Level V contouring
Many studies have suggested the potential oversight of the deep level V space [
64,
65,
69,
70]. The current atlas defines the posterior border of level V at the anterior border of trapezius, and it does not describe the region posterior to it (i.e., the region between the trapezius muscle and the levator scapulae). The incidence of involvement in the deep level V space ranged from 1.3% to 5% [
65,
69,
70]. Involvement of deep level V was associated with nodal disease at levels IVA [
69], Va, Vb, and Vc [
69,
70], and it was observed in 13.3% cases with level VB LNs [
64]. Consequently, adjustment of the posterior border of level VB should be considered in selected high-risk patients[
64,
69].
Accumulating evidence suggests that tailoring the nodal target volume could optimize coverage while minimizing unnecessary radiation. However, it is vital to acknowledge the inherent limitations of observational and retrospective studies. This underscores the need for careful consideration before incorporating any changes into practice.
Tailoring the RT dose and volume based on tumor response following IC holds potential for volume reduction. For locally advanced NPC, the current preferred treatment sequence involves IC followed by CCRT[
9,
71,
72] . Approximately 2-11% of patients achieved a complete response (CR), while 77-84.5% had a partial response (PR) after IC. [
6,
7,
18,
44] Furthermore, the reduction in volume of the primary tumor appeared to level off after two cycles of IC, whereas that of the affected LNs continued to decrease after the third cycle. A retrospective study documented volume reductions of 12.0%, 23%, and 20% in the NP tumor, 26%, 44%, and 42% in the RP LN, and 25%, 43%, and 55% in the cervical LN, following each successive cycle of IC [
73].
Xiang et al. [
74] reported the long-term outcome of 212 patients who were randomly assigned to receive RT using either pre-IC or post-IC volumes. In the post-IC arm, patients were administered 70 Gy in 33 fractions to the post-IC GTV
NP and 64 Gy in 33 fractions to the pre-IC GTV
NP. With a median follow-up period of 98 months, the 5-year estimated survivals in the pre-IC and post-IC arms were comparable. Of note, the locoregional recurrence-free survival (LR-RFS) were 90.2% and 93.5% in the pre-IC and post-IC arms respectively. All local recurrences in the post-IC arm were in-field. Importantly, patients who underwent volume reduction experienced fewer instances of xerostomia and hearing loss, along with an improved quality-of-life. However, it is worth noting that this study was conducted in a non-endemic region and predominantly included WHO type II NPC cases (79%).
In the endemic region, a phase II study [
75] involving 112 patients implemented a treatment approach that delivered 68 Gy in 30 fractions to the post-IC volume and 60 Gy in 30 fractions to the pre-IC volume, following 2 cycles of IC. The study's outcomes revealed remarkable 10-year LR-RFS, DMFS, and OS at 89.0%, 83.3%, and 75.9%, respectively. Notably, akin to the findings from Xiang’s study, all instances of local recurrence were in-field.
The contouring methods employed in the previous two reports had several similarities:
1. The pre-IC volume of GTVNP was considered high-risk and was encompassed within the intermediate dose CTV (treated with a dose of at least 60Gy).
2. The pre-IC skull base or bony invasion were included within the post-IC volume and received the full prescription dose.
3. GTV of cervical LN was defined using the post-IC volume.
It is well-documented that tumor response to IC, including tumor shrinkage and EBVDNA clearance, is prognostic for outcomes. [
76,
77,
78] This response reflects the tumor's biological behavior and inherent chemo-sensitivity, making it an important biomarker for treatment individualization. De-escalation strategies following IC include efforts to reduce concurrent chemotherapy intensity [
79]. In addition, a retrospective study has suggested that IMRT alone may suffice in a subset of patients who achieved CR/PR after IC. [
80] The omission of concurrent chemotherapy is currently being investigated in several phase III trials (ClinicalTrials.gov Identifiers: NCT05674305, NCT05527470, and NCT03015727).
