Ionizing radiation exerts cytotoxic effects primarily through DNA damage, including DNA double-strand breaks (DSBs) and the generation of reactive oxygen species through water radiolysis (10,11). Cellular responses to radiation-induced DSBs ultimately determine cell fate, with DNA repair pathways, cell-cycle checkpoints and cell-death mechanisms collectively shaping radiosensitivity (11). Mammalian cells employ two major DSB repair pathways: Homologous recombination (HR), which operates predominantly during S/G₂ and uses a sister chromatid as a repair template, and non-homologous end joining (NHEJ), which functions more broadly throughout the cell cycle (11). Radiation sensitivity also varies across the cell cycle, with cells in G₂/M generally more radiosensitive and those in late S phase more radioresistant (12). This temporal heterogeneity provides a rationale for radiosensitization strategies that modulate cell-cycle progression.

Radioresistance in NPC arises from an intricate network of tumor-intrinsic and microenvironmental factors that collectively attenuate radiation-induced cytotoxicity.

Enhanced DNA repair capacity represents a cardinal mechanism of radioresistance. NPC cells can exhibit activation of key repair proteins, including DNA-dependent protein kinase catalytic subunit (DNA-PKcs), ataxia telangiectasia mutated (ATM) and poly(ADP-ribose) polymerase-1 (PARP-1) (13,14). These proteins coordinate DSB repair and checkpoint signaling, enabling tumor cells to survive otherwise lethal radiation-induced damage. The long non-coding RNA linc00312 directly binds DNA-PKcs, and its downregulation in radioresistant NPC cells removes this inhibitory interaction and accelerates NHEJ repair (15).

Radiation normally activates cell-cycle checkpoints (G₁/S, intra-S and G₂/M) to halt cell-cycle progression and permit DNA repair (16). However, NPC cells often display checkpoint abnormalities that paradoxically promote survival. Although G₂/M arrest generally provides time for repair, prolonged arrest in radiosensitive phases can be therapeutically exploited. Conversely, aberrant checkpoint activation in radioresistant cells facilitates escape from radiation-induced growth arrest. Proteins such as cell division cycle 25C, CDK1 and cyclin B1 regulate G₂/M transition, and modulation of these molecules by radiosensitizers can override protective checkpoints (17).

CSCs constitute a critical reservoir of radioresistance in NPC. These cells, characterized by markers including CD133, CD44 and aldehyde dehydrogenase 1, exhibit enhanced DNA repair capacity, preferential activation of survival signaling [such as interleukin-6/signal transducer and activator of transcription 3 (STAT3)], and quiescence in the radioresistant G₀ phase (18,19). Circulating RNA networks, including hsa\_circRNA\_102115, regulate CSC properties and radioresistance, and curcumin-mediated suppression of these networks enhances radiosensitivity (20,21). The persistence of CSCs after radiotherapy likely underlies disease recurrence, highlighting the need for CSC-targeted radiosensitization strategies.

Tumor hypoxia, present in ~50% of solid tumors, including NPC, profoundly impairs radiation efficacy (22). Hypoxic cells exhibit two- to three-fold greater radioresistance than normoxic cells, because oxygen is required to ‘fix’ radiation-induced DNA damage through the formation of DNA-peroxy radicals (23). Hypoxia-inducible factor-1α (HIF-1α) coordinates adaptive responses, including upregulation of glycolysis (via glucose transporter 1 and lactate dehydrogenase A), angiogenesis (via vascular endothelial growth factor) and DNA repair genes [DNA repair protein RAD51 homolog 1 (RAD51) and BRCA2], collectively promoting survival under hypoxic stress (24,25). Moreover, hypoxia fosters immunosuppression through the recruitment of regulatory T cells and M2-polarized macrophages, further attenuating antitumor immunity (26).

EBV infection, etiologically associated with 95% of endemic NPC cases, directly contributes to radioresistance through multiple mechanisms (27). Latent membrane protein 1 (LMP1), a central EBV oncoprotein, activates NF-κB signaling and transcriptionally upregulates ATM, thereby accelerating DNA damage repair (28). LMP1 also drives metabolic reprogramming via the PI3K/Akt-GSK3β-F-box/WD repeat-containing protein 7 axis, stabilizing c-Myc and upregulating hexokinase 2, which enhances glycolysis (the Warburg effect) and reinforces radioresistance (29). In addition, LMP1 induces peroxisome proliferator-activated receptor g coactivator 1-a expression through protein arginine methyltransferase 1-mediated methylation, thereby inhibiting anoikis, facilitating metastasis and further promoting radioresistance (30).

