Prognostic and predictive value of EGFR in head and neck squamous cell carcinoma

EGFR is an extensively studied biomarker in head and neck squamous cell carcinoma (HNSCC). In this review, we discuss the prognostic and predictive role of EGFR in HNSCC, focusing on the different molecular alterations in specific treatment modalities such as radiotherapy alone (RT), combination of surgery, RT and chemotherapy (CT), EGFR inhibitors. We considered EGFR at different molecular levels: protein expression, protein activation, gene copy number, polymorphisms, mutation, EGFRvIII expression and EGFR ligand expression. Considering RT alone, evidence supports the predictive and prognostic role of high EGFR expression only when evaluated by quantitative assays: this may help select the patients who can mostly benefit from accelerated treatment. Conversely, no predictive biomarkers are available when treatment is a combination of surgery, CT and RT. For this combined treatment, several studies indicate that EGFR expression represents a good prognostic parameter only when measured by a “quantitative” or at least semi-quantitative method. With respect to EGFR inhibitors, neither EGFR expression nor increased gene copy number represent prognostic/predictive factors. If validated, nuclear EGFR, TGFα levels, EGFR phopshorylation and polymorphisms could represent additional prognostic factors in relation to combination of surgery, CT and RT, while EGFR polymorphisms and high amphiregulin levels could have prognostic value in patients treated with EGFR inhibitors.


INTRODUCTION
Head and neck squamous cell carcinoma (HNSCC) represents the sixth most frequent cancer worldwide with an incidence of 560,000 cases/year and over 350,000 deaths annually [1]. Treatment of locally-advanced HNSCC requires a multidisciplinary approach involving surgery followed by radiation (RT) with or without chemotherapy (CT) or curative RT with concomitant CT or anti-EGFR agents. TP53 gene status [2,3] and HPV [4] are the most studied biological markers with known prognostic value. The extensive studies of HPV have paved the way for tailored therapeutic strategies, with the aim of sparing toxicities in HPV-positive tumors and intensifying treatment in HPV-negative cancers.
As observed also for other malignancies, an extensively studied biomarker in HNSCC is the epidermal growth factor receptor (EGFR), a cell surface receptor member of the ErbB family. Activation of EGFR leads to a phosphorylation cascade mediated via tyrosine kinases which works downstream through the PI3K-PTEN-AKT, MAPK, ERK, and Jak/STAT pathways and promotes proliferation, invasion, angiogenesis, and metastatic spread. Evidence of EGFR activity has been reported in HNSCC cell lines, as well as in a high percentage of primary HNSCC [5][6][7]. Aberrant activation of EGFR signaling in HNSCC may be achieved by several mechanisms, including overexpression of EGFR and its ligands, establishing autocrine/paracrine loops, EGFR gene amplification, EGFR mutation/polymorphism and transactivation by other receptor tyrosine kinases (RTKs). The relevance of EGFR pathway in HNSCC led to the successful development of cetuximab in both the curative and palliative settings [8,9] and to the conduction of several trials with other antibodies directed against EGFR, such as panitumumab, zalutumumab and nimotuzumab [10][11][12][13], or RTK inhibitors involving downstream EGFR signaling [14,15]. Several studies have investigated the prognostic and predictive value of EGFR in HNSCC. In this review, we discuss available evidence on this topic, focusing on the different EGFR molecular alterations in tumor tissue, in relation with different treatments and settings.

