Patients with recurrent or refractory head and neck squamous cell carcinoma received cisplatin/epinephrine injectable gel or placebo gel injected directly into the clinically dominant tumour. The double-blind phase III trial comprised of up to 6 weekly treatments over 8 weeks, 4 weekly evaluation visits, and then monthly follow-up; open-label dosing began as needed after three blinded treatments. Tumour response was defined as complete (100% regression) or partial (50-99% regression) sustained for > or =28 day, and patient benefit as attainment of palliative or preventive goals prospectively selected by investigators and patients. With cisplatin/epinephrine gel, 25% (14 out of 57) of tumours responded (16% complete regression, 9% partial regression), vs 3% (one out of 35, complete regression) with placebo (P=0.007). Patient benefit was positively associated with target tumour response in the blinded period among cisplatin/epinephrine gel recipients (P=0.024): 43% (six out of 14) of responders benefited, vs 12% (five out of 43) of non-responders. The most frequent adverse event was pain during injection and the next most frequent was local cytotoxic effects consistent with the gel's mode of action. Systemic adverse events typical of intravenous cisplatin were uncommon. Intratumoural therapy with cisplatin/epinephrine gel provided safe, well-tolerated, effective palliative treatment for patients with locally advanced head and neck squamous cell carcinoma, who lack other satisfactory treatment options.