Regarding RT, a single-arm phase II trial [
81] treated low-risk stage III patients (defined as EBV DNA <4000 copies/ml) with 2 cycles of IC followed by 60Gy CCRT for those who achieved CR or PR with an undetectable EBV DNA. This study showed promising 2-year PFS and locoregional relapse-free survival at 94% and 95%, respectively. However, this study was only published in abstract form after a median follow-up of 25.8months. A full manuscript with updated analysis is eagerly awaited. Two ongoing randomized controlled trial are underway comparing reduced doses of either 60Gy or 63.6 Gy in 30 fractions to 70Gy in 33 fractions. Both studies recruited stage II/III patients who achieved CR or PR and EBVDNA clearance after IC (ClinicalTrials.gov Identifier: NCT04448522 and NCT05304468).
Tailoring the RT dose and volume based on tumor response during the course of RT represents a potential window for treatment adaptation. Proactive adaptive radiotherapy (ART) represents scheduled replanning tailored to anatomical changes. Anatomical alterations are commonly encountered during RT, often attributed to weight loss and tumor shrinkage. Interestingly, some studies also observed the shrinkage of OARs, including the parotid and submandibular glands [
82,
83]. Furthermore, it has been demonstrated that changes in neck contour and set-up errors during RT can lead to a notable increase in the spinal cord and brainstem dose [
84,
85,
86]. The introduction of replanning during the mid-course of RT using a new set of images may improve target coverage and better protect the normal tissues.
In a retrospective analysis of 290 patients who were enrolled in a prospective cohort [
87], proactive replanning at the 15th and/or 25th fraction was performed for half of the patients, while the other half declined. The replanning group demonstrated a higher 8-year LR-RFS rate of 87.4% compared to 75.6% in the non-replanning group, despite no significant improvement in OS. These patients also reported less dry mouth and sticky saliva. However, the effectiveness of ART in reducing xerostomia is debatable. The ARTIX trial [
88] randomized patients with locally advanced oropharyngeal squamous cell carcinoma to weekly replanning but failed to show a reduction in xerostomia in terms of stimulating salivary flow by paraffin.
Furthermore, the appropriate timing for proactive replanning remains unclear. Weekly kilovoltage cone-beam CT scans of 13 patients revealed that 11 cases (84.6%) experienced ≥50% shrinkage of GTV before the 21st fraction, which increased to 12 cases (92.3%) before the 26th fraction [
82]. Another study suggested two replans at the 5th and 15th fractions after assessing anatomic and dosimetric changes of target volume and OARs [
89].
However, the optimal method to adapt target volumes remains to be defined. Some clinicians adjust treatment volumes based solely on anatomical changes, while others advocate shrinking the high-dose volume to residual tumor. In this context, a two-phase technique has been described [
90,
91,
92]. In a report by Xie et al., the phase I delivered doses of 53-54 Gy, 47.5 Gy, and 45 Gy to the GTV, high-risk CTV and low-risk CTV, respectively, over 25 fractions. In phase II, doses of 15-15.5 Gy and 13.5 Gy were delivered to the residual GTV and high-risk CTV, respectively, over 7 fractions. Of note, the GTV was adapted in phase II while the high-risk CTV remains unchanged, ensuring that the regressed tumor receives a total dose of at least 65 Gy. Preliminary results indicated a local recurrence-free survival of 90.5% with no recurrence observed in the regressed area. [
90] Another emerging approach involves a mixed-beam arrangement [
91], in which the first IMRT phase targeted both the high- and low-risk CTVs, followed by a proton phase for the high-risk regions. This approach allows the proton therapy to target the upper neck, mitigating uncertainties linked to tissue inhomogeneity stemming from tumor shrinkage and positioning errors that are often encountered in the later stages of RT.
