Eighty-six observational studies (74 photon, five carbon-ion and seven proton) and eight comparative in-silico studies were included in the review. Individual study sample sizes ranged from 10 to 323 participants (total 4,643).
Nasopharyngeal cancer (15 studies, 1,372 participants): No statistically significant differences were observed for any of the outcomes. Pooled analysis for three-year overall survival showed evidence of significant statistical heterogeneity (I2=67.5%).
Oropharyngeal cancer (11 studies, 606 participants): No statistically significant differences were observed for any of the outcomes. Pooled analysis for two-year disease-free survival showed evidence of significant statistical heterogeneity (I2=49.9%).
Paranasal and sinonasal cancer (five studies, 235 participants): Five-year local control after proton therapy was significantly higher compared to intensity modulated photon therapy (88% versus 66%, p=0.035, effect size 0.216, 95% CI 0.025 to 0.407, I2 not reported).
Mucosal malignant melanomas (20 studies, 853 participants): Five-year overall survival was significantly higher after carbon-ion therapy compared to conventional photon therapy (44% versus 25%, p=0.007, effect size 0.185, 95% CI 0.058 to 0.313, I2=51.7%).
Adenoid cystic cancer (25 studies, 1,577 participants): No statistically significant differences were observed for any of the outcomes, but all of the pooled analyses showed evidence of significant statistical heterogeneity (I2=73.6% to 93.7%).
Comparative in-silico studies (eight studies, number of participants not reported): One study compared carbon-ion therapy and photon therapy and seven studies compared proton therapy and photon therapy. One in-silico study suggested that for paranasal sinus tumours, carbon-ion therapy had the ability to statistically significantly decrease the dose to the contralateral optic nerve. One proton therapy study reported a lower dose to the optic nerves compared with IMRT. For patients with nasopharyngeal, oropharyngeal, hypopharyngeal and squamous cell carcinomas, six studies consistently showed a lower dose to the parotid glands for proton therapy than with IMRT.
Toxicity: Toxicity was generally poorly reported but tended to be less frequent in carbon-ion and proton studies than with studies of photon therapy. In-silico studies showed that the organs at risk were exposed to a lower treatment dose that was independent of the tumour site.
The review reported further covariate analyses.