Effects of different drugs on radiation-induced oral mucositis in patients with head and neck tumors: a systematic review and a network meta-analysis

Abstract

Background

Radiation-induced oral mucositis (RIOM) affects over 80% of head and neck tumor patients receiving radiotherapy. Although four routine drugs are used for RIOM prevention, direct comparative evidence across multiple clinical outcomes remains insufficient. This study aims to compare the effects of different pharmacological interventions on RIOM in patients with head and neck tumors via a network meta-analysis of RIOM.

Methods

We conducted systematic searches in PubMed, EMBASE, and the Cochrane Library for RCTs published up to August 31, 2025. The search only included articles in English. We selected studies that reported on the severity of RIOM, the duration of RIOM, the severity of xerostomia, and the severity of dysphagia as primary outcomes. Secondary outcomes included the use of analgesics, weight loss, the use of feeding tubes, and treatment interruptions. We used the Risk-of-Bias 2 (RoB2) tool to assess the risk of bias in the included studies and performed a network meta-analysis by using Stata 17.0 software.

Results

A total of 29 RCTs involving 2,276 patients were included. Glutamine significantly reduced the risk of RIOM compared with amifostine [odds ratio (OR) =0.13, 95% confidence interval (CI): 0.03–0.51] and cytokine (OR =0.31, 95% CI: 0.10–0.99). Compared with amifostine, cytokine (OR: 10.28, 95% CI: 1.09, 96.79), glutamine (OR: 11.09, 95% CI: 1.10, 112.24), and placebo (OR: 12.54, 95% CI: 1.41, 111.20) all significantly increased the number of cases with dysphagia. Glutamine reduced the number of patients requiring feeding tubes and the incidence of treatment interruptions. However, no significant differences were found in the duration of RIOM, the number of reported ≥ grade II xerostomia cases, the number of patients using analgesics, or weight loss.

Conclusions

Glutamine is recommended as the preferred pharmacological intervention for preventing and alleviating RIOM in head and neck cancer patients, particularly in reducing the risks of severe RIOM, incidence of feeding tube, and treatment interruptions. Additionally, amifostine significantly reduced the risk of dysphagia. Based on the current evidence, glutamine appears to offer the most favorable risk-benefit profile among the evaluated interventions. Future research should adopt well-designed, unified RCTs with adequate samples to revalidate whether these measures are truly effective and safe.

Highlight box.

Key findings

• Glutamine exhibits the optimal risk-benefit profile to reduce severe radiation-induced oral mucositis (RIOM), feeding tube placement and radiotherapy interruption among four interventions. Amifostine uniquely minimizes severe dysphagia, while no agent shows obvious superiority in shortening RIOM duration or alleviating xerostomia.

What is known and what is new?

• Amifostine, cytokines, glutamine and calcium phosphate can relieve RIOM in head and neck tumor patients, yet direct head-to-head randomized controlled trial comparisons of these four agents are scarce. Traditional pairwise meta-analyses cannot rank their comprehensive efficacy across multiple clinical outcomes.

• This is the first network meta-analysis simultaneously comparing the four commonly used drugs across eight core endpoints. We quantified their relative efficacy and ranked them via surface under the cumulative ranking curve (SUCRA) values, providing evidence-based medication selection guidance for clinical RIOM prevention.

What is the implication, and what should change now?

• Based on existing clinical evidence, glutamine is preliminarily proposed as the primary oral prophylactic agent in the prevention of RIOM. Amifostine can be selectively applied for patients at high risk of severe dysphagia. Future trials should unify multi-timepoint toxicity assessment and incorporate patient-reported outcomes to verify our conclusions.

Introduction

Radiation-induced oral mucositis (RIOM) affects more than four out of five patients receiving treatment for head and neck cancer, a rate that has remained persistently high despite intensity-modulated fields and image guidance (1). The lesions turn eating into a painful task, drive malnutrition and weight loss, and frequently force unplanned breaks in radiation, eroding both tumour control and quality of life (2-7).

Head and neck cancer management encompasses three principal indications for radiotherapy or chemoradiotherapy: definitive (primary) treatment for organ preservation, adjuvant postoperative therapy to eradicate residual microscopic disease, and larynx-preservation strategies for selected patients with locally advanced laryngeal or hypopharyngeal carcinoma. The choice among these approaches is governed by tumour stage, subsite, functional status, and patient preference, yet all share the common denominator of high-dose irradiation to mucosal surfaces and the attendant risk of RIOM (8,9). Beyond pharmacological interventions, photobiomodulation (PBM)—commonly referred to as low-level laser therapy—has emerged as a non-invasive physical modality that serves dual roles: prophylactic preconditioning before radiation exposure to reduce mucositis incidence, and therapeutic application after lesion onset to accelerate healing and alleviate pain (10). Nevertheless, its adoption remains limited by heterogeneous dosing protocols, variable wavelengths and energy densities across studies, and inconsistent availability in clinical settings (4,8). Consequently, pharmacological prophylaxis continues to constitute the mainstay of RIOM prevention in routine practice. Cytokines, glutamine, amifostine, and supersaturated calcium phosphate are agents whose distinct mechanisms have shown early promise in lessening mucosal injury (11,12). The selection of these agents is based on three considerations. First, they are the most extensively investigated interventions for RIOM in the existing literature, with a sufficient number of randomized controlled trials (RCTs) to support a robust network meta-analysis. Second, they address distinct pathogenetic pathways: amifostine exerts cytoprotection through free radical scavenging; cytokines modulate immune responses and promote epithelial repair; glutamine maintains mucosal barrier integrity and reduces inflammatory stress; and calcium phosphate acts as a topical barrier protecting denuded mucosa. Third, despite their widespread use, head-to-head comparisons among these agents in the setting of head and neck radiotherapy are almost entirely lacking. The only available direct evidence derives from studies in hematopoietic stem cell transplantation, in which cytokines have been shown to be superior to calcium phosphate in reducing mucositis severity. This scarcity of direct comparisons underscores the need for a network meta-analysis to estimate the relative efficacy of these interventions (13).

