Evolution of a Paradigm Switch in Diagnosis and Treatment of HPV-Driven Head and Neck Cancer—Striking the Balance Between Toxicity and Cure

More than a decade after the discovery of p16 immunohistochemistry (IHC) as a surrogate for human papilloma virus (HPV)-driven head and neck squamous cell carcinoma (HNSCC), p16-IHC has become a routinely evaluated biomarker to stratify oropharyngeal squamous cell carcinoma (OPSCC) into a molecularly distinct subtype with favorable clinical prognosis. Clinical trials of treatment de-escalation frequently use combinations of biomarkers (p16-IHC, HPV-RNA in situ hybridization, and amplification of HPV-DNA by PCR) to further improve molecular stratification. Implementation of these methods into clinical routine may be limited in the case of RNA by the low RNA quality of formalin-fixed paraffin-embedded tissue blocks (FFPE) or in the case of DNA by cross contamination with HPV-DNA and false PCR amplification errors. Advanced technological developments such as investigation of tumor mutational landscape (NGS), liquid-biopsies (LBx and cell-free cfDNA), and other blood-based HPV immunity surrogates (antibodies in serum) may provide novel venues to further improve diagnostic uncertainties. Moreover, the value of HPV/p16-IHC outside the oropharynx in HNSCC patients needs to be clarified. With regards to therapy, postoperative (adjuvant) or definitive (primary) radiochemotherapy constitutes cornerstones for curative treatment of HNSCC. Side effects of chemotherapy such as bone-marrow suppression could lead to radiotherapy interruption and may compromise the therapy outcome. Therefore, reduction of chemotherapy or its replacement with targeted anticancer agents holds the promise to further optimize the toxicity profile of systemic treatment. Modern radiotherapy gradually adapts the dose. Higher doses are administered to the visible tumor bulk and positive lymph nodes, while a lower dose is prescribed to locoregional volumes empirically suspected to be invaded by tumor cells. Further attempts for radiotherapy de-escalation may improve acute toxicities, for example, the rates for dysphagia and feeding tube requirement, or ameliorate late toxicities like tissue scars (fibrosis) or dry mouth. The main objective of current de-intensification trials is therefore to reduce acute and/or late treatment-associated toxicity while preserving the favorable clinical outcomes. Deep molecular characterization of HPV-driven HNSCC and radiotherapy interactions with the tumor immune microenvironment may be instructive for the development of next-generation de-escalation strategies.

More than a decade after the discovery of p16 immunohistochemistry (IHC) as a surrogate for human papilloma virus (HPV)-driven head and neck squamous cell carcinoma (HNSCC), p16-IHC has become a routinely evaluated biomarker to stratify oropharyngeal squamous cell carcinoma (OPSCC) into a molecularly distinct subtype with favorable clinical prognosis. Clinical trials of treatment de-escalation frequently use combinations of biomarkers (p16-IHC, HPV-RNA in situ hybridization, and amplification of HPV-DNA by PCR) to further improve molecular stratification. Implementation of these methods into clinical routine may be limited in the case of RNA by the low RNA quality of formalin-fixed paraffin-embedded tissue blocks (FFPE) or in the case of DNA by cross contamination with HPV-DNA and false PCR amplification errors. Advanced technological developments such as investigation of tumor mutational landscape (NGS), liquid-biopsies (LBx and cell-free cfDNA), and other blood-based HPV immunity surrogates (antibodies in serum) may provide novel venues to further improve diagnostic uncertainties. Moreover, the value of HPV/p16-IHC outside the oropharynx in HNSCC patients needs to be clarified. With regards to therapy, postoperative (adjuvant) or definitive (primary) radiochemotherapy constitutes cornerstones for curative treatment of HNSCC. Side effects of chemotherapy such as bone-marrow suppression could lead to radiotherapy interruption and may compromise the therapy outcome. Therefore, reduction of chemotherapy or its replacement with targeted anticancer agents holds the promise to further optimize the toxicity profile of systemic treatment. Modern radiotherapy gradually adapts the dose. Higher doses are administered to the visible tumor bulk and positive lymph nodes, while a lower dose is prescribed to locoregional volumes empirically suspected to be invaded by tumor cells. Further attempts for radiotherapy de-escalation may improve acute toxicities, for example, the rates for dysphagia and feeding tube requirement, or ameliorate late toxicities like tissue scars (fibrosis) or dry mouth. The main objective of current de-
To this day, the biological basis of the heightened sensitivity of HPV-driven OPSCC toward treatment is not completely elucidated. To which extent does the interplay between intrinsic properties of the tumor cells vs. the tumor microenvironment affect this radiosensitivity is also an active area of research. Some studies have postulated that expression of wild-type p53 (though inactivated by E6 oncoprotein) persists at low levels and is activated after radiation-induced DNA damage, resulting in cell cycle arrest and death (Kimple et al., 2013). Another study postulated that p16 overexpression leads to an increase in misrepair of DNA double-strand breaks (DSBs) because it inhibits the binding of RAD51, a factor essential for homologous recombination (Dok et al., 2014). This results in a shift toward the non-homologous end-joining pathway (NHEJ) and increased misrepair of DSBs. Cell line experiments have also implicated the cell cycle redistribution of HPVpositive vs. HPV-negative cell lines. HPV + cells lines showed an extensive cell cycle arrest in G2, which could be associated with higher radiosensitivity Rieckmann et al., 2013). Additionally, tumor hypoxia is not an inverse prognosticator in HPV + OPSCC (Lassen et al., 2010), although studies have shown no significant difference in tumor hypoxia between HPV + OPSCC and HPV-negative tumors, whether by immunohistochemical staining (Kong et al., 2009), gene signatures (Toustrup et al., 2012), or PET-scans (Mortensen et al., 2012). Finally, the tumor immune microenvironment may play a crucial role in mediating this radiosensitivity. HPV-driven OPSCCs show higher levels of tumor-infiltrating lymphocytes (TILs CD8 T cells) (Balermpas et al., 2016). Radiation therapy causes cellular damage, releasing viral and tumor antigens, which may synergistically activate the immune antitumor response.