SEARCH CRITERIA
To identify the key publications on EGFR prognostic or predictive value in HNSCC, we conducted a comprehensive literature search in the online database Medline. The search was last updated on October 2015 and included only articles in English, with no limitation on the publication date. Articles were selected for inclusion and assigned to each single treatment section, as judged by the Authors.
Clinical outcome was evaluated in terms of clinical response, overall survival (OS), progression-free survival (PFS), disease-free survival (DFS), locoregional control (LRC), locoregional relapse (LRR), locoregional failure (LRF), disease control rate (DCR) or time to treatment failure (TTF), depending on the reported results of considered studies.
A prognostic biomarker was defined as any tumor characteristic that informs about cancer outcome. In more detail, a biomarker was defined as prognostic when patients with tumor showing a specific characteristic have different survival than subjects without that specific characteristic, independently from the treatment [16]. A predictive biomarker was defined as a tumor characteristic that can be used to predict the tumor response to a specific treatment. In particular, the biomarker is considered predictive if the treatment effect is different for patients with tumor showing a specific characteristic compared with patients without that specific characteristic [16]. Therefore, a predictive biomarker can be evaluated only in head-to-head studies presenting both treated and control arms.
In our analysis we considered EGFR at different cytogenetic/molecular levels: protein expression, protein activation, gene copy number, polymorphisms, mutation, EGFRvIII expression and EGFR ligand expression. From identified papers we retrieved prognostic and predictive information regarding EGFR alterations according to the treatment provided. Treatments were grouped as: radiotherapy alone (RT); combination of surgery, RT and chemotherapy (CT); EGFR inhibitors.

Methods of assessment
The expression of EGFR protein has been evaluated by several means (Table 1). EGFR immunohistochemistry (IHC), a relatively easy and inexpensive technique, represents the most frequent option; moreover, tissue microarrays enable multiple samples to be stained at once, favoring investigation of biomarkers in large case series of IHC. However, in available studies, EGFR immunoreactivity was heterogeneously evaluated using different cut-off values and following different criteria for intensity and/or extent of the staining, as well as cytoplasmic and/or membranous staining.
Quantitative in situ molecular-based methods have been developed to define EGFR expression by IHC thus avoiding the subjectivity of visual assessment. Another way of determining EGFR expression is the binding assay, a less-used method based on frozen samples processed with radioactive labeled EGF and used to estimate EGFR concentration.
Similarly, an intense EGFR expression assessed by IHC did not influence LRC, DFS and OS in two different series of oropharyngeal SCC patients treated with primary RT [21,22], as well as in response of oral tongue SCCs treated with preoperative RT [23].
By contrast, in two HNSCC series not differentiated by subsite and treated with exclusive RT, EGFR expression was a robust prognostic factor only when a "quantitative" EGFR image analysis-based IHC assay was performed. In one study, EGFR expression, quantitatively evaluated by using the SAMBA system, was a strong independent prognostic indicator, capable to improve the estimation of OS, DFS and LRC applied to pretreatment biopsy specimens from patients assigned to the standard therapy arm receiving conventional RT [24]. The same "quantitative" approach was used in a series of HNSCC patients enrolled into a Phase III trial who received accelerated or standard fractionated RT [25]. Regardless of treatment modalities, high EGFR expression was associated with higher LRR (Relative Risk: 1.91, P=0.0163) and lower OS (Relative Risk: 1.90, P=0.0010). Collectively, these findings show that high EGFR expression assessed by a quantitative method might  [25]. Once again, IHC analysis performed in pretreatment tumor biopsies from HNSCC patients revealed a beneficial role of high EGFR expression in patients assigned to continuous hyperfractionated accelerated RT (CHART) compared with conventionally-fractionated RT [26]. Particularly, among patients showing high EGFR expression (cells with EGFR membrane staining ≥25%), a significant benefit in 3-year LRC rate was observed in the CHART arm compared with the conventionally-fractionated RT arm. No difference between the two treatment arms was observed in the low EGFR expressing group. However, EGFR expression had no significant effect on OS or distant metastases [26].
Another study suggests the predictive value of EGFR expression in HNSCC patients treated exclusively with RT [27]. Patients with supraglottic larynx SCC were treated with primary RT at the same total dose but with different OTT: 9½ weeks, 6½ weeks or 5½ weeks. Using LRC as endpoint, the results showed that patients whose tumors had high EGFR expression at IHC assay benefitted from a reduction in RT overall treatment time from 6½ to 5½ weeks more than subjects with low EGFR levels.
Collectively, high EGFR expression may be useful for selecting HNSCC patients who will benefit from accelerated RT in terms of better LRC.