We therefore undertook a network meta-analysis to rank the four agents simultaneously for efficacy and toxicity in the head and neck cancer population. By integrating direct and indirect comparisons, we aimed to replace empirical choice with an evidence-based hierarchy that supports precision prevention and, ultimately, reduces the clinical burden of RIOM. The design and conduct of this network meta-analysis followed the methodological guidance provided by the Cochrane Handbook for Systematic Reviews of Interventions and we present this article in accordance with the PRISMA-NMA reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0538/rc) (14-16).

Methods

The protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO, CRD42024579718).

Search strategy and study selection

We searched PubMed, EMBASE, and the Cochrane Library for publications up to August 31, 2025 to identify potential studies. The search also included reference lists of other identified systematic reviews evaluating outcomes associated with the drugs of interest. To capture relevant records, we combined disease-oriented free-text words and controlled descriptors—“stomatitis”, “oral mucositis”, “radiation-induced oral mucositis”, and the abbreviation “RIOM”—with intervention-specific terms covering amifostine (WR-2721), cytokines, glutamine and calcium phosphate.

Two reviewers (S.L. and L.S.) independently screened titles and abstracts, assessed the full texts of potentially relevant articles, and extracted data using a piloted form; disagreements were resolved by discussion or, if necessary, by arbitration from a third senior investigator. The selection process, with reasons for exclusion at each stage, is reported in the PRISMA flow chart (Figure 1).

Figure 1.

The flowchart of study selection.

Inclusion criteria

All included studies followed the PICOS framework. The included studies were all RCTs, and the study subjects were patients aged ≥18 years with pathologically confirmed head and neck tumors who developed radiation-induced oral mucositis [meeting WHO, Radiation Therapy Oncology Group (RTOG), or Common Terminology Criteria for Adverse Events (CTCAE) criteria]. In addition, all included patients had received either definitive radiotherapy, adjuvant radiotherapy, or concurrent chemoradiotherapy. At least one intervention arm of the study consists of one of the four pharmacological interventions of interest (amifostine, cytokine, glutamine, or calcium phosphate), the other arm served as a control, either blank or placebo. Studies were included if they reported outcomes relevant to the assessment of the efficacy and safety of the interventions for RIOM. These outcomes typically included measures of RIOM severity, duration, and the incidence of adverse events. Only articles published in English were included.

Exclusion criteria

Non-RCTs (e.g., cross-sectional studies, cohort studies, case reports, review articles, conference abstracts, letters) were excluded to ensure comparability of the effect. Additionally, studies that did not report on the primary outcomes of interest were excluded.

Data extraction

Two reviewers independently extracted data onto a standardized, piloted form (Microsoft Excel); discrepancies were resolved by consensus or, when necessary, by a third senior investigator. Variables captured comprised bibliographic details (first author, year, journal), design elements (two-arm or multi-arm parallel trials), sample size, participant characteristics (mean age, proportion of males), intervention descriptors (agent, dose, schedule, duration), and outcome measures [event counts or means ± standard deviation (SD)] at the study’s primary reported time point.

Primary and secondary outcomes

The primary outcomes assessed in this meta-analysis included: severity of RIOM, duration of RIOM, severity of xerostomia, and severity of dysphagia. Secondary outcomes include use of analgesics, weight loss, use of feeding tube and treatment interruption were also counted in this study.

Outcomes were recorded as close to the end of the radiation therapy (week 7) as possible for all analyses. If information was not available, the time point closest to the end of the radiation therapy was preferred.

Risk-of-bias assessment

Two investigators (S.L. and L.S.) conducted the literature search and extraction of studies independently from each other using Risk-of-bias 2 (RoB 2) tool (17). We used RoB 2 (revised Cochrane risk-of-bias tool for randomized trials) to assess bias at the outcome level. The tool structures threats into five domains, each containing signalling questions with six response options: yes, probably yes, probably no, no, no information, or—not for mandatory items—not applicable. A domain-level algorithm maps the ordered responses to low risk, some concerns, or high risk. Two reviewers independently applied RoB 2, reconciled discrepancies by consensus, and if necessary, consulted a third reviewer.

Statistical analysis

Analyses were conducted in Stata 17.0. We pooled dichotomous outcomes as odds ratios (ORs) and continuous outcomes measured on different scales as standardised mean differences (SMDs), each reported with 95% confidence intervals (CIs). Between-study heterogeneity was handled with a random-effects model. Node-splitting tests were applied to check agreement between direct and indirect estimates. The network structure was visualised in a plot whose node areas reflect the number of participants randomised to each treatment and whose edge widths correspond to the number of trials providing head-to-head comparisons. Hierarchy for efficacy and safety was derived from the surface under the cumulative ranking curve (SUCRA) probabilities for every intervention on primary and secondary endpoints. League tables integrating direct and indirect evidence were constructed to present all pairwise effect estimates. Small-study effects were appraised with funnel plots.