The standard of care is based on data from trials conducted irrespective of tumor HPV status, and treatment of advanced stage HNSCC is multimodal par excellence. Non-resectable advanced stage HNSCC is treated with definitive radiochemotherapy (CRT), the standard conventional fractionation scheme being 70 Gray (Gy) in 2 Gy fractions (Fx) with concurrent cisplatin (100 mg/m2) on days 1, 22, and 43 (Pignon et al., 2009). In surgically operable disease, surgery (including reconstruction) is followed by postoperative RT up to 66 Gy (Gregoire et al., 2010). Patients with extracapsular extension (ECE) in the involved lymph nodes (LNs) or positive surgical margins (R) benefit from the addition of cisplatin (100 mg/m2) on days 1, 22, and 43 (Bernier et al., 2005;Gregoire et al., 2010).
The toxicity profile accrued per treatment modality (surgery, RT, or chemotherapy) is significant and increases whenever they are combined (summarized in Figure 1A) (Nguyen et al., 2002;Parsons et al., 2002;Pignon et al., 2009;Kelly et al., 2016). Given that patients with HPV-driven OPSCC are younger and will continue to live longer, de-escalation trials were conceived with the aim of decreasing treatment toxicity. Selection of appropriate candidates for treatment de-intensification is crucial to avoid compromising favorable survival outcomes. This article will briefly discuss the morbidity of treatment modalities in HNSCC. Then, the newest paradigms for diagnosis, risk stratification, and staging of HPV-driven OPSCC will be discussed. Finally, strategic principles behind current deescalation trials will be summarized, and data emerging from trials that have finished reporting will be discussed.

TOXICITY OF TREATMENT
Toxicity of treatment in HNSCC may be local (to the anatomical region) or systemic as a consequence of cancer burden or administration of chemotherapy. Interruptions or delays in completion of therapy are associated with worsened local control (LC) due to accelerated tumor repopulation (Bese et al., 2007).
Broadly speaking, toxicity can be conceptualized on several domains. Temporally, acute vs. late toxicities are defined as those occurring within 90 days vs. beyond 90 days of treatment completion (Trotti, 2000). Qualitatively, adverse events may be functional or emotional in nature (Trotti, 2000). Quantitatively, the landscape of toxicities (related to surgery, chemotherapy, or RT) can be graded using the Common Terminology Criteria for Adverse Events (CTCAE) (Bentzen and Trotti, 2007). Toxicities are organized according to System Organ Class (SOC) and vary in severity between grade 1 (mild, asymptomatic, and no intervention required), grade 2 (moderate, requiring minimal, local, or non-invasive intervention), grade 3 (severe or medically significant, significantly impairing Activities of Daily Living (ADL), and necessitating hospitalization), and grade 4 (lifethreatening and requiring urgent intervention). Grade 5 is death-causing toxicity (Bentzen and Trotti, 2007). Additionally, quality of life questionnaires (QoL) such as the European Organization for Research and Treatment of Cancer-Quality of Life core questionnaire (EORTC-QLQ-C30) or the head and neck-specific module (EORTC-QLQ-HN35) assess the impact of treatment on four domains: psychological, occupational, physical, and social.
The patterns of acute symptom burden have been recently described for patients receiving IMRT alone vs. concurrent CRT . Toxicities were evaluated using the MD Anderson Symptom Inventory-Head and Neck Module (MDASI-HN). For patients receiving IMRT only, in weeks 1-2, the top three most severe symptoms were fatigue, dry mouth, and drowsiness, in decreasing order of severity . During weeks 6-7, the top three most severe symptoms were problem tasting food, problems with mouth/throat mucus, and difficulty swallowing/chewing. For patients receiving concurrent CRT, there was a statistically significant increase in the overall severity of these symptoms (p < 0.001) .

Diagnosis of HPV-Driven Tumors
The first step in managing a patient presenting with a newly diagnosed OPSCC is establishing the presence of an HPV-driven tumor. A crucial distinction must be made between tumors harboring a passenger HPV infection versus those with a transcriptionally active virus. In an HPV-driven tumor, oncoproteins E6 and E7 are transcribed from the virus DNA and expressed in the tumor cells, leading to an interaction with growth regulatory proteins such as tumor suppressors TP53 (p53) and retinoblastoma (RB1), progression into the cell cycle, and acquisition of genomic instability (Münger et al., 2004;Doorbar et al., 2015).
Broadly speaking, there are two classes of HPV testing. Direct tests detect the presence of HPV DNA or RNA, whereas indirect tests establish the presence of HPV via molecular surrogates. In clinical settings, the most frequently used direct tests are performed on routine formalin-fixed paraffin-embedded (FFPE) tissue. In situ hybridization (ISH) or polymerase chain reaction (PCR) tests detect HPV DNA or RNA (Venuti and Paolini, 2012). Due to the low quality of RNA in FFPE material, detection of HPV E6 and E7 mRNA via reverse-transcriptase PCR is infrequently utilized in clinical routines. This method is favored for fresh frozen tissue (Venuti and Paolini, 2012). A promising ISH-based assay (HPV RNAscope) has shown optimal sensitivity and specificity in FFPE tissue but is still not broadly used in clinical practice (Mirghani et al., 2015b).
