Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer world wide, with almost 650,000 new cases and 350,000 disease related deaths annually [1]. At presentation, around half of these patients have advanced disease [2]. In this group there is a limited benefit from radiotherapy alone (5 year locoregional control 12.6-37.4%) [3]. Combined with chemotherapy, higher locoregional control rates of up to 65% can be achieved [4, 5, 6, 7, 8, 9]. However, the obvious benefit due to the addition of chemotherapy comes at the cost of higher grade III-IV toxicity. It is therefore essential to predict which patients will not benefit from chemoradiotherapy, which patients will become disease free, and in this last group, which patients would have been disease free with radiotherapy only.

Currently, clinical factors such as stage, site and tumor volume are used to predict response and select treatment [10, 11, 12, 13, 14, 15, 16, 17, 18, 19]. In the largest series analyzed so far, Knegjens et al. found tumor volume to be the most important predictor of outcome after chemoradiotherapy [20]. Like Knegjens, Chen et al. found a poorer outcome for patients with primary tumors above 30 cc [21]. However, the predictive power of clinical factors is still limited.

Apart from clinical factors, infection status with high risk Human Papilloma Virus (HPV) should be taken into account. HPV-associated tumors have a different pathogenesis, with different and less chromosomal aberrations than tumors caused by alcohol and tobacco abuse [22]. HPV-positive tumors arise more often in the oropharynx than in other sites. Patients with these tumors seem to have a better prognosis than HPV-negative patients [23, 24, 25, 26].

In recent years, gene expression profiling has been used to search for gene signatures correlating with outcome. These have the potential to provide insight into mechanisms and can monitor multiple biological processes. To date, such gene signatures as a single factor have shown prognostic potential [27, 28, 29].

Chung et al. [30, 31] found a gene expression profile containing mostly genes involved in epithelial-mesenchymal transition and NFκB pathway activation. This profile was highly prognostic for survival in two series of head and neck cancer patients treated with primary surgery with or without adjuvant therapy. This signature was subsequently validated in an independent dataset by Pramana et al. [32], who tested the signature in a series of HNSCC patients treated with combined radiation and cisplatin, with locoregional control as the endpoint. It therefore appears to be predictive in this setting, but its independence of clinical factors was not evaluated.

In this study, we further investigated whether a HPV profile (published by Slebos et al. [33]) and the Chung profile are able to add predictive power to the current prediction of local recurrence with just clinical factors.