Results

Study characteristics

The search strategy generated 2,605 articles, of which 74 were retrieved for further assessment. After excluding 45 articles, 29 eligible RCTs were included in this network meta-analysis (Figure 1). These 29 RCTs were published between 2000 and 2025 and involved a total of 2,276 participants (1,142 in the intervention groups and 1,134 in the control groups), with sample sizes ranging from 17 to 303. All studies enrolled patients with head and neck cancer, predominantly presenting with locally advanced disease (stage III–IV). The mean age of participants ranged from 47 to 67 years, and the male-to-female ratio was approximately 3:1 (with males generally exceeding 70% across studies). Treatment regimens included radiotherapy alone, concurrent chemoradiotherapy (primarily cisplatin-based, at doses ranging from 40 to 100 mg/m²), and adjuvant chemoradiotherapy. The pharmacological interventions included amifostine (n=5), cytokines (n=12), glutamine (n=10), and calcium phosphate (n=2), with follow-up times ranging from 5 weeks to 34 months. Table 1 summarizes the baseline characteristics of the included studies.

Table 1. Characteristics of randomised controlled trials of pharmacological interventions on radiation-induced oral mucositis in patients with head and neck tumors.

Study (first author, year, country/region)	Study type	Study population	Primary tumor site	Intervention vs. control (dose, route, frequency; N; RT regimen)	Primary endpoints and results	Secondary endpoints and results
Bourhis 2000, France (18)	RCT	HNSCC	NR	Intervention (n=13): amifostine 150 mg/m2 IV, 15–30 min before each RT session Control (n=13): no treatment (blank) RT regimen: very accelerated RT: 62–64 Gy, 2 Gy/fraction, b.i.d. (8–9 h interval)	Duration of RIOM: 25.1 vs. 49.2 days (P=0.03); feeding tube: 11/13 vs. 13/13	Feeding tube duration: 1 vs. 2.5 months (P<0.01); local-regional control: no difference [NMA: vs. placebo: MD −22.84 (−47.73, 2.06); vs. placebo: OR 0.17 (0.01, 4.13)]
Buentzel 2005, Germany (19)	RCT	Head and neck cancer	NR	Intervention (n=67): amifostine 200 mg/m2 IV, 15–30 min before each RT fraction Control (n=65): placebo RT regimen: CRT	Grade ≥3 RIOM: reduced (P<0.0001)	Xerostomia: reduced; grade ≥2 xerostomia: reduced [NMA: vs. placebo: OR 0.98 (0.39, 2.43); vs. placebo: MD −0.42 (−1.32, 0.48)]
Brizel 2000, USA (20)	RCT	Head and neck cancer (Stage II–III)	NR	Intervention (n=153): amifostine 200 mg/m2 IV daily, 15–30 min before irradiation Control (n=150): no treatment (blank) RT regimen: RT: 50–70 Gy, 1.8–2.0 Gy/fraction, q.d.	Grade ≥3 RIOM: not significantly reduced (P=0.48); grade ≥2 xerostomia: 51% vs. 78% (P<0.0001)	Late xerostomia grade ≥2: 34% vs. 57% (P=0.002); median saliva: 0.26 vs. 0.10 g (P=0.04); 2-year locoregional control: 58% vs. 63% [NMA: vs. placebo: OR 0.98 (0.39, 2.43); vs. placebo: MD −0.42 (−1.32, 0.48)]
Antonadou 2002, Greece (21)	RCT	Head and neck cancer (stage III–IV)	NR	Intervention (n=22): amifostine 300 mg/m2 IV, 15–30 min before RT Control (n=23): no treatment (blank) RT regimen: RT: 60–74 Gy, 2 Gy/fraction, 5 fractions/week + cisplatin 90 mg/m2 weekly	Grade ≥3 RIOM: reduced; grade ≥3 dysphagia: reduced; treatment interruption: reduced	Treatment duration shorter in amifostine group (P=0.013); complete response: 90.9% vs. 78.3% (P=0.414) [NMA: vs. placebo: OR 0.98 (0.39, 2.43); vs. placebo: OR 0.08 (0.01, 0.71); vs. placebo: OR 0.04 (0.01, 0.34)]
Haddad 2009, USA (22)	RCT	Locally advanced head and neck cancer	Oropharynx 60%, oral cavity 17%, larynx 10%, unknown primary 7%, other 3%	Intervention (n=29): amifostine 300 mg/m2 IV, 15–30 min before RT for 6–7.5 weeks Control (n=29): no treatment (blank) RT regimen: concomitant CRT: weekly carboplatin/paclitaxel	Grade ≥3 RIOM: reduced; grade ≥2 xerostomia: reduced	Xerostomia: reduced [NMA: vs. placebo: OR 0.98 (0.39, 2.43); vs. placebo: MD −0.42 (−1.32, 0.48)]
Saarilahti 2002, Finland (23)	RCT	Head and neck cancer	NR	Intervention (n=21): GM-CSF mouthwash 150 μg/day (swallowed after rinsing) Control (n=19): sucralfate mouthwash RT regimen: radical RT	Analgesics use: reduced; feeding tube: reduced; treatment interruption: reduced	Mucositis pain (VAS): reduced [NMA: vs. placebo: OR 0.50 (0.04, 6.04); vs. placebo: OR 0.94 (0.57, 1.55); vs. placebo: OR 0.57 (0.18, 1.86)]