ART is conventionally time and labour-intensive. Implementing ART necessitates meticulous technological considerations on image quality, deformable image registration and dose accumulation[
93]. Many studies had focused on predicting or selecting patients who may benefit the most from ART, and currently, ART for head and neck cancer (HNC) is predominantly offline and ad-hoc [
94]. However, online daily ART for HNC is gaining momentum with technologies such as MRI-LINAC and the Varian Ethos
TM system, and the integration of artificial intelligence for auto-segmentation [
95,
96] and re-optimization (e.g., RapidPlan). However, whether intensive adaptive planning would translate into clinical benefits in terms of improved tumor control and reduced toxicity remains to be determined. More studies akin to ARTIX are eagerly anticipated for NPC.
Locoregional failures observed in NPC are predominantly infield, which has prompted the exploration of dose escalation strategies to enhance local control. Data from the conventional RT era had suggested that a boost dose was correlated with enhanced local control. [
97] However, achieving dose escalation across the entire tumor while minimizing adjacent normal tissue toxicity is challenging. Target volume definitions have traditionally relied on anatomical volume. However, the emergence of biological imaging, which provides insights into the metabolic, biochemical, physiological, functional, molecular, genotypic, and phenotypic characteristics of tumors, has introduced a valuable tool for delineating functionally active or potentially radioresistant sub-volumes within tumors, referred to as the biological target volume (BTV). [
98] This approach allows for customized dose delivery.
18F-FDG-PET-CT is a molecular imaging technique reflecting cancer metabolism. Various thresholds have been proposed for tumor volume definition. [
99,
100,
101] A retrospective comparative study involving 292 patients who underwent PET-guided RT employed three distinct criteria for defining the GTV. In Group 1, visual criteria were used; Group 2 utilized a standardized uptake value (SUV) threshold of 2.5, while Group 3 employed the visual criteria for GTV, and defined a sub-volume (named GTV-PET) using threshold of 50% of the maximal SUV. Dose prescription for the GTV ranged from 70.4 to 72.6 Gy in 32 to 33 fractions. Additionally, the GTV-PET in Group 3 received simultaneous integrated boost of 75.2 to 77.55 Gy in 32 to 33 fractions. The results revealed that dose-painting in group 3 correlated with improved 5-year local and distant recurrence-free survival and OS, without additional G3-G4 toxicities [
99].
Intra-tumoral hypoxia is believed to contribute to radio-resistance, which dose escalation can potentially overcome.
18F-FMISO, a nitroimidazole derivative, accumulates in hypoxic viable cells but not necrotic cells [
102]. A feasibility study utilized
18F-FMISO PET-CT to deliver a boost dose of 14 Gy (to a total of 84 Gy) to the hypoxic sub-volume (defined as tumor-muscle ratio >1.3) while respecting the conventional OAR constraints. It was shown to be achievable using both IMRT and volumetric-modulated arc therapy (VMAT) techniques [
103]. A proton-based planning study explored the feasibility of delivering a stereotactic boost of 10 GyE in 2 fractions to an FMISO PET-defined hypoxic sub-volume before the course of standard 70 GyE radiation. However, in their cohort, 3 out of 8 patients failed to meet the constraint in the temporal lobe.[
104]
The utility of DWI in defining boost volume was investigated based on the theory that viable parts of a tumor exhibit restricted diffusion and lower apparent diffusion coefficient (ADC) compared to necrotic parts [
105]. In a randomized study of 260 locally advanced NPC cases, the dose-painting group received doses of 75.2 to 77.55 Gy in 32-33 fractions to parts of tumor with ADC below the mean ADC, according to the pre-IC MRI. As compared to the control group receiving conventional 70.4-72.6 Gy in 32-33 fractions, the dose-painting group demonstrated improved 2-year disease-free survival, local recurrence-free survival, distant metastasis-free survival (DMFS), and OS. No additional grade 3 or above acute or late adverse events were observed. [
106]
With the increasing availability of integrated
18F-FDG PET and MR (PET/MR) scanners, a pilot study demonstrated that volumes defined by DWI and PET did not completely overlap. More than 90% of volume of interest (VOI) defined by DWI was enclosed in PET-defined VOI (defined as SUVmax >40%), while only around half of PET-defined VOI was encompassed in DWI-defined VOI. [
107] The findings suggested that PET and DWI may complement each other in defining the optimal sub-volume for dose escalation.