1B) COMBINATION OF SURGERY, RADIOTHERAPY AND CHEMOTHERAPY Prognosis
Among the studies that used binding assays, some investigated the role of EGFR within organ preservation strategies such as neoadjuvant CT followed by RT or concurrent CT/RT. No correlation between EGFR levels and response to these treatments was disclosed. However, with such technique, most studies reported a positive correlation between high concentrations of EGFR and poorer prognosis [28][29][30][31][32]. The binding assay confirmed also EGFR expression to be an independent prognostic factor of neck node relapse in HNSCC patients undergoing surgical resection ±RT [33].
The first attempts to establish a relationship between EGFR expression measured by IHC and PFS and OS were performed by two independent groups several years ago [34,35]. In both cases, high EGFR expression was a predictor of reduced DFS and OS at multivariate analysis. However, while Maiorano et al [34] analyzed by EGFR IHC a series of oral cavity SCC treated with surgery and scored samples as positive when presenting at least 10% of either membranous or cytoplasmic stain, Rubin Grandis et al [35] relied on an automated image analysis system (SAMBA) to quantify EGFR expression by IHC in the primary tumor of patients who underwent surgery ±RT and CT. More recently, the prognostic impact of EGFR protein level measured by quantitative approaches was confirmed by two independent studies which assessed EGFR expression through the AQUA method [36,37]. In particular, Psyrri et al [36] composed a tissue microarray of primary, non-metastatic oropharyngeal SSC treated with primary external beam RT or gross total surgical resection and postoperative RT (±CT). The AQUA scoring system showed that patients with high cytoplasmic or nuclear EGFR expression were more likely to experience local recurrence. In a tissue microarray composed of HNSCC treated with CTRT, AQUA analysis revealed EGFR protein levels as a strong predictor of patient outcome [37].
On the other hand, discordant results were obtained when EGFR expression was assessed by IHC without a quantitative approach. In most of the studies involving case series of HNSCC treated with surgery ±CT/RT, semi-quantitative criteria were adopted considering both intensity and extent of membranous staining. However, no real concordance in the scoring system across studies was disclosed. For this reason, a direct comparison of different studies could lead to misleading results. Furthermore, these series were published over several decades, with cases treated as late as in the '70s. It is therefore reasonable to assume that the increasing quality of treatment in recent years, especially for what concerns RT, might have influenced outcomes thus introducing other confounding factors in the comparison of these case series. Given these premises, some studies showed a statistical association between clinical outcome and EGFR overexpression [38][39][40][41][42][43][44], while other analyses failed to disclose such correlation [45][46][47][48][49][50][51][52][53][54][55][56]. A recent meta-analysis of 33 studies showed that although HNSCC patients with high EGFR expression had a poorer OS regardless of the type of treatment, a large heterogeneity was reported, mainly related to tumor site and IHC scoring system [57]. With respect to IHC, it is worth mentioning that when the studies were stratified according to the score systems, only the group that evaluated the staining based on a combination of intensity and extent showed significant association between EGFR expression and OS.
Two other studies investigated EGFR expression by IHC as a prognostic factor in patients treated with combination of CT and RT [58,59]. They used several different chemotherapeutic agents as well as different scoring system, none of them relying on a quantitative approach. These studies failed in identifying EGFR overexpression as a prognostic factor. A positive study is represented by the work of Kumar et al [60], who prospectively evaluated EGFR expression, together with other markers, in advanced oropharyngeal SCC patients treated with one cycle of cisplatin or carboplatin and fluorouracil. Responders (i.e., those with 50% response at the primary site) received CT/RT; non-responders received surgery and RT. EGFR overexpression was associated with poor response to induction CT. Similar results were reported by Hitt et al [61].
The issue of EGFR expression deserves a further comment relative to oropharyngeal cancer, given the strong association between these tumor sites and the presence of HPV conferring a favorable prognosis. Several studies did not investigate HPV infection, therefore potentially altering the evaluation of the prognostic value of EGFR expression. An inverse correlation between HPV positivity and EGFR expression has been reported [22, 50,  showing that the use of EGFR expression in combination with HPV status provides additional prognostic information. The prognosis of patients with EGFRpositive/HPV-negative cancer was the poorest, while the EGFR negative/HPV-positive group showed the best outcome. However, in recent studies on oropharyngeal HNC patients, the prognostic role of EGFR expression was related to the association with HPV-negative tumors, and the added value of EGFR analysis seems to be marginal in respect to HPV [22,69].