Su 2006, USA (24)	RCT	Head and neck cancer	NR	Intervention (n=19): GM-CSF (dose NR) Control (n=22): placebo RT regimen: RT to head and neck	Grade ≥3 RIOM: reduced; grade ≥3 dysphagia: reduced; analgesics: reduced; feeding tube: reduced; treatment interruption: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: OR 0.88 (0.41, 1.91); vs. placebo: OR 0.50 (0.04, 6.04); vs. placebo: OR 0.94 (0.57, 1.55); vs. placebo: OR 0.57 (0.18, 1.86)]
Brizel 2008, Multinational (USA, Canada, Australia) (25)	RCT	HNSCC (advanced)	NR	Intervention (n=69): palifermin 120 μg/kg IV weekly, 3 days before and throughout CRT Control (n=32): placebo RT regimen: concurrent CRT: RT 70 Gy/2 Gy daily or 72 Gy/1.25 Gy b.i.d.; cisplatin 20 mg/m2 × 4 days + 5-FU 1,000 mg/m2/d × 4 days, weeks 1 & 5	Grade ≥3 RIOM: reduced; grade ≥2 xerostomia: reduced; grade ≥3 dysphagia: reduced; analgesics: reduced; feeding tube: reduced; treatment interruption: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: MD −0.18 (−1.08, 0.73); vs. placebo: OR 0.88 (0.41, 1.91); vs. placebo: OR 0.50 (0.04, 6.04); vs. placebo: OR 0.94 (0.57, 1.55); vs. placebo: OR 0.57 (0.18, 1.86)]
McAleese 2006, UK (26)	RCT	Head and neck cancer	Larynx (accelerated radiotherapy of laryngeal cancer)	Intervention (n=15): GM-CSF SC injection (total 2100 μg) Control (n=14): no treatment (blank) RT regimen: RT to head and neck	Grade ≥3 RIOM: reduced; feeding tube: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: OR 0.94 (0.57, 1.55)]
Ryu 2007, South Korea (27)	RCT	Head and neck cancer	NR	Intervention (n=55): palifermin (dose NR) Control (n=53): placebo RT regimen: concurrent CRT	Grade ≥3 RIOM: reduced; feeding tube: reduced; treatment interruption: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: OR 0.94 (0.57, 1.55); vs. placebo: OR 0.57 (0.18, 1.86)]
Henke 2011, Multinational (Europe, USA) (28)	RCT	Resected locally advanced head and neck cancer	Oral cavity, oropharynx, hypopharynx, or larynx (stage II–IVB)	Intervention (n=92): palifermin 120 μg/kg IV weekly from 3 days before and throughout RT Control (n=94): placebo RT regimen: postoperative CRT: RT 60 or 66 Gy, 2 Gy/fraction, 5 fractions/week + cisplatin 100 mg/m2 days 1 & 22	Grade ≥3 RIOM: 51% (41/92) vs. 67% (63/94) (P=0.027); median duration: 4.5 vs. 22.0 days; grade ≥2 xerostomia: reduced; feeding tube: reduced	Time to severe mucositis: 45 vs. 32 days; mortality: similar [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: MD −9.22 (−20.17, 1.73); vs. placebo: MD −0.18 (−1.08, 0.73); vs. placebo: OR 0.94 (0.57, 1.55)]
Le 2011, Multinational (Canada, USA, Europe) (29)	RCT	Locally advanced head and neck cancer	NR	Intervention (n=94): palifermin 120 μg/kg IV weekly from 3 days before and throughout CRT Control (n=94): placebo RT regimen: definitive CRT	Grade ≥3 RIOM: reduced; duration: reduced; grade ≥2 xerostomia: reduced; feeding tube: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: MD −9.22 (−20.17, 1.73); vs. placebo: MD −0.18 (−1.08, 0.73); vs. placebo: OR 0.94 (0.57, 1.55)]
Hoffman 2014, USA (30)	RCT	Head and neck cancer	NR	Intervention (n=53): cytokine (dose NR) Control (n=50): placebo RT regimen: RT or CRT	Grade ≥3 RIOM: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75)]
Shang 2014, China (31)	RCT	Head and neck cancer	NR	Intervention (n=50): cytokine (dose NR) Control (n=25): placebo RT regimen: RT or CRT	Grade ≥3 RIOM: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75)]
Yang 2016, China (32)	RCT	Head and neck cancer	NR	Intervention (n=40): cytokine (dose NR) Control (n=40): placebo RT regimen: RT or CRT	Grade ≥3 RIOM: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75)]
Pan 2017, China (33)	RCT	Head and neck cancer	NR	Intervention (n=39): cytokine (dose NR) Control (n=39): no treatment (blank) RT regimen: RT or CRT	Grade ≥3 RIOM: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75)]
Bao 2018, China (34)	RCT	Head and neck cancer	NR	Intervention (n=32): cytokine (dose NR) Control (n=32): placebo RT regimen: RT or CRT	Grade ≥3 RIOM: reduced; duration: reduced	NR [NMA: vs. placebo: OR 0.42 (0.23, 0.75); vs. placebo: MD −9.22 (−20.17, 1.73)]
Cerchietti 2006, Argentina (35)	RCT	Head and neck cancer	NR	Intervention (n=14): L-alanyl-L-glutamine IV ~30 g/day Control (n=15): placebo RT regimen: CRT	Weight loss: reduced; feeding tube: reduced	NR [NMA: vs. placebo: MD −0.62 (−2.85, 1.60); vs. placebo: OR 0.19 (0.08, 0.47)]