All in all, preliminary findings indicate that dose-painting holds promise for improving local tumor control. However, lessons learnt from other HNC underscored the potential late complications of dose escalation such as mucosal ulcers and dysphagia [
108,
109]. The importance of long-term safety data cannot be overstated. Moreover, prospective data is needed to assess the comparative efficacy and safety of different dose-painting strategies and to identify high-risk patients who could benefit from dose escalation. In addition to the conventional clinicopathologic features, radiomics [
110] and genomics [
111,
112] hold potential in predicting radio-resistance and selecting suitable patients for dose-painting.
Non-keratinizing NPC is consistently associated with EBV infection [
113] and the EBV in episomal forms released into the peripheral circulation upon tumor lysis. To date, most data on plasma EBVDNA predominantly utilized real-time-qPCR that targets the BamHI-W repeat region of the EBV genome [
115]. Plasma EBVDNA has emerged as an important biomarker implicated in NPC screening [
116], treatment [
117,
118], and surveillance [
118,
119]. Furthermore, EBVDNA levels are dynamic during treatment and demonstrate prognostic significance at various time-points, leading to their increasing integration into clinical trials for patient selection and treatment adaptation. However, challenges in harmonizing assays have hindered knowledge generalization.[
120]
Elevated pre-treatment levels of EBVDNA are indicative of a less favorable prognosis [
117,
121]. It has had implications for the consideration of induction, concurrent, or adjuvant chemotherapy in recent pivotal trials. Zhang et al.[
8] demonstrated that induction gemcitabine-cisplatin improved 5-year OS only in the subgroup with pre-treatment EBV DNA >4000 copies/mL. In a randomized study [
19] involving stage II and T3N0 NPC patients without adverse features, RT alone was shown to be non-inferior to CCRT in terms of 3-year failure-free survival. In this study, EBVDNA levels exceeding 4000 copies/ml were identified as one of the adverse features leading to patient exclusion. In the adjuvant setting, Miao et al.[
12] enrolled high-risk patients to receive adjuvant capecitabine, including those harbouring pre-treatment EBVDNA levels >17000 copies/ml.
A subsequent window for risk stratification and treatment adaptation emerges following IC. Patients achieving EBVDNA clearance after IC exhibit a more favorable prognosis compared to those without [
76]. A phase II non-inferiority randomized controlled trial suggested that two cycles of concurrent cisplatin (at 100mg/m2) was non-inferior to three cycles in patients who achieved EBVDNA clearance after IC. The 3-year PFS was 88% for the two-cycle group and 90.4% for the three-cycle group, resulting in a difference of 2.4% (95% CI: -4.3 to 9.1). The result has met the predefined non-inferiority margin of 10%. Notably, patients in the three-cycle group experienced significantly higher acute toxicity burden and late adverse events [
79]. Results of ongoing trials studying reduced dose RT for patients with EBVDNA clearance after IC are eagerly awaited (ClinicalTrials.gov Identifier: NCT04448522 and NCT05304468).
Post-RT EBVDNA is the most adverse prognostic factors among other predictors including pre-treatment EBVDNA, and T/N-category [
117]. It was postulated that adjuvant chemotherapy can eliminate residual tumor clones after RT, and the presence of which could be reflected in post-treatment EBVDNA level. In the NPC-0502 study [
122], Chan et al. recruited patients with positive EBVDNA levels 6-8 weeks after RT to receive adjuvant chemotherapy. Importantly, they discovered that approximately one third of these patients who had positive EBVDNA either demonstrated persistent or metastatic disease upon re-staging. However, this study failed to show clinical benefit of adjuvant gemcitabine-cisplatin. It was postulated that the lack of benefit could be due to the late commencement of adjuvant chemotherapy at a median of 13 weeks post-RT, and the selection of patients with extremely high risk of recurrence. Moving forward, many ongoing trials have incorporated post-treatment EBVDNA assessment to identify high-risk patients to receive adjuvant chemotherapy and/or immune check-point inhibitors (ICPi), for example, the NRG-HN001 ClinicalTrials.gov Identifier: NCT02135042, and NCT05517135.