Most recently, Shinn et al. has retrospectively analyzed the concordance between p16-IHC and HPV-mRNA and its impact on the clinical outcome of 467 patients with oropharyngeal tumors (Shinn et al., 2021). They found a rate of 4.9% discordance between p16-IHC and HPV mRNA (3.4% p16-IHC-/HPV mRNA+ and 1.5% p16 IHC+/HPV mRNA-). Both patient groups had an inferior clinical outcome to double positives. When stratified by HPV mRNA status alone, patients who were p16 negative but HPV mRNA positive had a better outcome than their p16-positive but HPV mRNA-negative counterparts (Shinn et al., 2021).
Post hoc analysis of RTOG0129 classified patients with HPV tumors into low or intermediate risk groups based on the N-stage and pack-years of smoking: • Patients with ≤10py were categorized as the low-risk group regardless of TN staging (Ang et al., 2010). • Patients with >10py and N0-N2a nodes were also low risk, with 3-year OS rates of 93% (95%CI 88.3-97.7%). • By contrast, patients with >10py and N2b-N3 tumors were considered intermediate risk with OS rates of 70.8% (95%CI 60.7-80.8%) (Ang et al., 2010).
Additionally, clinical studies revealed that the staging system for HNSCC was not suitable for prognosticating the outcome of HPV-driven tumors as it could not discriminate hazards (Huang et al., 2015;Dahlstrom et al., 2016 Using RPA and adjusted hazard ratios (AHRs), the novel eighth AJCC staging edition was derived (O'Sullivan et al., 2016). In this staging system, p16-IHC is the test for diagnosing an HPV-driven tumor. The T stage remains largely unmodified, and the main consequence is that there are differences between clinical and pathologic N staging, as the N stage was the strongest correlate of OS (Lydiatt et al., 2018). For clinically palpable or radiographically visible disease, the main difference was location of LNs and size (≥6 cm) (Lydiatt et al., 2018). Patients with unilateral LNs smaller than 6 cm are staged cN1 and those with contralateral or bilateral LNs <6 cm are cN2 and any LN ≥ 6 cm confers a cN3 stage (Lydiatt et al., 2018). For surgically resected tumors, the number of LNs (≥5) was the main prognostic factor (Lydiatt et al., 2018). Patients with 1-4 affected LNs and ≥5 LNs were pN1 and pN2, respectively (Lydiatt et al., 2018). ECE was not a prognostic factor in HPV-driven OPSCC and, therefore, is not considered in the updated eighth AJCC staging system (Lydiatt et al., 2018).
This staging system was first developed in patients who received primary CRT and later validated in patients who received surgery followed by adjuvant therapy (Huang et al., 2015;O'Sullivan et al., 2016;Lydiatt et al., 2018).
On this basis, the eighth AJCC staging system for HPV-driven OPSCC was adopted (Lydiatt et al., 2018): Based on the eighth AJCC, 48% of patients who would have been staged as Stage III or IV according to the seventh AJCC edition migrate to stage I (Lydiatt et al., 2018). Retrospective appraisal of hazard discrimination for the eighth AJCC staging system was conducted in the National Cancer Database (NCDB) for 3,745 patients (Zhan et al., 2017), revealing 4-year OS rates of Stage I (92%), II (81%), and III (63%) (Zhan et al., 2017).
In discussing the eighth AJCC system, it is important to keep in mind that tobacco consumption is not included. Beyond the smoking history, patients with seventh AJCC stage I and stage II OPSCC were candidates for single-modality treatment with excellent outcomes. Patients with stage III-IV 7th AJCC received multimodal therapy (surgery followed by adjuvant CRT or primary CRT). The eighth AJCC staging system was developed based on survival outcomes using retrospectively collected data. The patients with stage III-IVa seventh AJCC who migrated to stage I and II 8th AJCC received more intense therapy compared to patients who were stage I-II in the seventh AJCC and migrated to stage I in the eighth AJCC system.
Consequently, several questions remain to be elucidated. Are the favorable clinical outcomes of these patients related to multimodal therapies reserved for advanced stage OPSCC? Are all stage I HPV OPSCC eligible for treatment de-intensification? Who should receive multimodal therapy? Taken together, several parameters are relevant for evaluation and interpretation of currently completed and ongoing de-escalation trials. First, is the deescalation arm compared with a "standard of care" arm? What is the primary endpoint? Is the study statistically powered to detect differences in clinical outcomes? Which risk group is this trial targeting (low versus intermediate risk)? How is HPV diagnosis defined? How is the response monitored (clinical/radiographic vs. pathological)? In trials of surgery and adjuvant therapy, what constitutes a negative margin? Finally, questions of costeffectiveness should be kept in mind when evaluating these trials.

PRINCIPLES OF DE-ESCALATION TREATMENT
The overarching aim is the identification of appropriate treatment intensity that minimizes morbidity of cancer survivors without compromising their survival prospects, as seen in Figures 1B, 2 below.
