The molecular tumor progression model was initially proposed by Fearon and Vogelstein.³ This model states that
tumors progress by activating oncogenes and by silencing tumor suppressor genes (TSGs), each producing a growth advantage for a clonal population of cells, and that specific genetic events usually occur in a distinct order (multistep carcinogenesis) that is not necessarily the same for each tumor. For head and neck carcinomas, Califano et al.⁴ described a preliminary tumor progression model using allelic loss or imbalance as a molecular marker for oncogene amplification or TSG inactivation. They identified p16 (9p21), p53 (17p), and Rb (13q) as candidate TSGs, and cyclin D1 (11q13) as a candidate protooncogene. The results of this work support the initial observations of the colorectal molecular progression model, in that clonal genetic changes occur early in the histopathologic continuum of tumor progression. About one third of histopathologically benign squamous hyperplasias already consist of a clonal population of cells with shared genetic anomalies characterizing head and neck cancer. Identification of such early events facilitates discovery of genetic alterations associated with further transformation and aggressive clinical behavior.

The introduction of newer molecular assay techniques has greatly increased the ability to detect genetic changes and thereby improve the understanding of cancer biology in general. An overview by Ha et al.⁵ summarizes recent findings on genetic alterations in HNSCC grouped by assay techniques, such as comparative genomic hybridization, in situ hybridization, single nucleotide polymorphism, and microarray technology, and provides excellent illustrations of the complexity of HNSCC and how that complexity will require much more research to reveal the full picture. With further validation, however, this knowledge will contribute a great deal to the development of screening strategies focusing on the earlier steps of genetic alterations required to generate an invasive tumor phenotype and to the conception of early pharmacologic or genetic therapy approaches.

Tobacco and alcohol exposure have long been recognized as the dominant risk factors for HNSCC. Other risk factors include low fruit and vegetable consumption and betel quid chewing. In an overview, Petti⁶ estimated that, worldwide, 25% of HNSCCs are attributable to tobacco use, 7% to 19% to alcohol consumption, 10% to 15% to dietary deficiency, and, in regions of prevalence, >50% to betel quid chewing. Carcinogenicity is dose-dependent and magnified by exposures to multiple carcinogens.

Although tobacco and alcohol consumption is estimated to account for approximately three fourths of oral and pharyngeal carcinomas in the United States,⁷ neoplasms develop in only a small fraction of exposed individuals. This intriguing information raised the notion of the contribution of genetic susceptibility or predisposition and other cofactors (for examples of cofactors, see “Viral Etiology” section) to carcinogenesis. The potential pathways are thought to include genetic polymorphisms influencing environmental carcinogen absorption and detoxification, individual sensitivity to carcinogen-induced genotypic alterations, and so on. These ideas can now be tested more comprehensively because of recent progress in molecular biology concepts and assay methodology. For example, the ability to identify smokers at high risk for developing cancer will have important practical clinical implications in selecting individuals for more aggressive screening programs or for enrollment into intensive chemoprevention trials.

The distinction between prognostic and predictive biomarkers has not been widely appreciated. Therefore, until recently, these terms have been used rather loosely and interchangeably. Figure 1.1 illustrates the concept and definition for different classes of markers. The rates and extent of separation among the curves will vary with the disease type and stage and the efficacy of therapy, but the general principles and the relative ranking are applicable. In panel A, marker X represents an aggressive tumor feature, the presence of which is associated with poorer survival rate after both treatment (Rx) regimens 1 and 2, though Rx 2 is more effective than Rx 1. Marker Y (panel B), on the other hand, exemplifies a predictive marker for response to Rx 2. Hence, its presence is associated with better survival after Rx 2 (solid brown curve). Panel C illustrates that some markers could
have both prognostic and predictive values. The absence of marker Z is associated with better prognosis (solid and dotted black curves vs. dotted purple curve). However, since this marker also predicts response to Rx 2, its presence is associated with a better outcome after Rx 2 (solid purple curve) relative to Z+ after Rx 1 (dotted purple curve) and Z- after Rx 2 (solid black curve).

Figure 1.1 shows that carefully designed clinical trials incorporating patient stratification according to biomarkers and randomizing patients to received distinct therapy modalities are needed to yield a conclusive answer as to whether a marker is prognostic, predictive, both, or neither.

Three potent prognostic biomarkers for HNSCC have emerged in recent years. Two of these biomarkers are related to virus-associated head and neck carcinomas: circulating EBV titer for NPC and the presence of the HPV genome or its surrogate marker, p16, for OPSCC. The third biomarker, epidermal growth factor receptor (EGFR), seems to be more applicable for other HNSCCs.

