.

SNPs and association to treatment outcome

In univariate analysis, the SNP rs11942466 located near the AREG gene was associated with OS (HR: 1.44, 1.02–2.03) (Supplementary Table S4) but not LRF or DSS. None of the other investigated SNPs were associated to LRF, DSS, or OS in univariate analysis.

In multivariate analysis, the same covariates were included as in Table 2 (p16-status, T-stage, and performance status). Rs11942466 was significantly associated with LRF (HR: 1.44, 1.01–2.04), DSS (HR: 1.56, 1.02–2.38), and OS (HR: 1.5, 1.06–2.11) (Table 4). For all three endpoints, only the heterozygote genotype was significantly associated with increased risk (data not shown).

Table 4. Multivariate Cox analysis with SNPs for endpoints loco-regional failure, disease-specific survival, and overall survival.

	Loco-regional failure^a			Disease-specific survival^a			Overall survival^a
SNP	HR	95% CI	p value	HR	95% CI	p value	HR	95% CI	p value
rs2227983	0.71	(0.46–1.08)	.11	0.85	(0.52–1.38)	.51	0.84	(0.58–1.22)	.37
rs2293347	1.00	(0.59–1.69)	.99	0.90	(0.46–1.77)	.77	0.83	(0.49–1.41)	.49
rs4444903	1.01	(0.74–1.40)	.93	0.75	(0.50–1.13)	.18	0.80	(0.59–1.09)	.15
rs13104811	1.00	(0.73–1.36)	.98	0.78	(0.53–1.14)	.19	0.85	(0.63–1.13)	.26
rs9996584	0.89	(0.63–1.25)	.50	0.98	(0.66–1.46)	.92	0.81	(0.59–1.12)	.20
rs11942466	1.44	(1.01–2.04)	.042	1.56	(1.02–2.38)	.041	1.50	(1.06–2.11)	.021
rs1801274	0.91	(0.65–1.28)	.58	1.14	(0.76–1.71)	.53	0.95	(0.70–1.30)	.76
rs396991	0.90	(0.63–1.27)	.54	0.95	(0.63–1.44)	.80	0.93	(0.68–1.29)	.67
rs9344	1.09	(0.79–1.51)	.61	0.98	(0.66–1.45)	.91	1.07	(0.79–1.44)	.66

^aIncluded in multivariate analysis: p16 status, T-stage, and performance status.

Discussion

The results of applying EGFR-directed therapy with cetuximab in combination with radiotherapy with or without concurrent platinum-based chemotherapy (cisplatin) have been conflicting [22–27]. Recent studies have therefore investigated the possibility of identifying a subgroup of patients that may benefit from EGFR-directed therapy. Based on the RTOG experiences in head and neck cancer (trials 0234 and 0522), Bar-Ad et al. have shown an association between the severity of cetuximab-induced skin toxicity and OS in p16-negative patients, but not in p16-positive patients. No association with LRF was observed [6]. Others have identified SNPs that may be associated with toxicity and/or outcome in cancer patients treated with EGFR-directed treatment [3,4,6,17–21]. Lately, studies have suggested that activation of RAS signaling in the tumor is associated with response to cetuximab [28,29].

Based on the DAHANCA 19 phase III trial randomizing 619 HNSCC patients to RT/CRT with or without zalutumumab, we have studied the association between skin toxicity, treatment outcome, and SNPs. We evaluated skin toxicity during the first 3 weeks of treatment. In this period, we observed the highest percentage of skin toxicity and considered the data quality to be more valid since some patients may have had reduced treatment intensity during the following weeks due to side effects during the initial part of treatment.

In univariate analysis, we could confirm the association between development of moderate to severe skin toxicity and improved OS in agreement with the RTOG trials [6]. A similar effect has been observed with anti-EGFR treatment in other cancer diseases [17,30,31]. It has been proposed that systemic inhibition of EGFR reduces development of distant metastases which is a plausible explanation for improved OS while LRF remains unchanged [6].

In contrast to the RTOG trials, we found that the association was independent of p16-status. When adjusting for p16-status, performance status, and T-stage, there was still a trend for an association, but no longer statistically significant. Thus, the association between development of moderate to severe skin toxicity and improved OS may in part reflect individual biological differences, including different levels of immune activation. However, it may also reflect correlations between risk of skin toxicity and general characteristics associated with improved outcome (e.g., performance status and comorbidity) that are not linked with the EGFR system. Indeed, we found increased skin toxicity associated with good performance status, lack of comorbidity, short smoking history, and smaller T-stage.

When studying associations between candidate SNPs and the risk of skin toxicity, none of the SNPs displayed any significant associations. After adjusting for comorbidity and performance status, two SNPs near the AREG gene (rs9996584 and rs13104811) were found to be significantly associated with risk of skin toxicity. Rs9996584 has previously been shown to predict disease control and OS, and rs13104811 has been found to predict disease control in metastatic colorectal cancer patients treated with anti-EGFR plus irinotecan [17]. A third SNP near AREG (rs11942466) displayed a trend toward an association with skin toxicity. Whereas the minor alleles for rs9996584 and rs13104811 were associated with increased risk, the minor allele of rs11942466 was associated with decreased risk. This is in concordance with the previous observations [17]. We did observe an association between rs11942466 and treatment outcome. However, the associations were only significant for the heterozygous genotype, and we did not observe any associations between any of the other SNPs (including rs9996584 and rs13104811) and treatment outcome.

In summary, this study showed that moderate to severe skin toxicity was associated with improved OS in patients treated with zalutumumab. The mechanisms remain unclear and may reflect underlying biological differences in the EGFR system, the immune response, as well as so far unexplained associations between increased risk of skin toxicity and general patient and tumor characteristics linked with improved outcome. Among the SNPs tested, we could confirm the relevance of SNPs near the AREG gene encoding for the EGFR ligand amphiregulin, but the study did not confirm previously reported associations with skin toxicity and/or treatment outcome for SNPs in EGFR, EGF, FCGR2A, FCGR3A, and CCND1.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by The Danish Cancer Society (R146-A9519-16-S2), The Lundbeck Foundation Center for Interventional Research in Radiation Oncology (CIRRO), and the Danish Council for Strategic Research (0603-00136B).