Beyond glycolysis, NPC cells exhibit broader metabolic alterations, including enhanced fatty acid synthesis. Fatty acid synthase (FASN), which is frequently upregulated in NPC, is associated with radioresistance, and FASN inhibition sensitizes cells to radiation (31). Crosstalk between metabolism and DNA repair, exemplified by NADPH production that supports antioxidant defenses, creates metabolic vulnerabilities that can be therapeutically exploited (32).

The biological determinants of radioresistance, their clinical contexts and candidate biomarkers are integrated in Fig. 1.

For clinical interpretation, radiosensitization strategies should be considered according to the treatment context. In primary locoregionally advanced NPC, the benchmark remains definitive IMRT combined with platinum-based chemotherapy, with induction chemotherapy and concurrent chemoradiotherapy supported by guideline-level evidence in selected patients with stage II–IVA NPC (4). In recurrent or metastatic NPC, systemic chemotherapy and immunotherapy have stronger supporting evidence than most local radiosensitization strategies, although selected patients may benefit from reirradiation, surgery or focal radiotherapy when disease distribution and normal-tissue constraints permit (5,7). In older adult or medically frail patients, treatment intensity should be individualized according to renal function, bone marrow reserve, comorbidity, nutritional status and expected late toxicity. Accordingly, each radiosensitizing strategy in the present review is interpreted according to both its underlying mechanism and level of evidence rather than mechanistic plausibility alone.

Concurrent chemoradiotherapy (CCRT) represents the standard backbone for numerous patients with locoregionally advanced NPC, whereas induction chemotherapy followed by CCRT is recommended for selected higher-risk patients with stage II–IVA NPC according to contemporary guideline frameworks (4,5,7,33). The radiosensitizing effects of chemotherapy extend beyond simple additive cytotoxicity and involve mechanism-based synergies, including inhibition of radiation-induced DNA repair, redistribution of cells into radiosensitive cell-cycle phases and disruption of protective signaling pathways (34).

Lobaplatin, a third-generation platinum compound, has shown feasibility as a concurrent agent in older adult patients with NPC. In a small cohort of 29 patients aged ≥65 years, lobaplatin-based CCRT was associated with no liver or kidney dysfunction and a reported 3-year OS of 100%; however, this result should be interpreted cautiously because of the limited sample size, short-to-intermediate follow-up and non-randomized design (38). The gemcitabine plus cisplatin (GP) regimen has emerged as an established systemic option in recurrent/metastatic NPC, with superior progression-free survival (PFS) compared with fluorouracil plus cisplatin (7.0 vs. 5.6 months) in a phase III trial (39). Thus, platinum selection should be separated by clinical setting: Cisplatin remains a standard radiosensitizer in definitive CCRT, lobaplatin may be considered investigational or individualized in older adult/frail populations and GP is primarily supported in recurrent/metastatic systemic therapy.

Paclitaxel promotes microtubule polymerization and stabilization, inducing G₂/M-phase arrest, a relatively radiosensitive phase of the cell cycle (40). Low-dose paclitaxel pretreatment (0.05 or 1.0 nmol/l for 24 h) enhanced radiosensitivity in CNE-I NPC cells (41). G₂/M accumulation may drive cells toward mitotic catastrophe after irradiation. In locoregionally advanced NPC, GP and docetaxel, cisplatin and fluorouracil (TPF) sequential chemoradiotherapy showed similar long-term efficacy but different toxicity profiles, with more thrombocytopenia in the GP group and more leukopenia and neutropenia in the TPF group (42).

DNA methylation is implicated in NPC radioresistance, including interactions with microRNA (miRNA/miR)-mediated regulation (45). 5-Azacytidine enhanced the radiosensitivity of CNE2 and SUNE1 cells in vitro and in vivo, supporting further investigation of DNA-demethylating strategies in NPC (46).

Nicotinamide N-methyltransferase (NNMT) promotes Akt phosphorylation and enhances proliferation and migration. High NNMT expression is associated with shortened disease-specific survival and metastasis-free survival, suggesting utility as both a prognostic biomarker and a potential therapeutic target (48).

Overall, chemotherapy-based radiosensitization remains the most clinically mature strategy in NPC; however, its value differs by setting. In primary locoregionally advanced disease, platinum-based CCRT and induction chemotherapy are supported by the highest level of clinical evidence. In recurrent/metastatic disease, systemic disease control and immunochemotherapy have clearer roles than radiosensitization alone. Further gains are likely to depend on DNA repair, metabolic and EBV-related biomarkers that can identify patients most likely to benefit from treatment intensification while avoiding unnecessary toxicity.