1C) EGFR INHIBITORS Prognosis
Some studies have investigated the relation between EGFR expression and the outcome of recurrent/metastatic HNSCC patients treated with anti-EGFR antibody cetuximab [36, [70][71][72]. Even if limited to restricted series, these investigations consistently showed that EGFR levels detected by IHC have no impact on response, DCR, PFS or OS of patients treated with cetuximab.
By contrast, when EGFR expression was assessed by IHC and real time PCR, a prognostic value for high EGFR expression has been associated with reduced PFS and complete response in cetuximab+RT-treated HNSCC patients, in two studies [73,74].
However, in a phase II trial of induction CT with weekly cetuximab, paclitaxel, and carboplatin followed by chemoradiation in operable stage III/IV HNSCC, the identification of EGFR by a different approach, i.e. the quantitative AQUA method, confirmed the lack of prognostic value [75].
Last, the percentage of EGFR expression in the tumor cells was significantly associated with better OS in HNSCC patients receiving nimotuzumab in combination with CT [76], by contrast no association between EGFR expression and tumor outcome was observed in advanced HNSCC treated with nimotuzumab in combination with RT [77].

Prediction
In the EXTREME study, a trend to increased PFS and OS with cetuximab plus chemotherapy was observed for patients with higher IHC scores; however, the low number of patients does not allow any definitive conclusion [78]. The addition of cetuximab was able to counteract the dismal prognosis of the tumors with high EGFR expression showed by the group treated with chemotherapy alone.
Discordant results were obtained in Burtness' trial, in which patients were randomly assigned to receive cetuximab in combination with cisplatin or cisplatin alone [79]. Tumor samples were evaluated for EGFR cytoplasmatic expression by IHC and among the 52 patients categorized as EGFR low-to-moderate, the response rate was 41% for those treated with cisplatin plus cetuximab, compared with 12% for those treated with cisplatin and placebo (p=0.03). Thus, low-moderate EGFR expression seems to predict the response to cetuximab. However, in a logistic regression analysis of response, the interaction between EGFR and treatment group was found not to be significant. By contrast, Psyrri et al reported a correlation between EGFR expression (AQUA method) and response to cetuximab [36]. In a trial of concurrent chemoradiation ± cetuximab in advanced HNSCC, the analysis was not able to identify EGFR expression as a predictive biomarker because outcomes did not improve by adding cetuximab to RT-cisplatin and did not differ according to EGFR expression [72]. Considering treatment with nimotuzumab, a significant survival improvement was observed in EGFR positive unresectable HNSCC patients treated with this inhibitor and RT compared to control patients receiving placebo and RT [80].
It is worth mention the work by Del Campo et al. who investigated lapatinib in locally advanced HNSCC before chemo-radiotherapy. They reported that IHC EGFR overexpression seems to be predictive; however, the number of responding patients was too low to make any conclusions [81].
On the basis of the above-mentioned discordant data, EGFR expression is neither a prognostic nor a predictive factor in relation to EGFR inhibitors use.

EGFR PROTEIN ACTIVATION
Whereas many studies investigated the role of total EGFR expression, only few focused on quantitative evaluation of activated receptor (table 2), although phosphorylated EGFR (pEGFR) is thought to be a better biomarker for EGFR pathway activation. pEGFR can be assessed on frozen material by western blotting or reversephase protein and on fixed material by array fluorescence resonance energy transfer (FRET) or IHC.