Huang 2000, Taiwan (36)	RCT	Head and neck cancer	NR	Intervention (n=8): glutamine oral suspension (swish and expectorate) twice daily Control (n=9): placebo RT regimen: RT	Grade ≥3 RIOM: reduced; duration: reduced; analgesics: reduced; weight loss: reduced	NR [NMA: vs. placebo: OR 0.13 (0.05, 0.35); vs. placebo: MD −5.14 (−13.19, 2.92); vs. placebo: OR 0.28 (0.03, 2.51); vs. placebo: MD −0.62 (−2.85, 1.60)]
Sarumathy 2012, India (37)	RCT	Head and neck cancer	NR	Intervention (n=11): glutamine oral (dose NR) Control (n=15): no treatment (blank) RT regimen: RT	Duration of grade ≥3 RIOM: reduced	NR [NMA: vs. placebo: MD −5.14 (−13.19, 2.92)]
Chattopadhyay 2014, India (38)	RCT	Head and neck cancer	NR	Intervention (n=35): glutamine oral suspension 10 g in 1,000 mL water, 2 h before RT, daily Control (n=35): no treatment (blank) RT regimen: RT with or without concurrent chemotherapy	Grade 3 mucositis: 14.29% vs. NR (P=0.02); grade 4: 2.86% vs. NR (P=0.04); mean duration grade ≥3: 6.6 vs. 9.2 days (P<0.001)	Time to onset of mucositis: significantly delayed (P<0.001) [NMA: vs. placebo: OR 0.13 (0.05, 0.35); vs. placebo: MD −5.14 (−13.19, 2.92)]
Tsujimoto 2015, Japan (39)	RCT	Locally advanced head and neck cancer	Nasopharynx, oropharynx, hypopharynx or larynx	Intervention (n=20): L-glutamine oral 10 g, 3×/day Control (n=20): placebo 10 g, 3×/day RT regimen: CRT	Grade ≥3 RIOM: reduced; duration: reduced; analgesics: reduced; feeding tube: reduced	NR [NMA: vs. placebo: OR 0.13 (0.05, 0.35); vs. placebo: MD −5.14 (−13.19, 2.92); vs. placebo: OR 0.28 (0.03, 2.51); vs. placebo: OR 0.19 (0.08, 0.47)]
Lopez-Vaquero 2017, Spain (40)	RCT	Head and neck cancer	NR	Intervention (n=25): glutamine oral 10 g, 3×/day Control (n=25): maltodextrin 10 g, 3×/day RT regimen: RT or CRT	Grade ≥3 RIOM: reduced; weight loss: reduced	NR [NMA: vs. placebo: OR 0.13 (0.05, 0.35); vs. placebo: MD −0.62 (−2.85, 1.60)]
Huang 2019, Taiwan (41)	RCT	Head and neck cancer (stage I–IV)	NR	Intervention (n=31): L-glutamine 10 g + maltodextrin 5 g, 3×/day Control (n=33): placebo (maltodextrin 15 g), 3×/day RT regimen: RT	Analgesics: reduced	NR [NMA: vs. placebo: OR 0.28 (0.03, 2.51)]
Khanum 2019, India (42)	RCT	HNSCC	NR	Intervention (n=20): glutamine oral solution 2 h before RT Control (n=20): no treatment (blank) RT regimen: RT: 60–70 Gy, 2 Gy/fraction, 5 fractions/week + concurrent weekly cisplatin 40 mg/m2	Grade ≥3 RIOM: 22% vs. 55% (P=0.006); treatment interruption: reduced	Onset of mucositis delayed; grade 1–2 mucositis: no significant difference [NMA: vs. placebo: OR 0.13 (0.05, 0.35); vs. placebo: OR 0.02 (0.00, 0.10)]
Pathak 2019, India (43)	RCT	Oropharynx and larynx carcinoma (stage III–IV)	Oropharynx and larynx	Intervention (n=30): glutamine oral 10 g, 2 h before RT, 5 days/week Control (n=30): no treatment (blank) RT regimen: concurrent CRT: RT 70 Gy/35 fractions + weekly cisplatin 40 mg/m2	Grade ≥3 RIOM: reduced; grade ≥3 dysphagia: reduced; feeding tube: reduced	Treatment breaks: reduced; weight loss: reduced [NMA: vs. placebo: OR 0.13 (0.05, 0.35); vs. placebo: OR 0.09 (0.02, 0.32); vs. placebo: OR 0.19 (0.08, 0.47)]
Pattanayak 2016, India (44)	RCT	Head and neck cancer	NR	Intervention (n=81): glutamine oral 15 g, 2×/day Control (n=81): no treatment (blank) RT regimen: CRT	Onset of mucositis delayed; none developed grade ≥3 mucositis in glutamine arm vs. 92.35% in control; analgesics: reduced; feeding tube: reduced; treatment interruption: reduced	Pain, dysphagia, nausea, edema, cough: significantly lower in glutamine arm [NMA: vs. placebo: OR 0.28 (0.03, 2.51); vs. placebo: OR 0.19 (0.08, 0.47); vs. placebo: OR 0.02 (0.00, 0.10)]
Lambrecht 2013, Belgium (45)	RCT	Malignant neoplasms of the head and neck	NR	Intervention (n=27): calcium phosphate (Caphosol®) 15 mL rinse, 1 min, 4–10×/day Control (n=31): standard care (magic mouthwash, analgesics, etc.) RT regimen: (Chemo)radiotherapy	Grade ≥3 RIOM: reduced; grade ≥3 dysphagia: reduced	Pain (VAS): reduced [NMA: vs. placebo: OR 0.72 (0.21, 2.49); vs. placebo: OR 0.82 (0.49, 1.37)]
Wong 2017, UK (46)	RCT	Head and neck cancer	NR	Intervention (n=108): calcium phosphate (Caphosol®) 4×/day Control (n=107): standard care RT regimen: RT or CRT	Grade ≥3 RIOM: reduced; duration: reduced; grade ≥3 dysphagia: reduced; weight loss: reduced	Pain: reduced [NMA: vs. placebo: OR 0.72 (0.21, 2.49); vs. placebo: MD −0.70 (−17.23, 15.83); vs. placebo: OR 0.82 (0.49, 1.37); vs. placebo: MD −0.10 (−3.16, 2.96)]