The EGFR pathway is clinically relevant in NPC, and EGFR/phosphorylated (p-)EGFR expression has been evaluated as a prognostic biomarker (49). Nuclear EGFR is also linked to radiation response and DNA repair signaling (50). However, EGFR-targeted therapy should not be considered uniformly beneficial. A recent randomized phase 2 trial in patients with a suboptimal response to induction chemotherapy found no PFS improvement from adding nimotuzumab to concurrent chemoradiotherapy (51). These findings support biomarker development rather than unselected anti-EGFR intensification.

ANXA3, a calcium-dependent phospholipid-binding protein, is inversely associated with radioresistance in NPC. Low ANXA3 expression facilitates EGFR phosphorylation (Tyr1068) and nuclear translocation, activates DNA-PKcs and accelerates DSB repair. Clinical samples show that ANXA3-low/nuclear EGFR-high tumors are radioresistant, whereas cetuximab (an EGFR monoclonal antibody) reverses this phenotype (52). These findings support combination strategies pairing EGFR inhibitors with radiotherapy.

Mechanistic and early clinical data support continued evaluation of anti-EGFR agents with radiotherapy-based regimens, but efficacy remains uncertain. In a randomized phase 2 trial, 2-year PFS was 81.0% with nimotuzumab plus CCRT and 80.8% with CCRT alone (hazard ratio=0.93; P=0.70), with no improvement in survival outcomes and more frequent low-grade rash (51).

ANXA1 binds to the intracellular C-terminal tail of EGFR and stabilizes the receptor by competing with the Casitas B-lineage lymphoma (Cbl) E3 ubiquitin ligase, thereby reducing Cbl-mediated EGFR ubiquitination and subsequent degradation. This interaction activates STAT3, promotes programmed death-ligand 1 (PD-L1) expression and contributes to radioresistance. A nine-amino-acid peptide designed to disrupt the ANXA1-EGFR interaction downregulated EGFR and enhanced radiosensitivity in NPC cell and mouse models (53).

ATM and ATR are apical DDR kinases that activate downstream checkpoint and repair pathways. ATR kinase-dead cells exhibit impaired HR, defective G₂ checkpoint function and enhanced radiosensitivity independent of NHEJ status (14). These findings support ATR inhibition as a promising radiosensitization strategy, particularly in rapidly proliferating NPC cells with replication stress.

Bromodomain-containing protein 7 (BRD7), a tumor suppressor, competes with USP5 for METTL3 binding and promotes proteasomal degradation of METTL3. METTL3 reduction suppresses HR repair proteins (BRCA1 and RAD51), thereby enhancing radiosensitivity. Clinical correlations indicate that BRD7-high/METTL3-low tumors are associated with radiosensitivity and favorable prognosis (55).

Recombinant human endostatin (Endostar) is an anti-angiogenic strategy evaluated with PF chemotherapy and sequential IMRT in locally advanced NPC. A randomized, open-label multicenter phase II study found the regimen tolerable and reported improved PFS, although confirmatory trials are required (56).

Ferritin-encapsulated platinum nanoparticles target hypoxic NPC cells through transferrin receptor 1, which is upregulated under hypoxia. These nanozymes catalyze H₂O₂ decomposition into O₂, alleviating hypoxia and enhancing radiosensitivity. In a preclinical study, Pt-HFn combined with RT inhibited tumor growth and prolonged survival more effectively than RT combined with sodium glycididazole (57). Sodium glycididazole itself has shown radiosensitizing activity in NPC models by enhancing DNA damage and apoptosis (58).

JAB1, also known as CSN5, participates in signaling that supports NPC cell survival and radioresistance. The curcumin analogue T83 inhibited JAB1 expression, induced G₂/M arrest and apoptosis, and enhanced radiosensitivity in NPC models (60). miR-24 similarly targets JAB1, and its upregulation induces apoptosis and radiosensitization (61).

Leucine zipper tumor suppressor 2 (LZTS2), which is downregulated in NPC, binds the PI3K regulatory subunit p85, prevents p85-p110 interaction and inhibits Akt activation (62). LZTS2 overexpression suppresses proliferation and radioresistance, whereas p85 knockdown reverses these effects (62). Raf kinase inhibitor protein (RKIP), also downregulated in NPC, suppresses ERK and Akt pathways when expressed, and RKIP-low patients exhibit shorter OS and disease-free survival (63).