De-escalation trials follow one or a combination of the following strategies. In the primary RT/CRT treatment setting, strategies followed include the following: • Reduction of chemotherapy toxicity by replacing cisplatin with targeted agents (e.g., anti-EGFR treatment with cetuximab) • Reduction of chemotherapy and or RT dose/volume • Omission or modification of chemotherapy dose or RT dose/volume depending on clinical or pathologic response to ICT • Omission of chemotherapy In the surgical and adjuvant treatment setting, the strategy includes reduction or omission of RT, chemotherapy, or CRT after surgery. Additionally, emerging clinical trials are evaluating the combination of immunotherapy with radiotherapy (sequential or concomitant) (NCT02764593, 2016;Spreafico et al., 2018) and the use of particle therapy with protons instead of conventional photon radiotherapy to reduce toxicity to the surrounding tissue (Gunn et al., 2016).

Combining Radiotherapy With Cetuximab
Cetuximab is a monoclonal antibody targeting the epidermal growth factor receptor (EGFR), which mediates the activation of oncogenic pathways in HNSCC. In 2006, the Bonner RCT prospectively evaluated the impact of adding cetuximab to RT in patients with advanced-stage HNSCC (Bonner et al., 2006). Compared to patients who received RT alone, there was a statistically significant survival advantage without a concomitant increase in radiation-induced toxicity (median OS 29.3 vs. 40 months, respectively) (Bonner et al., 2006). The survival advantage was strongest among patients with clinical Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 753387 5 features suggestive of HPV-driven HNSCC, namely, young patients with oropharyngeal tumors, smaller primaries, and higher nodal involvement (Bonner et al., 2006), a finding subsequently confirmed upon secondary analysis based on p16-IHC status (Rosenthal et al., 2016).
At the present date, three prospective clinical trials, RTOG1016, De-ESCALaTE HPV, and TROG12.01, have evaluated the impact of adding cetuximab to primary RT of 70 Gy. In these trials, non-inferiority of cetuximab was not achieved and cisplatin-based CRT consequently remained the standard of care in HPV-driven OPSCC treated with primary RT (Gillison et al., 2019;Mehanna et al., 2019). The findings are summarized in Table 1 below.
For all de-escalation trials, the biomarkers used for selection of HPV-driven tumors, the AJCC staging (seventh and corresponding eighth when applicable), and the smoking status of patients enrolled will be described in the adjacent tables.
Consequently, a larger trial with 114 patients (NCT02281955) was planned with 2-year PFS as the primary endpoint (Chera et al., 2019). The study was powered to detect a 2-year PFS of 87% or greater, with the alternate hypothesis that the PFS was 80% or less. Patients with eighth AJCC stage I tumors (T1-T2 and N0-N1) received standalone RT (60 Gy/2 Gy Fx), and patients with stage II-III tumors received 60 Gy RT with concurrent weekly cisplatin (30 mg/m 2 ) (Chera et al., 2019). Clinical response was assessed using positron emitted tomography (PET) and computed tomography (CT) imaging at 10-16 weeks, omitting post-treatment biopsies and selective neck dissection (Chera et al., 2019). The PET/CT complete response rate was 93 and 80% at the primary tumor site and the neck, respectively (Chera et al., 2019). The 2-year PFS and OS were 86 and 97%, respectively. 34% of patients required feeding tubes acutely, with none developing feeding tube dependence (Chera et al., 2019). There were no grade 3 or higher late adverse events reported (Chera et al., 2019). Mouth dryness was the greatest symptom burden, with no return of function to the baseline after 1 year (Chera et al., 2019).
NRG-HN002 is another trial where 316 patients, classified as Ang low risk (Ang et al., 2010) (i.e., seventh AJCC T1T2-N1N2b or T3-N0N2b, <10py), were randomized to either IMRT (60 Gy/ 2 Gy Fx) with concomitant weekly cisplatin (40 mg/m2) or accelerated standalone IMRT 60 Gy in 5 weeks (Yom et al., 2021). For either arm to progress into a phase III trial, the coprimary endpoint was a 2-year PFS rate more than the historic control of 85% and an acceptable dysphagia toxicity measured by  (Yom et al., 2021). The trial has currently advanced to phase III, where de-intensified IMRT [60 Gy/2GFx) + weekly Cisplatin (40 mg/m2)], de-intensified IMRT (60 Gy/2Fx) + nivolumab, and 70 Gy IMRT + weekly Cisplatin (40 mg/m2) will be directly compared. The co-primary endpoints are PFS and the MDADI QoL score (Yom et al., 2021). Data from these trials (summarized in Table 2 below) are in agreement with De-ESCALaTE HPV and RTOG1016 regarding the importance of concurrent cisplatin in primary CRT. Nevertheless, with a 10Gy reduction in the RT dose and 20-40% reductions of the cisplatin dose (from 300 mg/m2 to 180-240 mg/m2), clinical and functional outcomes were encouraging. The main limitation is the short follow-up duration, given that distant metastases are detected in this patient population from 2 years on after treatment (Huang et al., 2013).

Modulation of Treatment According to Response to Induction Chemotherapy (ICT)
Historically, response to cisplatin-based ICT was considered a good predictor of radiation sensitivity (Mirghani et al., 2015a). The first trial exploring ICT in HPV-driven OPSCC was ECOG 2399 (Fakhry et al., 2008) 79 , whereby patients with oropharyngeal or laryngeal tumors (seventh AJCC T2-N1N3 or T3T4-N0N3) received two cycles of induction, paclitaxel and carboplatin, followed by CRT (70Gy RT with paclitaxel) (Fakhry et al., 2008). The primary endpoint was organ preservation, defined as freedom from primary site salvage surgery or primary tumor recurrence. For the subset of patients with HPV-driven OPSCC, 2-year OS and PFS were 95% and 86%, respectively (Fakhry et al., 2008). Nevertheless, high toxicity rates were observed, with 54-53% grade 3 or worse rates of dysphagia and mucositis (Cmelak et al., 2007). 26% of patients required gastrotomy tube placement during treatment, and 17% were dependent on tube feedings at 6 months (Cmelak et al., 2007).