NMA statistics represent the pooled effect of the respective drug class versus placebo from the network meta-analysis. b.i.d., twice daily; CRT, chemoradiotherapy; GM-CSF, granulocyte-macrophage colony-stimulating factor; HNSCC, head and neck squamous cell carcinoma; IV, intravenous; MD, mean difference; NMA, network meta-analysis; NR, not reported; OR, odds ratio; q.d., once daily; RIOM, radiation-induced oral mucositis; RCT, randomized controlled trial; RT, radiotherapy; SC, subcutaneous; VAS, visual analog scale.

Network meta-analysis results

All 29 RCTs provided dichotomous or continuous data on symptom severity or duration at the end of treatment. Network plots of eight outcomes are shown in Figure S1. Line widths represent the number of trials, and circle sizes represent the total number of participants. No trials provided head-to-head comparisons between pharmacological interventions; thus, direct evidence was only available for comparisons with placebo.

Primary outcomes

Number of reported RIOM cases

A total of 24 RCTs comparing the effect of pharmacological interventions on number of reported RIOM cases were included. Glutamine significantly lowered the number of reported cases of RIOM compared to amifostine (OR: 0.13, 95% CI: 0.03–0.51) and cytokine (OR: 0.31, 95% CI: 0.10–0.99). Calcium phosphate (OR: 5.52, 95% CI: 1.13–27.07) and placebo (OR: 7.67, 95% CI: 2.84–20.72) significantly raised number of reported cases of RIOM compared to glutamine (Figure 2). According to the SUCRA results, glutamine had the highest probability (98.7%) of being the best option for minimizing the occurrences of RIOM (Figure 3). The pairwise comparison results showed that glutamine significantly lowered the incidence of RIOM compared to amifostine (OR: 0.13, 95% CI: 0.03–0.51) and cytokine (OR: 0.31, 95% CI: 0.10–0.99). There was no significant global inconsistency (P=0.603) and no significant local inconsistency was found (Tables S1,S2).

Figure 2.

The forest plots of number of reported cases of RIOM of four pharmacological interventions. A: amifostine, B: cytokine, C: glutamine, D: calcium phosphate, E: placebo. CI, confidence interval; PrI, prediction interval; RIOM, radiation-induced oral mucositis.

Figure 3.

SUCRA cumulative rankograms for the eight outcomes, with higher values indicating a higher probability of being the most effective intervention. A: amifostine, B: cytokine, C: glutamine, D: calcium phosphate, E: placebo. SUCRA, surface under the cumulative ranking curve.

Duration of RIOM (days)

A total of 9 RCTs recruiting 825 patients comparing the effect of pharmacological interventions on the duration of RIOM (days) of patients were included. None of the four drugs showed a significant advantage in shortening the duration of RIOM compared with placebo (Figure S2). Treatment hierarchy was ranked as amifostine (90.8%) > cytokine (65.6%) > glutamine (48.7%) > calcium phosphate (28.4%) (Figure 3). The pairwise comparison results showed that no significant difference was seen with all four treatments when compared with placebo. There was no significant global inconsistency (P=0.9336) and no significant local inconsistency was found (all the P values were greater than 0.05) (Tables S1,S2).

Number of reported ≥ grade II xerostomia cases

A total of 6 trials with 966 participants assessed the number of reported ≥ grade II xerostomia cases. Compared with placebo, neither amifostine nor cytokines outperformed placebo (Figure S3). According to the SUCRA results, amifostine (71.9%) showed a higher probability of being the best treatment option regarding the incidence of ≥ grade II xerostomia (Figure 3). The pairwise comparison results showed that no significant differences were seen with amifostine and cytokines compared with placebo. There was no significant global inconsistency (P=0.6989) and no significant local inconsistency was found (all the P values were greater than 0.05) (Tables S1,S2).