Collectively, these strategies are most plausible when a measurable molecular vulnerability, such as aberrant EGFR trafficking, defective DNA damage response or hypoxia-driven survival, can be matched to an intervention whose dose, schedule and toxicity permit combination with radiotherapy.

Clinical translation of targeted radiosensitizers remains uneven. Anti-EGFR antibodies have early clinical data in selected locoregionally advanced NPC, whereas a number of DDR inhibitors, peptide inhibitors and nanozyme-based oxygenation strategies remain preclinical or early translational. Eligible patient subgroups should be defined by measurable pathway activation, such as EGFR trafficking, HR deficiency, replication stress, hypoxia markers or dynamic EBV DNA risk, rather than by diagnosis alone. Safety evaluation is essential because DDR inhibition or hypoxia modification may also sensitize normal mucosa, salivary tissue and marrow to radiation.

Radiotherapy can produce direct and bystander effects beyond the irradiated cells, although the clinical importance of these effects remains to be elucidated (64). Low-dose radiation can modulate antioxidant responses, DNA repair, apoptosis and antitumor immune responses (65). In NPC, epigenetic mechanisms contribute to immune evasion, providing a rationale for investigating DNA methyltransferase (DNMT) or histone deacetylase (HDAC) inhibitors, together with immune checkpoint blockade (66). These observations support further investigations of immunoradiotherapy, but do not by themselves establish a uniform abscopal effect or radiation-induced PD-L1 response in NPC. NPC-specific in vitro evidence indicates that high-dose irradiation can induce an immunogenic cell-death phenotype and enhance dendritic cell (DC) maturation; this supports the proposed pathway from immunogenic cell death to DC maturation, but tumor-antigen release and downstream T-cell priming were not directly demonstrated (67). Separately, radiotherapy-associated PD-L1 upregulation was observed in NPC cell lines and a patient-derived xenograft model, suggesting a context-dependent adaptive checkpoint response that remains to be validated clinically (68).

Camrelizumab, a PD-1 monoclonal antibody, achieved an objective response rate (ORR) of 28.2% in pretreated recurrent/metastatic NPC, with greater efficacy in PD-L1-positive patients (69). Tislelizumab combined with chemotherapy as first-line treatment for recurrent/metastatic NPC markedly prolonged PFS in the phase III RATIONALE-309 trial (hazard ratio=0.52) (70). These data support the use of checkpoint blockade most strongly in recurrent/metastatic disease. In locoregionally advanced NPC, induction chemoimmunotherapy before CCRT remains an evolving strategy; a small non-randomized study reported higher ORR (95.7 vs. 77.8%), complete response rate (39.1 vs. 22.2%) and 24-month event-free survival (88.9 vs. 62.6%) than induction chemotherapy alone (71).

In a retrospective cohort, TPF induction chemotherapy combined with a PD-1 inhibitor produced a higher ORR (88.5 vs. 71.2%) and complete response rate (29.5 vs. 11.3%) than TPF alone (72). Nevertheless, combined radiotherapy, chemotherapy and immunotherapy may increase mucositis, dermatitis, hematologic toxicity, endocrinopathy, hepatitis, pneumonitis or other immune-related adverse events. Prospective trials should define whether ICIs are best used as induction, concurrent, consolidation or salvage therapy.

Circulating EBV DNA serves as a dynamic biomarker in NPC and associates with tumor burden and treatment response (73). Early EBV DNA clearance (2 weeks) after induction chemotherapy predicts greater tumor shrinkage (78.9 vs. 56.7% nasopharyngeal volume reduction; 81.6 vs. 59.3% nodal volume reduction) and superior 3-year OS (89.2 vs. 71.4%) and PFS (85.7 vs. 64.3%) (74). Persistent EBV DNA positivity during follow-up, even in the absence of clinical recurrence, supports consideration of metronomic chemotherapy (such as capecitabine) in selected patients, which prolongs modified disease-free survival (12.9 vs. 6.8 months) (75). Longitudinal EBV DNA monitoring (0–2 weeks and 8–12 weeks post-radiotherapy) enables early recurrence detection, with persistent positivity associating with shorter PFS (73).

Although validated in other malignancies, data on TMB and MSI in NPC remain limited. Given NPC's viral etiology and distinct mutational landscape, these biomarkers require prospective validation (76).

DNMT and HDAC inhibitors can reverse immune suppression by upregulating antigen-presentation machinery and chemokine expression, potentially synergizing with ICIs (66). This epigenetic reprogramming may convert immunologically cold tumors into hot, immunogenic lesions that are more amenable to checkpoint blockade.

