The previous data show that treatment failures occur regularly in HPV+ HNSCC. Treatment failure can occur locoregionally or at distant sites. For clinical practice, it is important to look at differences in the clinical characteristics of these recurrences. This section therefore aims at identifying patterns of disease recurrence in the HPV+ population and its differences compared to HPV− cancers.

In oropharyngeal carcinomas, the study by Ang et al. (N Engl J Med. 2010) is a retrospective analysis of stage III/IV oropharyngeal carcinoma (no significant difference in survival after treatment with accelerated fractionation RTx+ cisplatin or standard fractionation radiotherapy + cisplatin). 206 (63.8 %) of 323 patients with oropharyngeal carcinoma were HPV positive (HPV ISH & p16). After 3 years, the tumor had relapsed locoregionally significantly more often in HPV-negative than in HPV-positive patients (35.1 %, 95 %CI 26.4–43.8 vs. 13.6 %, 95 %CI 8.9–18.3). The frequency of distant metastases did not differ significantly between both groups.

A retrospective analysis by Rischin et al. (J Clin Oncol. 2010) identified 106 (57 %) of 185 stage III or IV oropharyngeal carcinomas (treated with radiotherapy and cisplatin ± tiranzapine) that were p16 positive. After 2 years, locoregional failures were observed more often in the HPV-negative group (14 % vs. 7 %, p = 0.091) with similar rates of distant failure in both groups.

The study by Posner et al. (Ann Oncol. 2011) identified 56 (50 %) HPV-positive (HPV PCR) carcinomas among 111 patients with locally advanced oropharyngeal carcinomas. After 5 years, local-regional failure was significantly less common in HPV positive than in negative carcinomas but no significant difference was seen in the rate of distant metastases.

The study by Huang et al. (Oral Oncol. 2013) identified 457 p16+ patients among 624 patients with oropharyngeal cancer treated with definite radiotherapy or chemoradiation. The median follow-up was longer in p16+ patients (4.2 vs. 3.3 years). 27 (6 %) p16+ patients had locoregional failure as compared to 35 (21 %) p16− patients. Distant metastases (with or without concurrent locoregional recurrence) were identified in 54 (12 %) p16+ and 25 (15 %) p16− patients.

Taken together, these results suggest that locoregional recurrences were less common in HPV+ carcinomas, whereas the rate of distant metastases was similar.

Since the rate of distant metastases seems to be similar between HPV-positive and HPV-negative patients, it is of interest to investigate potential differences in clinical presentation between these groups.

The study by Fakhry et al. (J Clin Oncol. 2014) did not report significant differences in the anatomic site of metastasis between p16+ and p16– oropharyngeal carcinoma. Lung (p16+ vs. p16−; 72.9 % vs. 69.7 %), bone (14.6 % vs. 15.2 %), and liver (16.7 % vs. 12.1 %) were the most common sites in 81 patients with distant metastases.

Huang et al. (Int J Radiat Oncol Biol Phys. 2012) observed 36 distant metastases after a median 3.3 years of follow-up. The overall incidence did not differ significantly between p16+ and negative patients (10 % vs. 16 %). In both groups, lung, liver, and bone metastases were common sites of recurrence but in HPV+ patients metastatic spread to the skin (7 patients), intra-abdominal lymph nodes (n = 5), brain (n = 4), duodenum (n = 1), paraspinal muscle (n = 1), and axillary lymph nodes (n = 1) were observed. Multiple distant metastases were seen in 11 patients with p16+ tumors compared to 0 in p16− patients. The median time to distant metastases was also significantly longer in p16+ cancer patients (1.6 years vs. 0.5 years).

A similar pattern of metastasis has also been described by other authors. In a cohort of 11 HNSCC patients with brain metastases, 5 patients were p16+ (Bulut et al., Eur Arch Otorhinolaryngol. 2014). A later presentation of brain metastases in the p16+ subgroup was also described in this study (45.6 vs. 26.4 months). A study by Ruzevick et al. (J Neurooncol. 2013) identified 7 brain metastases from head and neck primaries. 4 of these were HPV+ and the mean time between treatment and brain metastases was 45 months.

A disseminating metastatic phenotype has been described in another study by Huang et al. (Oral Oncol. 2013). In this study, 457 p16+ and 167 p16− oropharyngeal cancer patients were followed for a median of 3.9 years. 54 and 25 distant metastases were observed in p16+ and p16- groups, respectively. Metastases to more than two organs were observed in 18 p16+ (0 p16− patients) with 11 of these exhibiting what the authors call an “explosive” character with rapid deterioration and large metastases. Oligometastatic spread to the lung was associated with a relatively indolent course in HPV+ cancers.

Taken together, these information suggest that the most common pattern of metastatic spread is similar between HPV+ and HPV− patients with lung, bone, and liver metastases. A subgroup of HPV+ patients that might be as large as 30 % presents with relevant differences. Atypical sites (brain, skin, intra-abdominal lymph nodes) and rapid clinical deterioration are of concern in these patients. A longer interval to metastasis is also of clinical relevance, because of its impact on follow-up schedules. An increased exposition to X-ray-based imaging in follow-up is of concern in these patients, since they are younger on average (Misiukiewicz et al., Clin Adv Hem Onc 2014). Taken together, the reported differences reflect a poor-prognosis subgroup in the HPV+ population. Therefore, prognostic factors identifying patients at risk for these patterns are warranted. Additional cigarette smoking might be a risk factor but the unusual timing, site, and dissemination of HPV-associated distant metastases do not differ significantly between patients with > 10 pack years or patients with 10 pack years or less (Huang et al. Int J Rad Oncol Biol Physics 2012). A stratification into high-, medium-, or low-risk groups according to p16 status, pack years and T and N stage has been proposed but did not find a significant difference in regard to distant metastases (Fakhry et al., J Clin Oncol. 2014). Circulating tumor cells and additional molecular aberrations including Bcl2 or TP53 might help in the future to identify patients at risk for poor prognosis. (Tinhofer et al., Ann Oncol. 2014; Tinhofer et al., Eur J Cancer 2016; Nichols et al. Clin Canc Res. 2010; Morris et al., JAMA Oncol. 2016).