Therefore, ICT-based de-escalation trials utilize the principle of monitoring tumor response after ICT to guide the decision toward a decrease in RT or CRT doses (selected trials in Table 3). In ECOG 1308 (NCT01084083), patients with resecteable OPSCC (seventh AJCC T3-T4b, N0-N3) received three cycles of ICT with cisplatin, paclitaxel, and cetuximab (Marur et al., 2017). Their next treatment was selected based on their clinical response to ICT. Patients with clinical complete response (CR was assessed by clinical examination using endoscopy and CT or magnetic resonance imaging (MRI)) received de-escalated RT 54Gy with concurrent cetuximab. Partial responders received 69.3 Gy with concurrent cetuximab (Marur et al., 2017). The 2-year OS and 2-year PFS were 94% (95%CI 82-98) and 80% (95%CI 65-89) for patients who achieved a primary site CR and were treated with 54 Gy of radiation. For all evaluated patients, the 2-year OS and PFS rates were 91% (95%CI 82-96) and 78% (95% CI 67-86), respectively (Marur et al., 2017). Additionally, this trial reported significantly lower rates of difficulties swallowing solids in patients receiving 54 vs. 69 Gy (40 vs. 89%, p 0.01) and impaired nutrition (10 vs. 44%, p 0.025), as measured by the Vanderbilt Head and Neck Symptom Survey-version 2 (VHNSSv2) (Marur et al., 2017). Nonetheless, 13/80 patients (16%) had strong protocol deviations in this trial (Marur et al., 2017), and several patients had dose reduction of cisplatin (17.5%), cetuximab (22.5%), and carboplatin (2.5%), respectively, due to grade 3 or more toxicity (CTCAE) during induction, raising the question of whether addition of ICT-associated toxicity for patient selection should not be considered to assess the net benefit of treatment de-escalation (Marur et al., 2017;Mirghani and Blanchard, 2018;Wirth et al., 2019). Finally, a post hoc analysis of this trial suggested worsened outcomes for patients with >10 py of smoking (Marur et al., 2017).
The Quarterback trial was a planned prospective randomized control trial, where patients received three cycles of induction with docetaxel, cisplatin, and 5-Fluorouracil (TPF) (Misiukiewicz et al., 2019). Complete or partial responders (as monitored by PET-CT or biopsies) would be randomized to 56 Gy IMRT or 70 Gy IMRT with weekly carboplatin (Misiukiewicz et al., 2019). Non-responders would receive the standard 70 Gy CRT arm (Misiukiewicz et al., 2019). The primary endpoint was non-inferiority with 3-year PFS. The trial closed early with 23 patients enrolled (Misiukiewicz et al., 2019). Although 20 patients developed significant response to ICT and were randomized, non-inferiority could not be demonstrated (p 0.8) (Misiukiewicz et al., 2019).

De-Escalation of Adjuvant Radiochemotherapy
Stratification of patients after surgery based on their pathological results aims to identify patients who can benefit from the complete omission of postoperative radiation and chemotherapy (Kelly et al., 2016), (see Table 4).
ECOG3311 is a phase II trial where 445 patients with intermediate risk OPSCC (seventh AJCC T1T2-N1N2b p16-IHC + OPSCC) were randomized into four clinical arms based on the presence/absence of pathological risk factors after TORS resection of the primary tumor and neck dissection (NCT01898494, 2013). Patients with 0-1 LNs, no ECE, and negative margins did not receive subsequent adjuvant treatment (arm A) . Patients with R0, N2 disease, or ECE <1 mm received de-escalated RT (in one of two possible arms: Arm B 50 Gy or Arm C 60 Gy). Arm D consisted of patients with R1, >4 involved LNs, or ECE who received CRT (66Gy RT + weekly cisplatin 40mg/m2) (Ferris et al., 2020;Ferris et al., 2021) Co-primary outcomes were 2-year PFS>85%, accrual rate, grade 3-4 bleeding events during surgery, and positive resection margins (Ferris et al., 2020). The positive margin rate was 3.3 and 5.9% grade III or IV oropharyngeal bleeding . This trial also met its primary endpoint for PFS: 2-year PFS for Arms A, B, C, and D were 96.9, 94.9, 96, and 90.7%, respectively .