These findings support immunoradiotherapy as a biologically plausible strategy; however, patient selection should be guided by PD-L1 expression, dynamic EBV DNA kinetics, and immune microenvironment features rather than treating checkpoint blockade as a uniform radiosensitizing approach. Mechanism-guided treatment strategies across clinical settings and levels of evidence are summarized in Fig. 2.

Several natural compounds and TCM-derived molecules converge on DDR and checkpoint regulation. Curcumin, the principal curcuminoid derived from Curcuma longa, exemplifies this mechanism-based group through multi-pathway modulation, but its clinical translation remains limited by bioavailability, formulation variability and a lack of definitive randomized NPC trials.

Curcumin downregulates miR-205-5p, thereby derepressing tumor protein p53-inducible nuclear protein 1 (TP53INP1), which promotes apoptosis in radioresistant NPC cells (C666-IR) (78). In addition, curcumin modulates circular RNA (circRNA) networks, particularly the hsa-circRNA-102115-hsa-miR-335-3p-MAPK1 axis, suppressing cancer stem cell properties and enhancing radiosensitivity (20,21). Curcumin also upregulates miR-593, which targets multidrug resistance protein 1 and reverses radioresistance (79).

The curcumin analogue T83 targets JAB1/CSN5, induces G₂/M arrest and apoptosis and enhances radiosensitivity in NPC models (60). The parent compound curcumin also alters cell-cycle and radiation-response pathways (80).

Tetrandrine, ginsenoside Rg3 and berberine illustrate a second mechanism-based group targeting checkpoints, EMT and EGFR-associated signaling. At non-cytotoxic concentrations, tetrandrine enhanced radiation-induced growth inhibition and apoptosis, increased DNA damage and abrogated G₂/M arrest in NPC cells; combined treatment also reduced xenograft growth (17).

Ginsenoside Rg3 from Panax ginseng suppresses radiation-induced EMT, a process that confers radioresistance and metastatic potential (81). Rg3 inhibits EGFR nuclear translocation and reduces DNA-PK expression, thereby impairing DSB repair while promoting apoptosis (81). Berberine similarly sensitizes NPC cells to radiation through Sp1 inhibition and EMT suppression (82). Compared with synthetic EGFR or DDR inhibitors, these agents may affect several pathways simultaneously but have weaker pharmacokinetic standardization and substantially lower clinical evidence.

Ras association domain-containing protein 1 (RASSF1A)-associated redox control and gambogic acid illustrate mechanisms that modulate oxidative stress and hypoxia-related survival. RASSF1A overexpression promoted FoxO3a activity and inhibited the Kelch-like ECH-associated protein 1/nuclear factor erythroid 2-related factor 2/thioredoxin reductase 1 axis, increasing oxidative stress and radiosensitivity in NPC models (83). Gambogic acid inhibited HIF-1α expression and sensitized hypoxic NPC cells to radiation (84). These findings remain largely preclinical.

Complex TCM formulations differ from single synthetic radiosensitizers because they are usually developed within multi-component and multi-target frameworks. Their effects should therefore be evaluated using quantifiable clinical and biological outcomes rather than by treating TCM syndrome categories as direct molecular entities. A 2024 observational study measured serum inhibin subunit α, inhibin subunit β B, EBV DNA and EBV viral capsid antigen immunoglobulin A in relation to clinical stage; however, its retrospective design precluded causal or treatment-predictive conclusions (85). For radiation injury, a meta-analysis of 30 randomized controlled trials involving 2,562 patients evaluated heterogeneous oral TCM herbal formulations categorized as ‘nourishing Yin and clearing Heat’ therapy, rather than a single standardized preparation. These formulations were administered as adjuncts to trial-specific control regimens, which varied across studies and included gentamicin, dexamethasone, local anesthetics, vitamin B12, compound borax or chlorhexidine gargles and recombinant human epidermal growth factor. Compared with the control regimens alone, adjunctive TCM therapy delayed the onset of grade I radiotherapy-induced oral mucositis by 10.80 days, increased the cumulative radiation dose delivered before its onset by 5.72 Gy and reduced the risk of grade III–IV oral mucositis (relative risk=0.25). However, the heterogeneity of the herbal formulations and control regimens limits the generalizability of these findings, and further large, rigorously designed randomized trials are required (86). A 2025 double-blind randomized trial included 24 evaluable patients receiving Danggui Buxue Tang (n=11) or placebo (n=13). Grade 2–3 neutropenia occurred in 0/11 and 4/13 patients, respectively (P=0.10), while grade 1 anemia occurred in 1/11 and 4/13 patients, respectively (P=0.33). Conversely, grade 2–3 anorexia was more frequent in the Danggui Buxue Tang group than in the placebo group (10/11 vs. 5/13 patients; P=0.01). RNA sequencing identified differential gene expression associated with B-cell receptor signaling. Due to the small sample size, mixed toxicity findings and absence of statistically significant differences for several outcomes, these results remain exploratory (87).