The association between EBV and NPC was summarized in a previous section. Lo et al.⁴⁵ have developed a real-time quantitative polymerase chain reaction assay for measuring circulating levels of tumor-derived EBV DNA in the serum or plasma of patients with NPC. They found in a longitudinal follow-up of 17 patients that elevations in serum EBV DNA titer could be detected as early as 6 months before clinical manifestation of recurrence, whereas the titer stayed low or undetectable in patients who remained in remission. A subsequent study of patients treated with radiation, with or without chemotherapy, at the same center⁴⁶ showed that having a pretreatment EBV DNA titer exceeding 4,000 copies per mL was associated with a 2.5-fold higher risk of NPC recurrence. More interestingly, having a high posttreatment EBV DNA titer was found to be an even stronger marker for poor overall outcome, that is, a 11.9-fold increase in recurrence rate. Conversely, having a posttreatment titer of <500 copies per mL was associated with favorable overall survival and relapse-free survival rates (Fig. 1.2) and also correlated with low relapse rate. Similar results were reported by Lin et al.⁴⁷ in a series of patients treated with weekly neoadjuvant chemotherapy (cisplatin alternating with fluorouracil for a total of 10 doses) followed by radiotherapy.

The combination of intensity-modulated radiotherapy (IMRT), as discussed in the section “High-Precision Radiotherapy” below, with concurrent cisplatin has yielded local-regional control rates of around 90% even among patients presenting with locally advanced NPC. Consequently, distant metastasis has become the main pattern of relapse for this neoplasm. Therefore, plans are underway to select high-risk populations, consisting of those with persistent circulating EBV DNA after receiving the combination of IMRT with cisplatin, for testing the efficacy of combinations of novel agents with conventional chemotherapy for eliminating occult metastatic disease.

As noted earlier in this chapter, the incidence of HPV-associated OPSCC is increasing, particularly in the Western world, and several retrospective case series have shown that patients with HPV-positive OPSCC treated with contemporary single or combined therapy modalities have a better prognosis than do patients with HPV-negative OPSCC. However, owing to small sample sizes, other favorable prognostic factors associated with tumor HPV status (e.g., earlier tumor stage, young age) cannot be excluded as potential explanations for the observed
difference in survival. Therefore, a thorough correlative study was undertaken to quantify the magnitude of impact of tumor HPV status on tumor outcome in patients enrolled in a large phase III trial of the Radiation Therapy Oncology Group (RTOG 0129) treated with a combination of radiation with concurrent cisplatin.⁴³ Patients with locally advanced HNSCC were stratified according to tumor site (larynx vs. other), nodal status (N0 vs. N1-N2b vs. N2c-N3), and Zubrod performance status (0 vs. 1) and assigned to receive either accelerated fractionation with a concomitant boost (72 Gy in 42 fractions over 6 weeks) or standard fractionation (70 Gy in 35 fractions over 7 weeks) regimens. Chemotherapy consisted of intravenous cisplatin at a dose of 100 mg/m² on days 1 and 22 in the accelerated fractionation group or on days 1, 22, and 43 in the standard-fractionation group.

Of the 743 patients enrolled, 60% had OPSCC. Pretreatment biopsy specimens from the patients with OPSCC were evaluated for HPV-16 DNA by using in situ hybridization, and HPV-16-negative tumors were further assayed for 12 additional oncogenic HPV types (types 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68). Tumor expression of the cyclin-dependent kinase inhibitor p16, induced as a consequence of pRb inactivation by viral oncoprotein E7,⁴⁸ was evaluated by immunohistochemical staining with a mouse monoclonal antibody. Strong agreement between tumor HPV status as determined by in situ hybridization and p16 expression was observed, but some discrepancies were noted as well. Overall, 64% of OPSCCs were found to be positive for HPV-DNA and 68% were positive for p16. Figure 1.3 shows the differences in overall survival according to HPV or p16 status. Because the sensitivity for detecting non-HPV-16 types, expected to account for 5% to 10% of HPV-positive OPSCC, was not well known at the time of the study, misclassification of HPV-positive tumors as HPV-negative tumors likely explains the slightly larger reduction in risk of death when p16 expression was used in the analysis. A strength of the p16 assay is that it is not HPV-type specific and is therefore an excellent surrogate for tumor HPV status. Analysis of patterns of failure among these groups showed that the local-regional failure rate at 3 years was significantly lower for patients with HPV-positive OPSCC (13.6% vs. 35.1%, P < 0.001) but rates for distant metastasis were not (8.7% vs. 14.6%, P = 0.23). Also, the cumulative incidence of second primary tumors (SPTs) was significantly lower for patients with HPV-positive OPSCC (3-year rates 5.9% vs. 14.6%, P = 0.02), largely because of lower rates of smokingrelated cancers in that group.