MC1273 is a trial which enrolled patients with intermediaterisk HPV-driven OPSCC and R0 surgeries to receive de-escalated adjuvant CRT (Ma et al., 2019). Intermediate risk criteria were defined as seventh AJCC stage III-IV and high-risk features such as ECE, lymphovascular invasion (LVI) or perineural invasion (PNI), ≥2LN, any LN > 3 cm, or ≥ T3 (Ma et al., 2019). Patients with >10py history were excluded (Ma et al., 2019). ECE was the stratifying factor whereby patients with no ECE (cohort A) received 30 Gy RT in 1.5 Gy twice daily fractions and concurrent docetaxel (Ma et al., 2019). Cohort B consisted of patients with ECE, who received 36 Gy in 1.8 Gy twice daily fractions and concurrent docetaxel (Ma et al., 2019). The primary endpoint was 2-year LC with rates of 100 and 93% in cohorts A and B, respectively (Ma et al., 2019). 2-year PFS and OS for all patients were 91.1 and 98.7%, respectively (Ma et al., 2019). No patient required a gastrotomy tube by 1 month after treatment (Ma et al., 2019). A subsequent phase III trial, DART-HPV, has been designed, where patients were randomized to RT (twice daily, 30 Gy/1.5 Gy or 36/Gy in 1.8 Gy with concomitant docetaxel) or RT 60Gy/2Gy once daily and cisplatin weekly (40 mg/m2) (NCT02908477, 2016; Ma et al., 2019) The primary endpoint was grade 3 AE preliminary results that were presented at the annual 2021 American Society of Radiation Oncology (ASTRO) meeting. Grade 3 AEs were 1.6% for the experimental arm vs. 7.1% for the standard of care (p 0.058). However, 2-year PFS was 86.9 vs. 95.8%. Particularly, patients with pN2 disease and ECE had the worst outcomes after de-escalation, with 42.9% PFS rates compared to 100% in the standard of care arm.
Taken together, the outcomes of ECOG3311 and MC1675 provide an encouraging basis for de-escalation of therapy in the adjuvant setting. However, patients with ECE + pN2 disease may not be suitable for treatment de-escalation. In MC167, the 2-year PFS was 42.9% for these patients who were in the de-escalation arm of the trial, with 77% LRC and 59.4% DMFS rates. By contrast, those patients had a 100% 2-year PFS in the standard of care arm. Similar results were seen in ECOG3311, whereby patients with either ECE or >4LNs received a standard dose CRT (66 Gy) with weekly cisplatin and had a 2-year PFS of 90.7%. This once again highlights the importance of adequate patient selection.
Further data are also awaited from two prospective trials, PATHOS and DELPHI. The PATHOS trial stratifies 242 patients with OPSCC seventh AJCC T1T3-N0N2b disease into four arms as well, depending on pathological results after TORS (Owadally et al., 2015). Patients with no risk factors go into an observational alone arm, patients with intermediate risk factors (seventh AJCC T3 stage, pN2a-N2b, LVI, pNI, or close margins) receive RT only (50 Gy or 60 Gy), high-risk patients with R1 resections receive 60 Gy RT, and patients with ECE receive CRT 60 Gy with cisplatin (100 mg/m2) (Owadally et al., 2015). Co-primary outcomes are 1-year MDADI and 1-year OS (Owadally et al., 2015). The ongoing DELPHI trial (NCT03396718, 2018) aims to enroll 384 patients into two clinical arms of radiation dose deescalation based on pathological risk factors. In the first level of the DELPHI trial, patients with intermediate risk (pT3, R0 margins, ≤ involved LNs, and no ECE) receive a 10% reduction of standalone RT (54 Gy to the tumor bed) and 45 Gy to the cervical LNs. Patients with at least one high-risk feature (R1 status, pT4, ≥4LNs, or ECE) will receive 59.4 Gy for the tumor bed, 45Gy to the cervical LNs, and additional chemotherapy. The primary endpoint is 2-year locoregional recurrence. If no more than three tumor recurrences in 30 patients occur in the first 2 years, further de-escalation of the RT dose will ensue, whereby patients with no high-risk features will receive 48.4 and 39.6 Gy to the tumor bed and cervical LNs, respectively. Patients with high-risk features will receive 55 and 39.6 Gy to the tumor bed and cervical LNs, respectively (NCT03396718, 2018).
The ADEPT trial (NCT01687413, 2012) was comparing 60 Gy RT alone vs. 60 Gy RT + weekly cisplatin 40 mg/m2 in stage I-III HPV-driven OPSCC with ECE. However, this trial has terminated due to slow accrual.

Comparison of Primary Radio(Chemotherapy) Versus Postoperative Adjuvant Radio(Chemo) Therapy
Between 2004 and 2013, the percentage of patients with T1-T2 OPSCC undergoing surgery increased from 56 to 82% in the US, with a meta-analysis suggesting decreased toxicity associated with surgery compared to definitive CRT (Nichols et al., 2019). Although not a de-escalation trial, ORATOR evaluated QoL outcomes in primary RT/CRT vs. surgical intervention (Nichols et al., 2019) (Supplementary Table S1). In the surgical arm, 34 patients underwent TORS + neck dissection (ND), with 47% receiving adjuvant RT up to 64 Gy and 23.5% receiving CRT (RT + cisplatin 100 mg/m2 every 3 weeks). In the primary arm, 26.5% received RT up to 70 Gy and 67.6% received CRT. The primary outcome was powered to detect a 10-point difference in MDADI total mean scores at 1 year (higher is better). 1-year scores were 86.9 in the primary CRT arm vs. 80.1 in the surgical arm (p 0.042). Grade 2 or higher adverse event rates (CTCAE) were similar in both arms, with preponderance for oral bleeding and trismus in the surgical arm and for neutropenia, hearing loss, tinnitus, and constipation in the primary CRT arm. TORS and ND were not associated with a superior QoL, and 3-year OS and PFS were 93 and 93.1%, respectively, with no differences between both arms (p 0.89 and p 0.63) (Nichols et al., 2019).