2A) RADIOTHERAPY Prognosis
One recent case series investigated pEGFR expression via IHC on a population of oropharyngeal SCC treated in most cases with accelerated RT [82]. EGFR phosphorylation at residue Tyr1148 and Tyr1068 was detected in 20% and 47% of patients, respectively; no association between Tyr1148 or Tyr1068 activation and OS or DFS was observed. Interestingly, pEGFR Tyr1148 and Tyr1068 were associated with absence of HPV infection.  [73]. Notably, in the same series, intermediate/high EGFR Y1068 phosphorylation and intermediate/high EGFR expression were both independently associated with reduced PFS, thus representing potential prognostic indicators. Y1068 pEGFR was confirmed by western blotting in 17% of HNSCC patients who underwent surgery±CT where EGFR-phosphorylated tumors relapsed significantly earlier than not-phosphorylated ones [83].
Kong et al [84] investigated a large series of HNSCC via FRET, an automated technique able to inform about EGFR phosphorylation status. Of interest, EGFR activation did not correlate with EGFR expression, which in turn was not found to be prognostic.
Last, in a single isolated study, pEGFR assessed by IHC using a specific antibody targeting the phosphotyrosine site 1086 was associated with prolonged survival [38].

2C) EGFR INHIBITORS Predictivity
No predictive effect was observed for pre-treatment pEGFR level in patients with locally-advanced HNSCC receiving lapatinib before chemo-radiotherapy [81].

EGFR GENE COPY NUMBER
In the studies reported in Table 3, EGFR gene status was assessed by FISH, and tumors showing high polysomy of chromosome or EGFR gene amplification were considered to have increased EGFR gene copy number and classified as FISH positive.

3A) RADIOTHERAPY Prognosis
In a study on oral tongue SCC patients treated with preoperative RT, high EGFR gene copy number tended to be higher in patients who did not achieve a pathological complete response, even if statistical significance was not reached [23]. The lack of other data does not allow to reach any definite conclusion.

3B) SURGERY, CHEMOTHERAPY, RADIOTHERAPY Prognosis
A number of studies have investigated the association between EGFR gene status and prognosis in HNSCC patients on different primary treatments, and overall discordant results were obtained.
Interestingly, in all these studies EGFR gene copy number did not correlate either with EGFR protein expression assessed by IHC or EGFR mRNA expression levels detected by microarray or RT-PCR. The fact that EGFR copy number status is a more reliable indicator than EGFR overexpression was confirmed by the observation that patients whose tumors co-exhibited increase of EGFR gene copy number and protein overexpression presented significantly shorter OS than patients with tumors negative for FISH status and positive for EGFR protein overexpression [46]. These results indicate that the alteration of EGFR copy number may not be related only to alterations of EGFR protein expression. Indeed EGFR copy number changes frequently occur in association with chromosome 7 aneusomy that confers worse survival rates. Therefore one possibility is that additional genes on chromosome 7 that are co-amplified with EGFR may be involved in promoting a more aggressive tumor behavior. In line with this hypothesis, in a series of 35 oral SCC patients, Gebhart found that a gain of short arm on chromosome 7 was linked to a higher rate of relapse and worse OS [87]. Alternatively, an increased EGFR copy number may be a surrogate marker of chromosomal instability, which confers an adverse prognosis in many tumors [88]. However, the mechanism by which EGFR FISH status contributes to the oncogenic effect in expressing cells remains unclear.
On the other hand, a number of studies reported lack of association between EGFR gene copy number and OS [23,37,40,73,89]. Of note, most of these studies also showed the correlation between EGFR gene copy number and EGFR protein expression levels. Thus, despite some studies are available, the actual prognostic value of the EGFR gene copy number remains controversial. However, when EGFR gene copy number was assessed by FISH, the different scores used to define the FISH-positive and negative cases could explain these inconsistent findings. www.impactjournals.com/oncotarget