Number of reported dysphagia cases

A total of 6 RCTs recruiting 479 patients comparing the effect of pharmacological interventions on the number of reported dysphagia cases were included. Compared with amifostine, glutamine (OR: 11.09, 95% CI: 1.10–112.24), cytokine (OR: 10.28, 95% CI: 1.09–96.79) and placebo (OR: 12.54, 95% CI: 1.41–111.20) significantly increased the number of dysphagia cases. Compared with glutamine, calcium phosphate significantly reduced number of dysphagia cases (OR: 0.10, 95% CI: 0.02–0.45). Compared with calcium phosphate, cytokine (OR: 9.20, 95% CI: 2.30–36.86) and placebo (OR: 11.23, 95% CI: 3.10–40.72) significantly increased the number of dysphagia cases (Figure S4). According to the results of SUCRA (Figure 3), amifostine showed the best efficacy in reducing number of dysphagia cases (87%). There was no significant global inconsistency (P=0.4497) and no significant local inconsistency was found (all the P values were greater than 0.05) (Table S1). The pairwise comparison results showed that significant differences were seen with both amifostine and glutamine compared with calcium phosphate, cytokine, and placebo (1 was not included in the 95% CIs) (Table S2).

Number of patients with analgesics

A total of 7 trials with 485 participants assessed the number of patients with analgesics. Neither cytokine nor glutamine showed significant superiority in reducing the use of analgesics over placebo (Figure S5). Glutamine (74.4%) showed higher probability of becoming the better treatment option than cytokine (53.8%) based on SUCRA (Figure 3). The pairwise comparison results showed that neither glutamine nor cytokine showed a significant difference in reducing the use of analgesics (1 was not included in the 95% CIs) (Table S2). There was no significant global inconsistency (P=0.2563) and no significant local inconsistency was found (all the P values were greater than 0.05) as shown in Table S1.

Secondary outcomes

Weight loss

A total of 4 RCTs with 310 participants assessed the weight loss (kg). Compared with placebo, neither glutamine nor calcium phosphate showed significant advantage of lowering weight loss (Figure S6). Glutamine (68.7%) showed a higher probability to become the better treatment option than calcium phosphate (41%) regarding to weight loss (Figure 3). The pairwise comparison results showed that neither glutamine nor calcium phosphate showed statistically better performance in minimizing weight loss (0 was not included in the 95% CIs) (Table S2). There was no significant global inconsistency (P=0.9490) and no significant local inconsistency was found (all the P values were greater than 0.05), as shown in node splitting analysis provided in Table S1.

Number of patients with feeding tube

A total of 12 trials with 1,020 participants assessed the number of patients with feeding tube. Compared with placebo, glutamine significantly reduced the number of patients with feeding tube (OR: 0.20, 95% CI: 0.07–0.55), while placebo significantly increased the number of patients with feeding tube compared to glutamine (OR: 5.30, 95% CI: 2.13–13.16) (Figure S7). According to the SUCRA results, glutamine (82.3%) showed a higher probability of decreasing the number of patients with feeding tube (Figure 3). The pairwise comparison results showed that glutamine was significantly better than cytokine and placebo in minimizing the use of feeding tube (Table S2). There was significant global inconsistency (P=0.0003) and no significant local inconsistency was found; node splitting analysis provided in Table S1.

Number of treatment interruption

A total of 7 RCTs with 542 participants assessed the number of treatment interruption cases. In terms of reducing the risk of treatment interruption, cytokine (OR:13.01, 95% CI: 1.24–136.69) and placebo (OR: 22.75, 95%CI: 2.98–173.84) significantly increased number of treatment interruption cases compared to amifostine. Glutamine significantly decreased number of treatment interruption cases compared to cytokine (OR: 0.03, 95% CI: 0.00–0.31). While placebo (OR: 54.45, 95% CI: 10.07–294.42) significantly increased number of treatment interruption cases compared to glutamine (Figure S8). Based on the results of SUCRA, treatment hierarchy was ranked as glutamine (91.3%) > amifostine (74.7%) > cytokine (27.9%) in regarding to treatment interruption (Figure 3). This indicates that glutamine had the highest probability of preventing treatment interruption.

There was significant global inconsistency (P<0.0001), and no significant local inconsistency was found. The node-splitting analysis showed no significant local inconsistency, with all P values >0.05 (Table S1). The full pairwise comparison results are summarized in Table S2.

Discussion

In this systematic review and network meta-analysis of 29 RCTs, we evaluated pharmacological interventions for the prevention of RIOM in head and neck cancer patients receiving radiotherapy or chemoradiotherapy. The four agents of interest—amifostine, cytokines, glutamine, and calcium phosphate—target distinct pathogenetic pathways and have never been compared directly in head and neck radiotherapy trials. This is the first NMA to rank these agents across eight clinically relevant outcomes, providing an evidence-based hierarchy for precision prevention and reduces the clinical burden of RIOM. Funnel plots (Figure 4) showed no significant asymmetry, indicating minimal publication bias. The main findings of this network meta-analysis are as follows: (I) glutamine significantly reduced the incidence of severe RIOM (≥ grade III), feeding tube placement, and treatment interruption compared with placebo and several active interventions; (II) amifostine significantly decreased the incidence of severe dysphagia (≥ grade III), outperforming all other agents for this specific outcome; (III) overall, glutamine demonstrated the most favorable risk-benefit profile and is recommended as the preferred pharmacological intervention for RIOM prevention and management.