A follow-up prospective trial, ORATOR2, was planned to randomize patients to de-escalated CRT vs. de-escalating adjuvant treatment, the primary outcome being 2-year OS (Nichols et al., 2020). The results of a direct comparison between TORS and definitive CRT were eagerly awaited. Unfortunately, the trial was terminated due to unacceptable toxicity in the TORS + ND arm (two treatment-related deaths) (Palma et al., 2021), establishing primary CRT as a safe approach for treatment de-escalation. 61 patients were randomized in total. The 2-year OS was 100% for the RT arm vs. 83.5% in the TORS + ND arm (Palma et al., 2021). The findings of E3311, MC1765, and ORATOR are encouraging. Nonetheless, the data from ORATOR2 suggest that further studies will be needed to answer the question of surgery versus primary radiochemotherapy.

De-Intensification Schemes Using Immune Therapy
An emerging strategy is the combination of primary or adjuvant RT with modulators of the immune response, predominantly immune-checkpoint blockers (ICB, Supplementary Table S2).
HCC 18-034 (NCT03715946, 2018) is evaluating the addition of postoperative adjuvant reduced dose, moderately accelerated RT (45 or 50 Gy, in daily dose fx, six fx per week), and nivolumab (monoclonal antibody against Programmed cell Death protein 1 (PD-1)) in patients with advanced stage p16-IHC + OPSCC (seventh AJCC: T0, T3 + >2Nb and <10py or T0, T3 with >N1 and >10py) with intermediate risk features (ECE or positive margins). Nivolumab will be administered in two doses of 240 mg/m2 during weeks two and four of RT, and up to six doses afterward, of 480 mg/m2. The primary outcome is PFS at 3 years and gastrotomy tube dependence at 1 year (NCT03715946, 2018).
NCT03799445 will evaluate the impact of upfront dual ICB with nivolumab and ipilimumab [monoclonal antibody against cytotoxic T-lymphocyte-associated protein 4 (CTLA4)], followed by RT (50-66 Gy) in patients with the eighth AJCC (stage I-II). In this trial, patients' tumors must test positive for both p16-IHC and HPVDNA or RNA by ISH (NCT03799445, 2019). Primary endpoints include dose-limiting toxicity [DLT, defined as any ≥ grade III toxicity (CTCAE) related to immunotherapy not resolving within 28 days after treatment], complete response rate at 6 months, and 2-year PFS (NCT03799445, 2019).
The results of these trials are eagerly awaited, in light of the negative results from the Javelin Head and Neck 100 (.NCT02952586) and GORTEC 2017-01 (REACH, NCT02999087.) phase III trials. Javelin Head and Neck 100 compared the combination of avelumab, a PD-L1 inhibitor, + standard-of-care CRT (RT70Gy/2Gy + cisplatin 100 mg/m 2 every 3 weeks) against standalone CRT in locally advanced HNSCC (Lee et al., 2021). The trial included HPV-positive and HPVnegative disease. The primary endpoint of PFS prolongation was not met, and there was no benefit seen upon stratification in HPV-positive disease (Lee et al., 2021). Similarly, in the GORTEC 2017-01-Reach trial, the combination of avelumab and cetuximab-based chemoradiotherapy did not improve PFS, further cementing the role of cisplatin-based radiochemotherapy as the standard of care in the treatment of locally advanced HNSCC .
Beyond ICB combinations, exploration of radiotherapyinduced immune activation and unmasking of HPV-associated neoepitope may, together with the growing arsenal of immunoncology (IO) drugs, facilitate the development of effective antitumor specific vaccines.

MOLECULAR STRATIFICATION OF HPV-DRIVEN OPSCC
Beyond p16-IHC, direct HPV testing, and tobacco smoking, deescalation trials can also contribute to a deeper understanding of the biology of HPV-driven OPSCC.
In one approach, PET imaging (18F-MISO PET detecting tumoral hypoxia) was used in a pilot study from the Memorial Sloan Kettering Cancer Center (MSKCC) to modulate the RT dose to the LNs in patients with p16-IHC + OPSCC receiving CRT (Lee et al., 2016). Patients with no baseline tumor hypoxia or with resolution of hypoxia after week 1 (per 18F-MISO PET scans) were candidates for 10-Gy dose de-escalation to the LNs (Lee et al., 2016). The primary tumor site received the standard RT dose (70 Gy). 10 patients (30%) were eligible for dose deescalation (Lee et al., 2016). The 2-year clinical outcomes were as follows: 100% LC, 97% DM, and 100% for OS (Lee et al., 2016).
Additionally, targeting hypoxic HPV-driven tumors with heavy charged particles that are less dependent on the oxygen enhancement ratio (OER) (Klein et al., 2017;Chiblak et al., 2019) may provide another attractive venue to specifically escalate the dose while sparing normal tissue in this subgroup.
The association between the mutational landscape of HPV-driven OPSCC and patient outcomes is still under investigation. Beaty et al. performed next generation sequencing (NGS) of tumor samples from 78 patients enrolled in de-escalation trials of primary RT to investigate the prognostic role of PIK3CA mutations (Beaty B. T. et al., 2019). PIK3CA was the most significantly mutated gene in 21.8% of patients (Beaty B. T. et al., 2019;. Patients with mutated PIK3CA had significantly lower 3-year DFS (65%) compared to patients with wild-type PIK3CA (93%, p 0.0009), suggesting that this patient population is not suitable for deescalation trials (Beaty B. T. et al., 2019;. However, conflicting data emerged from studies of patients with metastastic HPV-driven OPSCC where mutations in the PI3K pathway (PI3KCA, PIK3CA2B, and PIK3R1) were associated with an improved overall survival outcome at 5 years (Hanna et al., 2018).