Modern observations involving oxidative stress, inflammation, epithelial injury, immunity and the oral microbiome should therefore be interpreted as measurable parallel domains rather than literal biomedical translations of TCM syndromes. Within this cautious framework, Shengmai Jianghuang San (SMJHS) inhibits cancer stem cells and hypoxic adaptation through the tenascin C/HIF-1α/YAP1 pathway. SMJHS suppresses proliferation, migration and cancer stem cell biomarker (CD44) expression under hypoxia while downregulating HIF-1α and YAP1, thereby enhancing radiosensitivity (88). Metabolomic profiling indicates multi-target engagement across lipid metabolism and oxidative stress pathways (89).

SMY reversed radiation-associated EMT in radioresistant NPC cells by inhibiting lipocalin 2 expression (90). DNA methylation analysis also indicates that SMY modulates epigenetic landscapes in radioresistant cells (91).

Dendrobium officinale was evaluated for radiation-induced oral mucositis and salivary/oral-microbiome changes in patients with NPC (92). Separately, modified Sijunzi decoction was associated with reduced radiation-induced sinusitis in a retrospective study, and network pharmacology implicated JAK-STAT and NF-kB-related pathways (93).

Schisandrin B and emodin provide additional preclinical examples. Schisandrin B targets CDK4/6, induces cell-cycle arrest and enhances NPC radiosensitivity (94). Emodin inhibits EBV reactivation and NPC tumorigenic phenotypes, although the cited study did not test ionizing-radiation sensitization (95).

Photodynamic therapy (PDT) uses non-ionizing visible light and photosensitizers and therefore should not be classified as radiosensitization for ionizing radiation. Studies of Zn-BC-AM, hypericin and hypocrellin are best interpreted as related light-based cytotoxic approaches (96–99). They are retained only to clarify this distinction and are not included in Table I as ionizing-radiation radiosensitizers.

Compared with synthetic agents that usually target a defined molecule such as EGFR, PARP, ATR or PD-1, natural compounds and TCM-derived agents may engage multiple pathways, including DDR, EMT, oxidative stress, hypoxia, inflammation and microecology. This breadth may be advantageous for heterogeneous tumors but also complicates dose standardization, target attribution and regulatory evaluation. A large matched observational cohort reported an association between adjunctive Chinese herbal medicine and clinical outcomes, supported by experimental work, but its non-randomized design cannot establish causality (100). Key challenges therefore include limited bioavailability, batch variability, insufficient pharmacokinetic characterization, uncertain herb-drug interactions and a lack of randomized multicenter validation.

Therefore, natural compounds and TCM-derived agents should be regarded primarily as hypothesis-generating or adjunctive candidates until standardized preparations, quality control, optimal dosing, safety with CCRT and interactions with immunotherapy or targeted therapy are prospectively clarified. Natural products and TCM-derived interventions, together with their evidence limitations and translational requirements, are summarized in Fig. 3. Representative mechanism-guided radiosensitization strategies for NPC are summarized in Table I.

The present review indicates that mechanism-guided radiosensitization in NPC has progressed from empirical chemotherapy-based approaches toward mechanism-guided, multimodal therapeutic strategies. The principal conclusions are as follows: i) Mechanistic convergence: Radioresistance emerges from interconnected mechanisms, including enhanced DNA repair, cell-cycle checkpoint adaptation, CSC enrichment, hypoxic microenvironments, EBV-driven oncogenic signaling and metabolic reprogramming; effective radiosensitization therefore requires pathway-informed interventions rather than non-specific treatment intensification; ii) clinical setting matters: Platinum-based chemoradiotherapy remains most established in primary locoregionally advanced NPC, whereas chemoimmunotherapy has stronger evidence in recurrent/metastatic disease and selected induction settings; iii) precision targeting: EGFR inhibition, DDR pathway inhibition and hypoxia-directed strategies provide mechanism-based opportunities to enhance radiotherapy, although numerous approaches still require prospective clinical validation; iv) biomarker hierarchy: Plasma EBV DNA currently has the strongest disease-specific and dynamic clinical readiness; PD-L1 is context- and cut-off-dependent, whereas TMB, ERCC1/XPA, EGFR-axis and hypoxia markers require further prospective validation for treatment selection; and v) natural compounds and TCM-derived agents provide multi-target mechanistic hypotheses and emerging supportive-care signals, but clinical adoption requires standardized preparations, pharmacokinetic characterization and rigorous multicenter trial validation.