Figure 4.

Funnel plots for each outcome assessing publication bias across pairwise comparisons of different interventions. A: amifostine; B: cytokine; C: glutamine; D: calcium phosphate; E: placebo

Glutamine’s superior efficacy for severe RIOM, feeding tube placement, and treatment interruption aligns with its established biological role as a primary energy source for rapidly proliferating mucosal epithelial cells and its anti-inflammatory and antioxidant properties (39,47-50). Glutamine serves as a precursor for glutathione synthesis, enhances mucosal barrier function, and modulates inflammatory cytokine production, thereby attenuating radiation-induced mucosal injury (51,52). The oral administration route (10–15 g, two to three times daily; Table 1) enhances clinical practicality and patient compliance. Our findings are consistent with a prior meta-analysis. Leung and Chan reported a significant role for oral glutamine in reducing the risk and severity of severe oral mucositis (risk ratio 0.17; 95% CI: 0.06–0.47) (51). Our NMA extends these findings by establishing glutamine’s relative superiority over amifostine and cytokines, a comparison not possible in previous pairwise meta-analyses.

Amifostine’s selective efficacy for severe dysphagia prevention (SUCRA 87.0%) is consistent with its well-established cytoprotective mechanism via its active thiol metabolite WR-1065, which scavenges free radicals and protects normal tissue from radiation-induced damage (52,53). This finding echoes Sasse et al., who demonstrated that amifostine significantly reduced the risk of dysphagia (OR 0.26, 95% CI: 0.07–0.92; P=0.04) among multiple radiation-related toxicities in a meta-analysis of 14 RCTs (54). However, amifostine’s clinical applicability is limited by adverse effects (transient hypotension, nausea) and the logistical burden of intravenous administration before each radiation fraction (55,56); its use should be reserved for patients at the highest dysphagia risk who can tolerate these drawbacks. However, amifostine showed no significant protective effect on xerostomia, whereas prior meta-analyses reported obvious relief of acute and late xerostomia with this agent (OR 0.24 and 0.33 for acute and late xerostomia, respectively). This discrepancy likely reflects differences in radiation techniques: the predominance of intensity-modulated radiotherapy (IMRT) in more recent trials (included in our NMA) reduces baseline xerostomia risk through parotid gland sparing, potentially creating a ceiling effect that obscures drug benefits.

Across the weighted composite of severe RIOM, feeding-tube placement, and unplanned treatment interruption, glutamine offered the clearest net clinical gain and is the only orally administered agent without additional monitoring—making it the most practical first-line choice for RIOM prophylaxis. Amifostine should be considered as a targeted intervention for patients at highest risk of severe dysphagia, balanced against its side effect profile and administration requirements. Clinicians must weigh patient preference, drug availability, and side-effect profiles when selecting among the four agents.

Several limitations should be acknowledged. First, due to the constraints of the available data in the included studies, our primary outcome assessment was limited to the 7th week of radiotherapy, which was the most consistently reported time point across trials. However, it is clinically well established that radiation-induced oral mucositis (RIOM) typically peaks between the 4th and 5th weeks of treatment. Consequently, our reliance on week-7 data may capture the healing phase rather than the peak severity of mucositis, potentially leading to a conservative underestimation of the true intervention effects. Data on peak severity (weeks 4–5) were either unavailable, inconsistently defined, or missing in a substantial proportion of the original trials, which precluded a more temporally precise analysis within the framework of this network meta-analysis. Future primary studies should standardize the reporting of RIOM severity at multiple predefined time points, including weeks 4–5, to enable more accurate meta-analytic modeling of the peak toxicity period. Second, our network meta-analysis focused exclusively on clinician-graded mucositis severity and could not incorporate patient-reported outcomes (PROs), such as pain scales [visual analog scale (VAS) or numerical rating scale (NRS)] or quality of life (QoL) metrics. These outcomes are critically important for understanding the full clinical burden of RIOM and the patient-perceived benefits of interventions. In addition, future studies should also strengthen research on promising non-pharmacological interventions such as PBM for the management of RIOM (57). Finally, subgroup analyses by tumor site, radiation technique, or chemotherapy regimen were precluded by sparse data and inconsistent reporting.

Conclusions

Based on the network meta-analysis findings, glutamine is recommended as the preferred pharmacological intervention for preventing and alleviating RIOM in head and neck cancer patients, particularly in terms of reducing the risks of severe RIOM, feeding tube dependence, and treatment interruptions. Moreover, amifostine significantly reduced the risk of ≥ grade III dysphagia. Future research should focus on conducting high-quality RCTs with standardized protocols to further confirm the efficacy and safety of glutamine and explore ways to optimize the clinical use of amifostine by improving its drug delivery systems and managing its side effects more effectively. Treatment selection should be individualized, taking into account efficacy and safety outcomes as well as patient, caregiver, and clinician preferences.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.