BACKGROUND Treatment of cancer is increasingly more effective but is associated with short and long term side effects. Oral side effects remain a major source of illness despite the use of a variety of agents to prevent them. One of these side effects is oral mucositis (mouth ulcers). OBJECTIVE To evaluate the effectiveness of prophylactic agents for oral mucositis in patients with cancer receiving treatment, compared with other potentially active interventions, placebo or no treatment. METHODS The Cochrane Oral Health Group's Trials Register, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE and EMBASE were searched. Reference lists from relevant articles were scanned and the authors of eligible studies were contacted to identify trials and obtain additional information. Date of most recent searches June 2002. METHODS Trials were selected if they met the following criteria: design - random allocation of participants; participants - anyone with cancer receiving chemotherapy or radiotherapy treatment for cancer; interventions - agents prescribed to prevent oral mucositis; outcomes - prevention of mucositis, pain, amount of analgesia, dysphagia, systemic infection, length of hospitalisation, cost and patient quality of life. METHODS Information regarding methods, participants, interventions and outcome measures and results were independently extracted, in duplicate, by two reviewers. Authors were contacted for details of randomisation and withdrawals and a quality assessment was carried out. The Cochrane Oral Health Group statistical guidelines were followed and relative risk values calculated using random effects models. RESULTS One hundred and nine studies were eligible. Fifty-seven were excluded for various reasons, usually as there was no useable information on mucositis. Of the 52 useable studies all had data for mucositis comprising 3594 randomised patients. Interventions evaluated were: acyclovir, allopurinol mouthrinse, amifostine, antibiotic pastille or paste, benzydamine, camomile, chlorhexidine, clarithromycin, folinic acid, glutamine, GM-CSF, hydrolytic enzymes, ice chips, oral care, pentoxifyline, povidone, prednisone, propantheline, prostaglandin, sucralfate and traumeel. Of the 21 interventions included in trials, nine showed some evidence of a benefit (albeit sometimes weak) for either preventing or reducing the severity of mucositis. Interventions where there was more than one trial and a significant difference compared with a placebo or no treatment were allopurinal with unreliable evidence for a reduction in the severity of mucositis OR = 0.01 (95% CI: 0 to 0.03), amifostine provided minimal benefit in preventing mucositis RR = 0.95 (95% CI: 0.91 to 0.99), antibiotic paste or pastille demonstrated a moderate benefit in preventing mucositis RR = 0.87 (95% CI: 0.79 to 0.97), GM-CSF prevented mucositis RR = 0.51 (95% CI: 0.29 to 0.91), hydrolytic enzymes reduced the severity of mucositis RR = 0.49 (95% CI: 0.30 to 0.81), and ice chips prevented mucositis OR = 0.42 (95% CI: 0.19 to 0.93). Other interventions showing some benefit with only one study were: benzydamine, oral care protocols and povidone. The NNT to prevent one patient experiencing mucositis over a baseline incidence of 60% for amifostine is 33 (95% CI: 20 to 100), antibiotic paste or pastille 13 (95% CI: 8 to 50), GM-CSF 3 (95% CI: 2 to 20) and ice chips 5 (95% CI: 2 to 31). When the baseline incidence is 40%/90% the NNTs for amifostine are 50/20, for antibiotic paste or pastille 20/8, for GM-CSF 5/2 and for ice chips 6/10. The general reporting of RCTs was poor. However, the quality of the randomisation improved when the authors provided additional information. CONCLUSIONS Several of the interventions were found to have some benefit at preventing or reducing the severity of mucositis associated with cancer treatment. The strength of the evidence was variable and implications for practice include consideration that benefits may be specific for certain cancer types hat benefits may be specific for certain cancer types and treatment. There is a need for well designed and conducted trials with sufficient numbers of participants to perform subgroup analyses by type of disease and chemotherapeutic agent.