A central requirement for clinical translation is a practical biomarker hierarchy that distinguishes prognostic association from demonstrated treatment-predictive utility. A 2025 systematic review identified plasma EBV DNA and its kinetics as a robust candidate for risk stratification after chemoradiotherapy, while emphasizing the need to standardize assays and cut-off values (101). In a 2025 meta-analysis of 13 cohorts comprising 1,507 patients receiving ICIs, high pretreatment EBV DNA was associated with worse PFS (hazard ratio=1.72) and OS (hazard ratio=2.03) (102). By comparison, a 2024 meta-analysis of prospective trials reported a higher pooled ORR in later-line immunotherapy when PD-L1 expression was at least 1% (0.37 vs. 0.22), but first-line PFS benefit was not notably different between PD-L1 subgroups and cut-offs of 10 or 25% did not discriminate outcomes (103). A broader 2023 meta-analysis similarly associated lower baseline or decreasing EBV DNA with improved immunotherapy response, whereas positive vs. negative PD-L1 status and high vs. low TMB were not notably associated with ORR (104). Thus, plasma EBV DNA has the highest current clinical readiness; PD-L1 remains context-dependent; ERCC1/XPA is a retrospective platinum-response candidate (36,37,47); EGFR/p-EGFR and nuclear EGFR have mechanistic and prognostic support but no validated selection threshold (49–51); and hypoxia markers remain predominantly preclinical or investigational (25,57). Table II provides a direct comparison of these biomarkers and their supporting evidence (25,36,37,47,49–51,57,69,73,74,76,101–104). Despite notable advances, several obstacles impede optimal radiosensitization.

The main gap is no longer the absence of candidate markers, but the lack of standardized assays, prospectively validated thresholds, head-to-head incremental-value analyses and trials showing that biomarker-directed treatment improves outcomes. Multiomics panels should therefore be compared against the clinically simpler EBV-DNA benchmark before routine implementation.

Optimal sequencing, dosing and partner selection remain incompletely defined. NPC cell-line experiments demonstrate that cisplatin exposure duration, sequence and interval relative to radiation can materially alter cytotoxicity, supporting systematic clinical optimization (35).

Failure of treatment benefit cannot be summarized by a single percentage across chemotherapy, targeted therapy, immunotherapy and natural compounds because the clinical settings and endpoints differ. Resistance reflects interacting DNA-repair, hypoxia, metabolic, stemness and immune mechanisms (9,22,26), while proposed immunotherapy biomarkers remain heterogeneous and variably validated (76).

Combination regimens may produce additive or synergistic toxicities, including mucositis, dermatitis, xerostomia, marrow suppression, nephrotoxicity, ototoxicity and immune-related adverse events. Older adult patients, patients with renal dysfunction, malnutrition, heavy comorbidity burden, prior irradiation or recurrent disease require individualized assessment of therapeutic index rather than automatic treatment intensification.

Current guidelines and evidence summaries support intensity-modulated radiotherapy-based CCRT for locoregionally advanced NPC, induction chemotherapy for selected high-risk disease, and platinum-based chemotherapy plus a PD-1 inhibitor for recurrent or metastatic disease (4,5,7). Existing real-world evidence discussed in the present review includes a 29-patient older adult lobaplatin cohort (38), a retrospective study of a PD-1 inhibitor plus TPF induction chemotherapy (72), a matched observational cohort of adjunctive Chinese herbal medicine (100) and a prognostic model integrating nutritional, inflammatory and post-treatment EBV DNA variables (105). These studies broaden representation of older patients and routine-care populations, but heterogeneous treatment settings and endpoints, non-randomized treatment allocation and residual confounding prevent causal comparisons across strategies.

Accordingly, real-world data should complement rather than replace randomized evidence. Priority analyses should use prespecified treatment definitions, propensity-based or target-trial methods, standardized toxicity and quality-of-life outcomes, and stratification by age, renal function, prior irradiation, endemic region and EBV DNA kinetics. Until such evidence is available, guideline-supported CCRT or chemoimmunotherapy should remain the clinical benchmark, whereas experimental radiosensitizers should be restricted to trials or carefully documented individualized use.