Another biomarker trial approach is based on monitoring of circulating free DNA (cfDNA) or circulating tumor HPVDNA (ctHPVDNA) detected in patients' blood. In NCT0316182, 115 patients were prospectively followed up for a median duration of 23 months after being treated with curative intent chemoradiotherapy (Chera et al., 2020). The trial estimated the positive predictive value (PPV) and negative predictive value (NPV) of ctHPV-DNA for determining disease recurrence (Chera et al., 2020). Undetectable levels of ctHPVDNA at post-treatment time points had an NPV of 100%. Conversely, two consecutively positive ctHPVDNA blood tests had a PPV of 94%. The median time from ctHPVDNA positivity to biopsy-proven recurrence was 3.9 months (Chera et al., 2020). cfHPV-DNA was confirmed as a highly specific biomarker of surveillance in a recent metaanalysis of 11 studies (Hanna et al., 2018). cfHPV-DNA had a pooled sensitivity of 0.81 (95% CI 0.78-0.84) and 0.98 (95%CI 0.96-0.99) at the first diagnosis. At follow-up, it had a sensitivity of 0.73 (95%CI 0.57-0.86) and a specificity of 1 (Hanna et al., 2018). Interestingly, one study found a significant association between levels of cfHPV-DNA and N status, as well as the extent of disease involvement (Tanaka et al., 2022). Levels of cfHPV-DNA increased as function of involvement, with the lowest in locally advanced disease, followed by locoregional spread, and the highest for distant metastases (Tanaka et al., 2022).
Similarly, HPVDNA may be detected from oral rinses. Patients with persistent oral HPVDNA after the end of therapy had a decreased 2-year OS (HR 1.86, p 0.003) compared to patients without detectable DNA in a prospective phase II clinical trial (Fakhry et al., 2019). Finally, the association between seropositivity to HPV16 antigens and clinical outcomes has been demonstrated in several studies (Dahlstrom et al., 2015;Nelson et al., 2017). Furthermore, a recent study investigated differential patterns of antibody response to cancer antigens in HPV-driven versus HPV-negative HNSCC: antibodies against IMP-1 (found in n 9/153, 6% of patients) were adversely prognostic only in HPV-driven OPSCC (HR 3.28, p < 0.001) (Laban et al., 2019). Detecting relevant tumor immune microenvironment (TIME) parameters may also assist in stratifying risk for recurrence and inferior OS in HPV-driven HNSCC. A multicentric retrospective study from the German Cancer Consortium (DKTK) identified enrichment in CD8 + infiltrating immune cells as an independent prognostic factor both in p16-IHC+/HPVDNA + OPSCC tumors and HPVnegative tumors, in a cohort of 161 patients treated with surgery and postoperative CRT (Balermpas et al., 2016). Disappointingly, high levels of PD-L1 were not prognostic in the Javelin 100 Head and Neck trial, although this finding was not stratified by HPV status (Lee et al., 2021).
Taken together, these findings reaffirm the importance of validating all biomarkers in prospective phase III clinical trials.

CONCLUSION
The identification of HPV as a protoypic predictive marker for molecular stratification of patients has paved the way for the Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 753387 development of several avenues of treatment de-intensification. Individualized therapy may be tailored by de-escalation or adaption of local radiotherapy and de-intensification/ replacement of systemic therapy. This field is evolving at a rapid pace. In this dynamic era, deeper understanding of the biology of HPV-driven tumors shapes physicians' approach toward improved diagnosis, staging, risk stratification, and management of this disease. With a steadily increasing complexity and a plethora of opportunities, a consensus is needed to assure better comparability, for example, homogenizing inclusion criteria, exclusion criteria, risk stratification (most noteworthy, the impact of smoking), staging, and diagnosis of an HPV-driven tumor, replacing the single p16-immunohistochemistry test with a combination of direct HPV tests or including more advanced molecular methods that better assist in stratifying patients at low risk for locoregional or distant recurrence.
In postoperative adjuvant treatment of OPSCC, the design of current trials accounts for extracapsular extension, although it was not included as an adverse prognosis factor in the eighth AJCC staging system. In primary definitive CRT treatment, data from phase III trials, where substitution of cisplatin with cetuximab leads to inferior survival outcomes, have established cisplatin-based radiochemotherapy as the standard of care. Similarly, results from NRG-HN002 consolidated the role of cisplatin, whereby a higher rate of locoregional recurrence was found in patients treated with moderately accelerated radiotherapy alone (60 Gy in 5 weeks) and complete omission of cisplatin. Nevertheless, numerous other strategies are ongoing, with promising data emerging from phase II clinical trials. Confirmation in randomized phase III clinical trials is awaited. Maturity of follow-up will be an issue to address in these trials, given the main pattern of distant relapse after 2 years in HPVdriven OPSCC.
Multiparametric tumor characterization may be needed for accurate patient selection to avoid de-escalating patients at high risk of recurrence. In parallel, broadening the therapeutic window with targeted tumor-specific agents, a growing immunoncology arsenal, and novel radiation dose/quality painting via heavy charged particles may navigate the therapy of HPV-driven HNSCC toward high-precision oncology.

AUTHOR CONTRIBUTIONS
Conceptualization, BT and AA; writing-original draft preparation, BT; writing-review and editing, JD and AA. All authors have read and agreed to the published version of the manuscript.

ACKNOWLEDGMENTS
We would like to thank Katrin Rein and Yara Falakha for their editorial assistance and help with the preparation of the graphical presentations.