3C) EGFR INHIBITORS Prognosis
The results obtained in different cohorts of locallyadvanced or recurrent HNSCC patients treated with cetuximab-based regimens indicate that patients with FISH positive tumors do not show a better response compared with FISH-negative subjects [70] and that EGFR gene status is not associated with PFS or OS [70,73,90]. Furthermore, EGFR amplification was not associated with TTP or OS also in HNSCC patients treated in two phase II trials on erlotinib [91]. By contrast, EGFR FISH-positive status was associated with shorter OS in locally-advanced HNC patients receiving CT/RT and gefitinib [92]. Prediction The phase III EXTREME study demonstrated that combining cetuximab with platinum-based CT significantly prolonged OS in the first-line treatment of patients with recurrent and/or metastatic HNSCC compared with CT alone. Samples deriving from patients receiving cetuximab plus CT or CT alone were investigated by EGFR FISH to evaluate the role of increased EGFR copy number as predictive biomarker [93]. No correlation between EGFR FISH score and response rate was observed in both study arms: therefore, the results of this unique-to-date study support the lack of utility of EGFR FISH status in predicting the response to cetuximab.
Furthermore, EGFR amplification showed no predictive role in patients treated with lapatinib in locally advanced HNSCC before chemio-radiotherapy [81].

4A) SURGERY, CHEMOTHERAPY, RADIOTHERAPY
EGFR polymorphisms were uniformly assessed by DNA PCR-based methods. A recent study reported the prognostic role for EGF and EGFR polymorphisms in a cohort of locally-advanced HNSCC patients treated with post-operative CT-RT (Table 4) [94]. In more details, the 5-year OS rates of patients with EGFR R521K G/G (11.1%) and G/A (15.9%) were lower than those for the A/A (62.5%) genotype. Patients carrying one or two unfavorable alleles had worse 5-year OS than those without unfavorable allele (not available versus 20% versus 71.4%, p=0.002). Multivariate analysis revealed that the highest risk of death was associated with the coexistence of two unfavorable genotypes (hazard ratio 25.7, 95% confidence interval =3. 4-193.4; p=0.002).
The prognostic value of EGFR R521K polymorphism was further investigated in the study of Bandrés et al., where it was defined as R497K according to the older nomenclature [95]. Patients with G/G genotype of the polymorphism R521K in exon 13 showed the highest risk of disease-related mortality. In contrast, in the same patient cohort, the (CA) n repeat polymorphism in intron 1 was not associated with OS. However, this single study does not allow any firm conclusion.

4B) EGFR INHIBITORS Prognosis
Fifty-one recurrent/metastatic HNSCC patients enrolled in a single-arm, phase II study of second-line treatment with cetuximab/docetaxel were genotyped for EGFR polymorphisms R521K in exon 13 and (CA) n repeat. The results revealed that R521K G/G genotype is significantly associated with increased skin toxicity, DCR and better PFS [96]. However, this genotype had no influence on OS.
However, in HNSCC patients treated with cetuximab±CT, the presence of R521K G/G genotype was associated with longer OS. In detail, the median OS was 6.7 months in patients with at least one K allele compared with 13.3 months in patients homozygous for the wildtype R allele [97].
The CA repeat polymorphism was not associated with DCR, PFS or OS [96]. At present, the promising prognostic value of EGFR R521K G/G polymorphism in patients treated with combination of surgery, CT and RT or EGFR inhibitors deserves further investigations. The predictive role of these polymorphisms has not been investigated yet.

EGFR MUTATION
EGFR mutation is usually considered a rare event in HNSCC. However, the actual frequency of EGFR mutation is not well defined, even though analogous techniques, mainly PCR and sequencing, have been applied to investigate this issue [3,98,99].

5A) SURGERY, CHEMOTHERAPY, RADIOTHERAPY Prognosis
Hama et al [83] prospectively analyzed a population of surgically-treated HNSCC (Table 4). Among 13 patients with EGFR phosphorylation, 4 with EGFR mutation had a longer survival without recurrence than patients with EGFR wild type (p=0.023). However, the mutational analysis performed on the largest number of cases (17 mutated cases out of 108 total cases) indicated that EGFR mutation is not a significant prognostic factor [99].