ICPi targeting programmed death receptor 1 (PD-1) or its ligand (PDL1) or cytotoxic T-cell lymphocyte-associated protein 4 (CTLA-4) have shown effectiveness in treating recurrent/metastatic NPC[
123,
124,
125,
126]. Ongoing research is now exploring the role of ICPi in the definitive treatment. Phase II single-arm trials for pembrolizumab[
127] and tislelizumab[
128] (both PD-1 inhibitors) have studied the efficacy of integrating PD-1 inhibitors with IC and CCRT, followed by maintenance therapy, with results presented in abstract form. The phase III CONTINUUM[
129] study recruited patients with stage III-IVA NPC (except T3/T4N0 or T3N1). The experimental arm received up to 12 cycles of induction, concurrent and maintenance sintilimab (a PD-1 inhibitor), and it demonstrated improved 3-year event-free survival compared to the control arm treated with IC +CCRT (86.1% vs. 76%; stratified HR 0.59). The full manuscript of this study is eagerly anticipated. Moving forward, the TIRA[
130] trial evaluated treatment deintensification by omitting concurrent cisplatin. In the experimental arm, patients received IC and induction-concurrent-maintenance toripalimab (a PD-1 inhibitor).
In the era of immunotherapy, there is a pressing need to identify biomarkers for PD1/PDL1 therapy in NPC. Conventional markers like PDL1 expression and tumor mutation burden have not demonstrated strong predictive significance in NPC[
123,
124,
131]. The CONTINUUM [
129] study proposed that the clinical benefit of sintilimab was observed only in patients with tertiary lymphoid structure (TLS), the ectopic lymphoid tissues that can be found in the tumor or neighbouring peripheral tissue [
132]. Genomic studies on tumor microenvironment may offer further insights [
124,
133]. Chen et al. studied the gene expression patterns in NPC and identified 3 immune subtypes: active, evaded, and non-immune. They found that patients with an active immune subtype responded better to ICPi. [
133] Further research is needed to define this landscape.
The interaction between RT and immunotherapy has become a prominent subject of research in solid tumors, including HNC. Preclinical data suggests that radiation can have immunostimulatory effects, for example, by triggering immunogenic cell death, enriching immune tumor microenvironment, and overexpressing MHC class-I and Fas receptors on tumor to activate T cell response [
134,
135]. This RT-induced immune response is thought to synergize with immunotherapy. Yet, the optimal combination of RT and immunotherapy in terms of dose, volume, schedule, and sequence is still under investigation.
The lack of clinical benefit of ICPi in the definitive treatment of HNC[
136] has led to the postulations that elective nodal treatment inhibits the priming of T-cells naturally harboured in the lymph node chains, thus dampening tumor immune response. This phenomenon has been observed in mouse models. [
137,
138] Consequently, a strategy has been developed based on preclinical models for HNC, which involves delivering stereotactic body RT to the primary tumor in combination with immunotherapy, followed by delayed nodal treatment. Their analysis suggests that this lymph node-sparing approach can induce a systemic immune response and produce anti-tumor responses at local, regional, and distant sites.[
138] In the context of NPC, the recent movement towards volume-reduced RT, as discussed in previous sections, in particular studies focusing on limited neck irradiation [
43], may offer a potential avenue for optimizing the combination of immunotherapy and RT. However, further research is essential to fully understand how RT interacts with immunotherapy in NPC and to determine the optimal treatment strategy.