EBV biology should be integrated as a unifying axis that distinguishes NPC from most other head and neck cancers. Etiopathogenetically, EBV latent proteins such as LMP1 promote NF-kB activation, ATM-mediated DNA repair, PI3K/Akt-driven metabolic reprogramming, immune escape and metastatic fitness. As a biomarker, plasma EBV DNA provides a dynamic measure of tumor burden, early response, minimal residual disease and recurrence risk, although assay and cut-off standardization remain necessary (101,102). As a therapeutic target, EBV-associated antigens such as LMP2A and Epstein-Barr virus nuclear antigen 1 support vaccine and T-cell-based approaches, whereas EBV-driven signaling may identify tumors vulnerable to metabolic, immune or DDR-directed radiosensitization. A cohesive EBV-centered framework may therefore integrate etiopathogenesis, risk stratification and therapeutic targeting more effectively than discussing these aspects separately.

Administration sequence is a central unresolved issue. Cisplatin and taxanes may require defined lead times before radiotherapy to maximize DNA damage or cell-cycle redistribution, whereas ICIs may be used as induction, concurrent, consolidation or salvage therapy depending on stage and trial design. EGFR inhibitors combined with radiotherapy may increase skin and mucosal toxicity; DDR inhibitors may compromise normal-tissue repair; platinum agents may worsen renal, auditory and marrow toxicity; and immunotherapy may add immune-related adverse events. For older adults or medically frail patients, treatment choice should incorporate comprehensive geriatric assessment, renal function, hearing status, nutrition, baseline EBV DNA, comorbidity, patient preference and expected late toxicity.

First, multiomics biomarker integration should combine genomic profiling of DNA repair alterations, transcriptomic signatures including miRNA, long non-coding RNA and circRNA networks and proteomic or phosphoproteomic quantification of DDR proteins, checkpoint kinases and signaling states (106–109). Second, rational combination optimization should include experimental evaluation of fraction timing and dose rate (110), biomarker-selected investigation of dual checkpoint blockade in NPC (111), and disease-specific validation before extrapolating multi-target miRNA radiosensitization findings from other tumor types (112). Third, artificial intelligence models may facilitate individualized induction-chemotherapy selection and risk stratification for distant metastasis in NPC (113,114). Stimuli-responsive nanocarriers may improve tumor selectivity, but the cited pH-activated fullerene nanogel is a photodynamic, light-activated platform rather than an ionizing-radiation radiosensitizer (115). Ultra-high-dose-rate radiotherapy has shown partial reversal of radioresistance in preclinical head and neck squamous cell carcinoma models, but NPC-specific clinical validation is required before extrapolation (116). Long-term comparative studies should also track late toxicity and second-primary cancer risk after NPC radiotherapy (117). Fourth, translational implementation should incorporate validated artificial intelligence-assisted target segmentation (118). Careful mechanistic evaluation of DNA damage checkpoints is also required. In a previous NPC study, CHK1/WEE1 inhibitors were not effective radiosensitizers when administered after irradiation (119), underscoring the importance of treatment sequencing. Standardized post-induction target delineation should also be incorporated (120). Fifth, survivorship research should focus on radioprotective and regenerative strategies to reduce late toxicities, including xerostomia, osteoradionecrosis and cognitive impairment, as well as mechanistic studies of radiation-induced second primary cancers (117,121). Finally, implementation studies should prospectively validate risk models integrating nutritional, inflammatory and post-treatment EBV DNA variables (105) and combined nasopharyngeal-brush plus plasma EBV DNA surveillance (122). Their feasibility, clinical utility and cost-effectiveness require dedicated evaluation rather than being assumed from the current retrospective and diagnostic studies.

Mechanism-guided radiosensitization in NPC reflects the integration of molecular oncology, immunology, pharmacology, radiobiology and computational science. Its clinical value will depend less on adding more agents indiscriminately and more on matching strategies to disease setting, EBV-related risk, biomarker-defined vulnerabilities and patient fitness. Chemotherapy, targeted therapy, immunotherapy, and selected natural compounds constitute a broad therapeutic armamentarium, but their optimal use depends on rigorous biomarker validation, rational regimen sequencing, toxicity-aware clinical trial design and equitable access to advanced diagnostics and therapeutics. By aligning therapeutic interventions with the biological determinants of radioresistance, individualized radiosensitization may improve disease control while preserving quality of life.