5B) EGFR INHIBITORS Prognosis
A case report showed complete response in a patient with HNSCC treated with cetuximab monotherapy after initial surgery and RT. The authors identified a somatic mutations in the ligand-binding domain (P546S) and in the kinase domain (R705G) [100]. Interestingly, in vitro experiments indicated that the P546S mutation in the EGFR ligand-binding domain enhances NIH-3T3 cell line sensitivity to cetuximab compared with cells expressing wild-type EGFR; on the other hand, R705G mutation did not seem to contribute to drug sensitivity. In another study, a deletion in exon 19 of EGFR was disclosed in two out of 29 HNSCC patients treated with cetuximab and RT. These two patients presented poor clinical outcome, suggesting that this mutation may contribute to the limited response [74].
Despite this intriguing but overall weak evidence, it is rather unlikely that EGFR mutations could provide information about HNSCC response to anti EGFR therapy, mostly due to their low and variable rate in HNSCC.

6A) SURGERY, CHEMOTHERAPY , RADIOTHERAPY Prognosis
EGFRvIII is a mutant form of EGFR due to a deletion of exons 2-7 resulting in a frame deletion variant with a truncated extracellular domain exerting ligandindependent constitutive activity.
Wheeler et al. investigated the expression levels of EGFRvIII mRNA, as detected by real time RT-PCR, in HNSCC patients treated with surgery and/or CT (Table 4) [73]. In this cohort of HPV-negative tumors, intermediate or high EGFRvIII mRNA levels did not provide any prognostic information.
Another study by Szabo et al evaluated EGFRvIII cell expression by IHC in a group of HNSCC patients [38]. Multivariate analysis showed no significant correlation between the presence of EGFRvIII and patient survival.

6B) EGFR INHIBITORS Prognosis
Biopsies from patients enrolled in a single-arm phase II study investigating cetuximab plus docetaxel as second-line treatment of recurrent/metastatic HNSCC were analyzed by IHC to measure EGFRVIII expression. EGFRvIII expression was detected in 17% of cases and there was a significant association between EGFRvIII levels and treatment efficacy. In the group of patients with low EGFRvIII IHC score, DCR was 65% whereas patients with high EGFRvIII score showed a DCR of 13% (p=0.02). Multivariate logistic regression analysis confirmed the independent association of EGFRvIII with lack of response and shorter PFS [71]. Moreover, EGFRvIII level, detected by real time PCR, was associated with increased DCR but not with time to progression or OS in recurrent or metastatic HNSCC patients treated with erlotinib [91]. However, Smilek et al reported no association between RT-PCR detected EGFRvIII levels and response to cetuximab combined with RT [74].
It must be emphasized that the presence of EGFRvIII in HNSCC reported in the mentioned studies may be questioned. In fact, two studies recently analyzed a large number of tumors (638 and 531 HNSCC samples) for EGFRvIII expression using IHC with the antibody L8A4, RT-PCR and the RNA-Seq analysis [101,102]. Both studies provided a strong evidence showing that EGFRvIII is absent or very rare (frequency of 0.37%) in HNSCC. In any case, when present, EGFRvIII expression level is low in HNSCC [103] and thus difficult to detect. Cumulatively, the different methodologies applied, coupled with the intratumoral heterogeneity, may explain the discordant results and support the notion that an accurate detection of EGFRvIII needs multiple methodologies including DNA, RNA and protein assessment [103].

7A) RADIOTHERAPY Prognosis
In a group of patients with oropharyngeal cancer who underwent curative RT, the levels of the EGFR ligand TGFα assessed by IHC in pretreatment tumor biopsies was not a prognostic marker (Table 4) [21]. By contrast, in a group of 68 patients with early laryngeal cancer treated with RT, the recurrence rate was significantly higher in patients with tumor showing TGFα expression detected by IHC [19].

7B) SURGERY, CHEMOTERAPY, RADIOTHERAPY Prognosis
TGFα levels were quantified by IHC and computerized image analysis system on primary tumors from HNSCC patients surgically treated (±RT-CT). At multivariate analysis, high TGFα levels were predictors of reduced DFS and disease-related mortality [35].