Tobacco smoking was also found to be independently associated with overall survival and progression-free survival, both in patients with OPSCC and in the entire study population. Risk of death, for example, increased significantly by 1% per each pack-year increase in tobacco use, and magnitudes of effect were similar for patients with HPV-positive OPSCCs (hazard ratio [HR] 1.01, 95% confidence interval [CI] 1.00 to 1.02) and HPV-negative OPSCCs (HR 1.01, 95% CI 1.00 to 1.03).

The RTOG 0129 study also showed that HPV-associated OPSCCs were more common among people who had never smoked or had sporadically smoked and were also significantly associated with several favorable prognostic factors, including younger age, white race, better performance status, absence of anemia, and smaller primary tumors. In multivariate analysis, age, race, performance status, tumor classification, nodal classification, and tobacco pack-years were also significant determinants of overall survival. When the unadjusted hazard ratios (HR 0.38, 95% CI 0.26 to 0.55) were compared with the adjusted hazard ratios for HPV (HR 0.42,
95% CI 0.27 to 0.66), factors other than HPV were estimated to account for about 9% of the difference in overall survival between patients with HPV-positive and HPV-negative OPSCCs (Fig. 1.3A).

Recursive partitioning analysis indicated that tumor HPV status was the major determinant of overall survival, followed by tobacco smoking (≤10 vs. >10 pack-years) and then nodal category (N0-2a vs. N2b-3) for patients with HPV-positive OPSCCs and primary tumor category (T2-3 vs. T4) for patients with HPV-negative OPSCCs (Fig. 1.4A). Recursive partitioning led to the classification of patients with OPSCC into three risk groups: low risk (reference group, with a 3-year overall survival rate of 93%), intermediate risk (HR 3.54, 95% CI 1.91 to 6.57; 3-year overall survival rate of 70.8%), and high risk (HR 7.16, 95% CI 3.97 to 12.93; 3-year overall survival rate of 46.2%) of death (Fig. 1.4B). Patients with HPV-positive OPSCC were generally at low risk except for those who smoked or had N2b-3 nodes, which put them at intermediate risk. Patients with HPV-negative OPSCC were generally at high risk, but those having no history of tobacco use and having T2-3 tumors were at intermediate risk. Based on these findings, several trials are being designed that focus specifically on patients with HPV-related OPSCC.

EGFR is composed of four extracellular domains (I to IV), including the ligand-binding regions (domains I and III), a hydrophobic transmembrane domain, a juxtamembrane domain, an intracellular protein tyrosine-kinase domain containing the ATP binding pockets, and a regulatory carboxyl terminal domain. It is monomeric in the absence of ligands. Binding of a ligand to the extracellular domains I and III alters the spatial configuration of these domains, creating an extended and stabilized conformation that promotes homodimerization and heterodimerization⁴⁹ and activates signal transduction.

The EGFR signaling pathway has evoked considerable attention as a potential biomarker for radiation response. EGFR is overexpressed in many neoplasms, for example, in 80% to 100% of HNSCCs, and perturbation of EGFR signaling is regarded as a major cause of malignant transformation and progression.^50,51 An extensive correlative biomarker analysis using tumor biopsy specimens from patients with locally advanced HNSCC enrolled in a phase III trial of conventionally fractionated radiation (70 Gy in 2-Gy fractions, five times a week) showed no correlation between EGFR expression and T or N classification, American Joint Committee on Cancer (AJCC) disease stage grouping, and recursive partitioning analysis classes⁵² (r: -0.07 to +0.17).

As shown in Figure 1.5, EGFR overexpression, defined in terms of some level above the median mean optical density or a staining index measured using an image-analysis-based immunohistochemical assay, was found to be a strong and independent marker for higher local-regional relapse rate (68% vs. 50% at 5 years, P = 0.0031) and inferior overall survival rate (20% vs. 40%, P = 0.006) but not for the incidence of metastasis.⁵³ This finding was supported by those from another recent study using automated quantitative assessment of EGFR expression in a tissue microarray from a cohort of patients with oropharyngeal cancer treated with radiation alone, postoperative radiation, or chemoradiation.⁵⁴ The investigators revealed a strong correlation between EGFR expression, defined as above median level, and worse local recurrence rate (58% vs. 17%, P < 0.01) and worse disease-free survival rate (19% vs. 43%, P = 0.0016).

Two European studies, also conducted using specimens from patients with HNSCC enrolled in major radiation
fractionation trials, showed that high EGFR expression was predictive for increased tumor response in patients treated with accelerated radiation, which suggests that EGFR is functionally related to accelerated tumor cell repopulation during fractionated radiation.^55,56 In one of these studies, the HR for local-regional recurrence after conventional versus hyperfractionated and accelerated radiation was estimated to be 1.8 in the group with an EGFR index (i.e., the proportion of EGFR-positive tumor cells) above the median value (2P = 0.010, 95% CI 1.14 to 2.8). These data indicate that EGFR inhibitors could be especially beneficial in combination with hyperfractionated and accelerated radiation regimens.