7C) EGFR INHIBITORS Prognosis
Expression of the EGFR ligand amphiregulin was assessed by IHC on biopsies from recurrent/metastatic HNSCC patients treated with cetuximab plus docetaxel. A trend towards a reduction in DCR (40%) was observed (p=0.09) in patients with high amphiregulin IHC score, as compared with patients with low score (65%) [71]. In addition, patients that showed high expression of amphiregulin had significantly shorter PFS and OS. Due to the paucity of data, no firm conclusion can be made about the prognostic role of EGFR ligands.

CONCLUSION
We reviewed the literature on the clinical significance of EGFR alterations in relation to the HNSCC treatment, with the aim to identify emerging prognostic or predictive factors. Our conclusions are summarized in table 5.
The majority of the trials explored the prognostic role of EGFR alterations, while only limited data exist on the predictive role of EGFR alterations in relation to a specific treatment. At present, the predictive role of this marker has been exclusively explored in HNSCC patients treated with RT or EGFR inhibitors.
Concerning RT, high EGFR expression is a negative prognostic factor associated with poor outcome. Convincing evidence supports the robust predictive role of high EGFR expression when evaluated by a quantitative assay, in order to select patients who can most benefit of accelerated RT.
For HNSCC patients receiving multiple treatments such as surgery, CT and RT or a combination of them, no predictive biomarkers are available. However, several studies indicated that EGFR expression represents a good prognostic parameter, even if there is heterogeneity mainly due to IHC scoring system and tumor site variability. Notably, the significant association of high EGFR expression with shorter PFS and OS was reproducible only when protein expression was measured by a "quantitative" or at least semi-quantitative method. It appears advisable to estimate EGFR protein by a IHC method considering both intensity and extent of the staining, as well as stratifying patients by specific tumor sites. Nuclear EGFR and TGFα levels seem to be additional promising negative prognostic factors, as well as EGFR activation and polymorphisms, however they have not been widely validated yet. EGFR gene copy number as well as EGFR mutations has no prognostic value and do not deserve further investigation.
Within the field of EGFR inhibitors, mainly cetuximab, available studies established that both EGFR expression and an increased gene copy number are neither predictive nor prognostic biomarkers. EGFR polymorphisms and high amphiregulin levels may be promising prognostic factors and validation trials are necessary. However, it must be taken into account that the immune system plays a role in the clinical response to the EGFR inhibitor cetuximab too. Indeed, in addition to the inhibition of downstream signaling pathways, cetuximab mediates its effects also by immunogenic mechanisms such as the antibody-dependent cell mediate cytotoxicity (ADCC), the complement-mediate cytotoxicity, the modulation of the human leukocyte antigen class I and antigen-processing machinery component expression [104]. This fact should be taken into account also considering the emerging role of checkpoint inhibitors in head and neck cancer and the possible synergistic effect of cetuximab and immunotherapeutic drug combinations.
Moreover, when interpreting these data, the inverse relationship between EGFR expression or EGFR increased gene copy number and the stronger prognostic factor of HPV positivity should be taken into account. HPV-positive oropharyngeal tumors tend to present decreased EGFR expression [50,60,63,64,67,105], and increased EGFR gene copy number is restricted to HPV-negative cancer [3,65]. Although the rationale underlying those findings has not been fully elucidated yet, it is widely accepted that HPV-positive tumors have a different genetic profile compared with HPV-negative counterparts, thus contributing to the different clinical behavior and to the higher chemo-and radiosensitivity [68,105,106]. Therefore, the greater impact of EGFR deregulation in HPV negative tumors should be analyzed within the context of this different genetic pattern, and the "combined" effect of HPV status and EGFR expression on prognosis both of HPV-positive and negative or only of HPV-negative patients remains to be defined.
Lastly, it should be stressed the importance to increase the research regarding the molecular mechanisms underlying EGFR deregulation, in order to improve the HNSCC therapeutic approaches and to reduce the discrepancy sometimes existing between preclinical and clinical data.

ACKNOWLEDGMENTS
The Authors have no conflicts of interest directly relevant to this manuscript. Editorial assistance for the preparation of this manuscript was provided by Luca Giacomelli, PhD and Sara Parodi, PhD; this assistance was supported by internal funds.